2 * Pixel and vertex shaders implementation using ARB_vertex_program
3 * and ARB_fragment_program GL extensions.
5 * Copyright 2002-2003 Jason Edmeades
6 * Copyright 2002-2003 Raphael Junqueira
7 * Copyright 2004 Christian Costa
8 * Copyright 2005 Oliver Stieber
9 * Copyright 2006 Ivan Gyurdiev
10 * Copyright 2006 Jason Green
11 * Copyright 2006 Henri Verbeet
12 * Copyright 2007-2011, 2013-2014 Stefan Dösinger for CodeWeavers
13 * Copyright 2009 Henri Verbeet for CodeWeavers
15 * This library is free software; you can redistribute it and/or
16 * modify it under the terms of the GNU Lesser General Public
17 * License as published by the Free Software Foundation; either
18 * version 2.1 of the License, or (at your option) any later version.
20 * This library is distributed in the hope that it will be useful,
21 * but WITHOUT ANY WARRANTY; without even the implied warranty of
22 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
23 * Lesser General Public License for more details.
25 * You should have received a copy of the GNU Lesser General Public
26 * License along with this library; if not, write to the Free Software
27 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
31 #include "wine/port.h"
36 #include "wined3d_private.h"
38 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader
);
39 WINE_DECLARE_DEBUG_CHANNEL(d3d_constants
);
40 WINE_DECLARE_DEBUG_CHANNEL(d3d
);
41 WINE_DECLARE_DEBUG_CHANNEL(d3d_perf
);
43 static BOOL
shader_is_pshader_version(enum wined3d_shader_type type
)
45 return type
== WINED3D_SHADER_TYPE_PIXEL
;
48 static BOOL
shader_is_vshader_version(enum wined3d_shader_type type
)
50 return type
== WINED3D_SHADER_TYPE_VERTEX
;
53 static const char *get_line(const char **ptr
)
58 if (!(q
= strstr(p
, "\n")))
80 static const char *arb_get_helper_value(enum wined3d_shader_type shader
, enum arb_helper_value value
)
82 if (shader
!= WINED3D_SHADER_TYPE_VERTEX
&& shader
!= WINED3D_SHADER_TYPE_PIXEL
)
84 ERR("Unsupported shader type '%s'.\n", debug_shader_type(shader
));
88 if (shader
== WINED3D_SHADER_TYPE_PIXEL
)
92 case ARB_ZERO
: return "ps_helper_const.x";
93 case ARB_ONE
: return "ps_helper_const.y";
94 case ARB_TWO
: return "coefmul.x";
95 case ARB_0001
: return "ps_helper_const.xxxy";
96 case ARB_EPS
: return "ps_helper_const.z";
104 case ARB_ZERO
: return "helper_const.x";
105 case ARB_ONE
: return "helper_const.y";
106 case ARB_TWO
: return "helper_const.z";
107 case ARB_EPS
: return "helper_const.w";
108 case ARB_0001
: return "helper_const.xxxy";
109 case ARB_VS_REL_OFFSET
: return "rel_addr_const.y";
112 FIXME("Unmanaged %s shader helper constant requested: %u.\n",
113 shader
== WINED3D_SHADER_TYPE_PIXEL
? "pixel" : "vertex", value
);
116 case ARB_ZERO
: return "0.0";
117 case ARB_ONE
: return "1.0";
118 case ARB_TWO
: return "2.0";
119 case ARB_0001
: return "{0.0, 0.0, 0.0, 1.0}";
120 case ARB_EPS
: return "1e-8";
121 default: return "bad";
125 static inline BOOL
ffp_clip_emul(const struct wined3d_context
*context
)
127 return context
->lowest_disabled_stage
< 7;
130 /* ARB_program_shader private data */
149 struct wined3d_shader_loop_control loop_control
;
153 struct arb_ps_np2fixup_info
155 struct ps_np2fixup_info super
;
156 /* For ARB we need an offset value:
157 * With both GLSL and ARB mode the NP2 fixup information (the texture dimensions) are stored in a
158 * consecutive way (GLSL uses a uniform array). Since ARB doesn't know the notion of a "standalone"
159 * array we need an offset to the index inside the program local parameter array. */
163 struct arb_ps_compile_args
165 struct ps_compile_args super
;
167 WORD clip
; /* only a boolean, use a WORD for alignment */
168 unsigned char loop_ctrl
[WINED3D_MAX_CONSTS_I
][3];
171 struct stb_const_desc
173 unsigned char texunit
;
177 struct arb_ps_compiled_shader
179 struct arb_ps_compile_args args
;
180 struct arb_ps_np2fixup_info np2fixup_info
;
181 struct stb_const_desc bumpenvmatconst
[MAX_TEXTURES
];
182 struct stb_const_desc luminanceconst
[MAX_TEXTURES
];
183 UINT int_consts
[WINED3D_MAX_CONSTS_I
];
186 unsigned char numbumpenvmatconsts
;
190 struct arb_vs_compile_args
192 struct vs_compile_args super
;
198 unsigned char clip_texcoord
;
199 unsigned char clipplane_mask
;
201 DWORD boolclip_compare
;
206 unsigned char samplers
[4];
207 DWORD samplers_compare
;
209 unsigned char loop_ctrl
[WINED3D_MAX_CONSTS_I
][3];
212 struct arb_vs_compiled_shader
214 struct arb_vs_compile_args args
;
216 UINT int_consts
[WINED3D_MAX_CONSTS_I
];
218 char need_color_unclamp
;
222 struct recorded_instruction
224 struct wined3d_shader_instruction ins
;
228 struct shader_arb_ctx_priv
233 /* plain GL_ARB_vertex_program or GL_ARB_fragment_program */
235 /* GL_NV_vertex_program2_option or GL_NV_fragment_program_option */
237 /* GL_NV_vertex_program3 or GL_NV_fragment_program2 */
241 const struct arb_vs_compile_args
*cur_vs_args
;
242 const struct arb_ps_compile_args
*cur_ps_args
;
243 const struct arb_ps_compiled_shader
*compiled_fprog
;
244 const struct arb_vs_compiled_shader
*compiled_vprog
;
245 struct arb_ps_np2fixup_info
*cur_np2fixup_info
;
246 struct list control_frames
;
250 unsigned int num_loops
, loop_depth
, num_ifcs
;
252 BOOL ps_post_process
;
254 unsigned int vs_clipplanes
;
258 /* For 3.0 vertex shaders */
259 const char *vs_output
[MAX_REG_OUTPUT
];
260 /* For 2.x and earlier vertex shaders */
261 const char *texcrd_output
[8], *color_output
[2], *fog_output
;
263 /* 3.0 pshader input for compatibility with fixed function */
264 const char *ps_input
[MAX_REG_INPUT
];
269 struct wined3d_shader_signature sig
;
271 struct wine_rb_entry entry
;
274 struct arb_pshader_private
{
275 struct arb_ps_compiled_shader
*gl_shaders
;
276 UINT num_gl_shaders
, shader_array_size
;
277 DWORD input_signature_idx
;
278 DWORD clipplane_emulation
;
282 struct arb_vshader_private
{
283 struct arb_vs_compiled_shader
*gl_shaders
;
284 UINT num_gl_shaders
, shader_array_size
;
288 struct shader_arb_priv
290 GLuint current_vprogram_id
;
291 GLuint current_fprogram_id
;
292 const struct arb_ps_compiled_shader
*compiled_fprog
;
293 const struct arb_vs_compiled_shader
*compiled_vprog
;
294 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
[WINED3D_MAX_VS_CONSTS_F
];
307 char pshader_const_dirty
[WINED3D_MAX_PS_CONSTS_F
];
308 const struct wined3d_context
*last_context
;
310 const struct wined3d_vertex_pipe_ops
*vertex_pipe
;
311 const struct fragment_pipeline
*fragment_pipe
;
312 BOOL ffp_proj_control
;
315 /* Context activation for state handlers is done by the caller. */
317 static BOOL
need_rel_addr_const(const struct arb_vshader_private
*shader_data
,
318 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
)
320 if (shader_data
->rel_offset
) return TRUE
;
321 if (!reg_maps
->usesmova
) return FALSE
;
322 return !gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
];
325 /* Returns TRUE if result.clip from GL_NV_vertex_program2 should be used and FALSE otherwise */
326 static inline BOOL
use_nv_clip(const struct wined3d_gl_info
*gl_info
)
328 return gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]
329 && !(gl_info
->quirks
& WINED3D_QUIRK_NV_CLIP_BROKEN
);
332 static BOOL
need_helper_const(const struct arb_vshader_private
*shader_data
,
333 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
)
335 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
)) return TRUE
;
336 if (!gl_info
->supported
[NV_VERTEX_PROGRAM
]) return TRUE
; /* Need to init colors. */
337 if (gl_info
->quirks
& WINED3D_QUIRK_ARB_VS_OFFSET_LIMIT
) return TRUE
; /* Load the immval offset. */
338 if (gl_info
->quirks
& WINED3D_QUIRK_SET_TEXCOORD_W
) return TRUE
; /* Have to init texcoords. */
339 if (!use_nv_clip(gl_info
)) return TRUE
; /* Init the clip texcoord */
340 if (reg_maps
->usesnrm
) return TRUE
; /* 0.0 */
341 if (reg_maps
->usespow
) return TRUE
; /* EPS, 0.0 and 1.0 */
342 if (reg_maps
->fog
) return TRUE
; /* Clamping fog coord, 0.0 and 1.0 */
346 static unsigned int reserved_vs_const(const struct arb_vshader_private
*shader_data
,
347 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
)
349 unsigned int ret
= 1;
350 /* We use one PARAM for the pos fixup, and in some cases one to load
351 * some immediate values into the shader. */
352 if (need_helper_const(shader_data
, reg_maps
, gl_info
)) ++ret
;
353 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
)) ++ret
;
357 /* Loads floating point constants into the currently set ARB_vertex/fragment_program.
358 * When constant_list == NULL, it will load all the constants.
360 * @target_type should be either GL_VERTEX_PROGRAM_ARB (for vertex shaders)
361 * or GL_FRAGMENT_PROGRAM_ARB (for pixel shaders)
363 /* Context activation is done by the caller. */
364 static unsigned int shader_arb_load_constants_f(const struct wined3d_shader
*shader
,
365 const struct wined3d_gl_info
*gl_info
, GLuint target_type
, unsigned int max_constants
,
366 const struct wined3d_vec4
*constants
, char *dirty_consts
)
368 struct wined3d_shader_lconst
*lconst
;
369 unsigned int ret
, i
, j
;
371 if (TRACE_ON(d3d_constants
))
373 for (i
= 0; i
< max_constants
; ++i
)
375 if (!dirty_consts
[i
])
377 TRACE_(d3d_constants
)("Loading constant %u: %s.\n", i
, debug_vec4(&constants
[i
]));
383 /* In 1.X pixel shaders constants are implicitly clamped in the range [-1;1] */
384 if (target_type
== GL_FRAGMENT_PROGRAM_ARB
&& shader
->reg_maps
.shader_version
.major
== 1)
387 /* ps 1.x supports only 8 constants, clamp only those. When switching between 1.x and higher
388 * shaders, the first 8 constants are marked dirty for reload
390 for (; i
< min(8, max_constants
); ++i
)
392 if (!dirty_consts
[i
])
396 if (constants
[i
].x
> 1.0f
)
398 else if (constants
[i
].x
< -1.0f
)
399 lcl_const
[0] = -1.0f
;
401 lcl_const
[0] = constants
[i
].x
;
403 if (constants
[i
].y
> 1.0f
)
405 else if (constants
[i
].y
< -1.0f
)
406 lcl_const
[1] = -1.0f
;
408 lcl_const
[1] = constants
[i
].y
;
410 if (constants
[i
].z
> 1.0f
)
412 else if (constants
[i
].z
< -1.0f
)
413 lcl_const
[2] = -1.0f
;
415 lcl_const
[2] = constants
[i
].z
;
417 if (constants
[i
].w
> 1.0f
)
419 else if (constants
[i
].w
< -1.0f
)
420 lcl_const
[3] = -1.0f
;
422 lcl_const
[3] = constants
[i
].w
;
424 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, i
, lcl_const
));
427 /* If further constants are dirty, reload them without clamping.
429 * The alternative is not to touch them, but then we cannot reset the dirty constant count
430 * to zero. That's bad for apps that only use PS 1.x shaders, because in that case the code
431 * above would always re-check the first 8 constants since max_constant remains at the init
436 if (gl_info
->supported
[EXT_GPU_PROGRAM_PARAMETERS
])
438 /* TODO: Benchmark if we're better of with finding the dirty constants ourselves,
439 * or just reloading *all* constants at once
441 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type, i, max_constants, constants + (i * 4)));
443 for (; i
< max_constants
; ++i
)
445 if (!dirty_consts
[i
])
448 /* Find the next block of dirty constants */
451 for (++i
; (i
< max_constants
) && dirty_consts
[i
]; ++i
)
456 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type
, j
, i
- j
, &constants
[j
].x
));
461 for (; i
< max_constants
; ++i
)
466 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, i
, &constants
[i
].x
));
470 checkGLcall("glProgramEnvParameter4fvARB()");
472 /* Load immediate constants */
473 if (shader
->load_local_constsF
)
475 if (TRACE_ON(d3d_shader
))
477 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
479 GLfloat
* values
= (GLfloat
*)lconst
->value
;
480 TRACE_(d3d_constants
)("Loading local constants %i: %f, %f, %f, %f\n", lconst
->idx
,
481 values
[0], values
[1], values
[2], values
[3]);
484 /* Immediate constants are clamped for 1.X shaders at loading times */
486 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
488 dirty_consts
[lconst
->idx
] = 1; /* Dirtify so the non-immediate constant overwrites it next time */
489 ret
= max(ret
, lconst
->idx
+ 1);
490 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, lconst
->idx
, (GLfloat
*)lconst
->value
));
492 checkGLcall("glProgramEnvParameter4fvARB()");
493 return ret
; /* The loaded immediate constants need reloading for the next shader */
495 return 0; /* No constants are dirty now */
499 /* Loads the texture dimensions for NP2 fixup into the currently set
500 * ARB_[vertex/fragment]_programs. */
501 static void shader_arb_load_np2fixup_constants(const struct arb_ps_np2fixup_info
*fixup
,
502 const struct wined3d_gl_info
*gl_info
, const struct wined3d_state
*state
)
504 GLfloat np2fixup_constants
[4 * MAX_FRAGMENT_SAMPLERS
];
505 WORD active
= fixup
->super
.active
;
511 for (i
= 0; active
; active
>>= 1, ++i
)
513 const struct wined3d_texture
*tex
= state
->textures
[i
];
514 unsigned char idx
= fixup
->super
.idx
[i
];
515 GLfloat
*tex_dim
= &np2fixup_constants
[(idx
>> 1) * 4];
522 ERR("Nonexistent texture is flagged for NP2 texcoord fixup.\n");
528 tex_dim
[2] = tex
->pow2_matrix
[0];
529 tex_dim
[3] = tex
->pow2_matrix
[5];
533 tex_dim
[0] = tex
->pow2_matrix
[0];
534 tex_dim
[1] = tex
->pow2_matrix
[5];
538 for (i
= 0; i
< fixup
->super
.num_consts
; ++i
)
540 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
541 fixup
->offset
+ i
, &np2fixup_constants
[i
* 4]));
545 /* Context activation is done by the caller. */
546 static void shader_arb_ps_local_constants(const struct arb_ps_compiled_shader
*gl_shader
,
547 const struct wined3d_context
*context
, const struct wined3d_state
*state
, UINT rt_height
)
549 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
552 for(i
= 0; i
< gl_shader
->numbumpenvmatconsts
; i
++)
554 int texunit
= gl_shader
->bumpenvmatconst
[i
].texunit
;
556 /* The state manager takes care that this function is always called if the bump env matrix changes */
557 const float *data
= (const float *)&state
->texture_states
[texunit
][WINED3D_TSS_BUMPENV_MAT00
];
558 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
559 gl_shader
->bumpenvmatconst
[i
].const_num
, data
));
561 if (gl_shader
->luminanceconst
[i
].const_num
!= WINED3D_CONST_NUM_UNUSED
)
563 /* WINED3D_TSS_BUMPENVLSCALE and WINED3D_TSS_BUMPENVLOFFSET are next to each other.
564 * point gl to the scale, and load 4 floats. x = scale, y = offset, z and w are junk, we
565 * don't care about them. The pointers are valid for sure because the stateblock is bigger.
566 * (they're WINED3D_TSS_TEXTURETRANSFORMFLAGS and WINED3D_TSS_ADDRESSW, so most likely 0 or NaN
568 const float *scale
= (const float *)&state
->texture_states
[texunit
][WINED3D_TSS_BUMPENV_LSCALE
];
569 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
570 gl_shader
->luminanceconst
[i
].const_num
, scale
));
573 checkGLcall("Load bumpmap consts");
575 if(gl_shader
->ycorrection
!= WINED3D_CONST_NUM_UNUSED
)
577 /* ycorrection.x: Backbuffer height(onscreen) or 0(offscreen).
578 * ycorrection.y: -1.0(onscreen), 1.0(offscreen)
583 val
[0] = context
->render_offscreen
? 0.0f
: (float) rt_height
;
584 val
[1] = context
->render_offscreen
? 1.0f
: -1.0f
;
587 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, gl_shader
->ycorrection
, val
));
588 checkGLcall("y correction loading");
591 if (!gl_shader
->num_int_consts
) return;
593 for (i
= 0; i
< WINED3D_MAX_CONSTS_I
; ++i
)
595 if(gl_shader
->int_consts
[i
] != WINED3D_CONST_NUM_UNUSED
)
598 val
[0] = (float)state
->ps_consts_i
[i
].x
;
599 val
[1] = (float)state
->ps_consts_i
[i
].y
;
600 val
[2] = (float)state
->ps_consts_i
[i
].z
;
603 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, gl_shader
->int_consts
[i
], val
));
606 checkGLcall("Load ps int consts");
609 /* Context activation is done by the caller. */
610 static void shader_arb_vs_local_constants(const struct arb_vs_compiled_shader
*gl_shader
,
611 const struct wined3d_context
*context
, const struct wined3d_state
*state
)
613 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
614 float position_fixup
[4];
617 /* Upload the position fixup */
618 shader_get_position_fixup(context
, state
, position_fixup
);
619 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, gl_shader
->pos_fixup
, position_fixup
));
621 if (!gl_shader
->num_int_consts
) return;
623 for (i
= 0; i
< WINED3D_MAX_CONSTS_I
; ++i
)
625 if(gl_shader
->int_consts
[i
] != WINED3D_CONST_NUM_UNUSED
)
628 val
[0] = (float)state
->vs_consts_i
[i
].x
;
629 val
[1] = (float)state
->vs_consts_i
[i
].y
;
630 val
[2] = (float)state
->vs_consts_i
[i
].z
;
633 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, gl_shader
->int_consts
[i
], val
));
636 checkGLcall("Load vs int consts");
639 static void shader_arb_select(void *shader_priv
, struct wined3d_context
*context
,
640 const struct wined3d_state
*state
);
643 * Loads the app-supplied constants into the currently set ARB_[vertex/fragment]_programs.
645 * We only support float constants in ARB at the moment, so don't
646 * worry about the Integers or Booleans
648 /* Context activation is done by the caller (state handler). */
649 static void shader_arb_load_constants_internal(struct shader_arb_priv
*priv
,
650 struct wined3d_context
*context
, const struct wined3d_state
*state
,
651 BOOL usePixelShader
, BOOL useVertexShader
, BOOL from_shader_select
)
653 const struct wined3d_d3d_info
*d3d_info
= context
->d3d_info
;
654 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
656 if (!from_shader_select
)
658 const struct wined3d_shader
*vshader
= state
->shader
[WINED3D_SHADER_TYPE_VERTEX
];
659 const struct wined3d_shader
*pshader
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
662 && (vshader
->reg_maps
.boolean_constants
663 || (!gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]
664 && (vshader
->reg_maps
.integer_constants
& ~vshader
->reg_maps
.local_int_consts
))))
666 TRACE("bool/integer vertex shader constants potentially modified, forcing shader reselection.\n");
667 shader_arb_select(priv
, context
, state
);
670 && (pshader
->reg_maps
.boolean_constants
671 || (!gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
]
672 && (pshader
->reg_maps
.integer_constants
& ~pshader
->reg_maps
.local_int_consts
))))
674 TRACE("bool/integer pixel shader constants potentially modified, forcing shader reselection.\n");
675 shader_arb_select(priv
, context
, state
);
679 if (context
!= priv
->last_context
)
681 memset(priv
->vshader_const_dirty
, 1,
682 sizeof(*priv
->vshader_const_dirty
) * d3d_info
->limits
.vs_uniform_count
);
683 priv
->highest_dirty_vs_const
= d3d_info
->limits
.vs_uniform_count
;
685 memset(priv
->pshader_const_dirty
, 1,
686 sizeof(*priv
->pshader_const_dirty
) * d3d_info
->limits
.ps_uniform_count
);
687 priv
->highest_dirty_ps_const
= d3d_info
->limits
.ps_uniform_count
;
689 priv
->last_context
= context
;
694 const struct wined3d_shader
*vshader
= state
->shader
[WINED3D_SHADER_TYPE_VERTEX
];
695 const struct arb_vs_compiled_shader
*gl_shader
= priv
->compiled_vprog
;
697 /* Load DirectX 9 float constants for vertex shader */
698 priv
->highest_dirty_vs_const
= shader_arb_load_constants_f(vshader
, gl_info
, GL_VERTEX_PROGRAM_ARB
,
699 priv
->highest_dirty_vs_const
, state
->vs_consts_f
, priv
->vshader_const_dirty
);
700 shader_arb_vs_local_constants(gl_shader
, context
, state
);
705 const struct wined3d_shader
*pshader
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
706 const struct arb_ps_compiled_shader
*gl_shader
= priv
->compiled_fprog
;
707 UINT rt_height
= state
->fb
->render_targets
[0]->height
;
709 /* Load DirectX 9 float constants for pixel shader */
710 priv
->highest_dirty_ps_const
= shader_arb_load_constants_f(pshader
, gl_info
, GL_FRAGMENT_PROGRAM_ARB
,
711 priv
->highest_dirty_ps_const
, state
->ps_consts_f
, priv
->pshader_const_dirty
);
712 shader_arb_ps_local_constants(gl_shader
, context
, state
, rt_height
);
714 if (context
->constant_update_mask
& WINED3D_SHADER_CONST_PS_NP2_FIXUP
)
715 shader_arb_load_np2fixup_constants(&gl_shader
->np2fixup_info
, gl_info
, state
);
719 static void shader_arb_load_constants(void *shader_priv
, struct wined3d_context
*context
,
720 const struct wined3d_state
*state
)
722 BOOL vs
= use_vs(state
);
723 BOOL ps
= use_ps(state
);
725 shader_arb_load_constants_internal(shader_priv
, context
, state
, ps
, vs
, FALSE
);
728 static void shader_arb_update_float_vertex_constants(struct wined3d_device
*device
, UINT start
, UINT count
)
730 struct wined3d_context
*context
= context_get_current();
731 struct shader_arb_priv
*priv
= device
->shader_priv
;
733 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
734 * context. On a context switch the old context will be fully dirtified */
735 if (!context
|| context
->device
!= device
)
738 memset(priv
->vshader_const_dirty
+ start
, 1, sizeof(*priv
->vshader_const_dirty
) * count
);
739 priv
->highest_dirty_vs_const
= max(priv
->highest_dirty_vs_const
, start
+ count
);
742 static void shader_arb_update_float_pixel_constants(struct wined3d_device
*device
, UINT start
, UINT count
)
744 struct wined3d_context
*context
= context_get_current();
745 struct shader_arb_priv
*priv
= device
->shader_priv
;
747 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
748 * context. On a context switch the old context will be fully dirtified */
749 if (!context
|| context
->device
!= device
)
752 memset(priv
->pshader_const_dirty
+ start
, 1, sizeof(*priv
->pshader_const_dirty
) * count
);
753 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, start
+ count
);
756 static void shader_arb_append_imm_vec4(struct wined3d_string_buffer
*buffer
, const float *values
)
760 wined3d_ftoa(values
[0], str
[0]);
761 wined3d_ftoa(values
[1], str
[1]);
762 wined3d_ftoa(values
[2], str
[2]);
763 wined3d_ftoa(values
[3], str
[3]);
764 shader_addline(buffer
, "{%s, %s, %s, %s}", str
[0], str
[1], str
[2], str
[3]);
767 /* Generate the variable & register declarations for the ARB_vertex_program output target */
768 static void shader_generate_arb_declarations(const struct wined3d_shader
*shader
,
769 const struct wined3d_shader_reg_maps
*reg_maps
, struct wined3d_string_buffer
*buffer
,
770 const struct wined3d_gl_info
*gl_info
, DWORD
*num_clipplanes
,
771 const struct shader_arb_ctx_priv
*ctx
)
774 char pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
775 const struct wined3d_shader_lconst
*lconst
;
776 unsigned max_constantsF
;
779 /* In pixel shaders, all private constants are program local, we don't need anything
780 * from program.env. Thus we can advertise the full set of constants in pixel shaders.
781 * If we need a private constant the GL implementation will squeeze it in somewhere
783 * With vertex shaders we need the posFixup and on some GL implementations 4 helper
784 * immediate values. The posFixup is loaded using program.env for now, so always
785 * subtract one from the number of constants. If the shader uses indirect addressing,
786 * account for the helper const too because we have to declare all available d3d constants
787 * and don't know which are actually used.
791 max_constantsF
= gl_info
->limits
.arb_ps_native_constants
;
792 /* 24 is the minimum MAX_PROGRAM_ENV_PARAMETERS_ARB value. */
793 if (max_constantsF
< 24)
794 max_constantsF
= gl_info
->limits
.arb_ps_float_constants
;
798 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
799 max_constantsF
= gl_info
->limits
.arb_vs_native_constants
;
800 /* 96 is the minimum MAX_PROGRAM_ENV_PARAMETERS_ARB value.
801 * Also prevents max_constantsF from becoming less than 0 and
803 if (max_constantsF
< 96)
804 max_constantsF
= gl_info
->limits
.arb_vs_float_constants
;
806 if (reg_maps
->usesrelconstF
)
808 DWORD highest_constf
= 0, clip_limit
;
810 max_constantsF
-= reserved_vs_const(shader_data
, reg_maps
, gl_info
);
811 max_constantsF
-= wined3d_popcount(reg_maps
->integer_constants
);
812 max_constantsF
-= gl_info
->reserved_arb_constants
;
814 for (i
= 0; i
< shader
->limits
->constant_float
; ++i
)
817 DWORD shift
= i
& 0x1f;
818 if (reg_maps
->constf
[idx
] & (1u << shift
))
822 if(use_nv_clip(gl_info
) && ctx
->target_version
>= NV2
)
824 if(ctx
->cur_vs_args
->super
.clip_enabled
)
825 clip_limit
= gl_info
->limits
.user_clip_distances
;
831 unsigned int mask
= ctx
->cur_vs_args
->clip
.boolclip
.clipplane_mask
;
832 clip_limit
= min(wined3d_popcount(mask
), 4);
834 *num_clipplanes
= min(clip_limit
, max_constantsF
- highest_constf
- 1);
835 max_constantsF
-= *num_clipplanes
;
836 if(*num_clipplanes
< clip_limit
)
838 WARN("Only %u clip planes out of %u enabled.\n", *num_clipplanes
,
839 gl_info
->limits
.user_clip_distances
);
844 if (ctx
->target_version
>= NV2
)
845 *num_clipplanes
= gl_info
->limits
.user_clip_distances
;
847 *num_clipplanes
= min(gl_info
->limits
.user_clip_distances
, 4);
851 for (i
= 0, map
= reg_maps
->temporary
; map
; map
>>= 1, ++i
)
853 if (map
& 1) shader_addline(buffer
, "TEMP R%u;\n", i
);
856 for (i
= 0, map
= reg_maps
->address
; map
; map
>>= 1, ++i
)
858 if (map
& 1) shader_addline(buffer
, "ADDRESS A%u;\n", i
);
861 if (pshader
&& reg_maps
->shader_version
.major
== 1 && reg_maps
->shader_version
.minor
<= 3)
863 for (i
= 0, map
= reg_maps
->texcoord
; map
; map
>>= 1, ++i
)
865 if (map
& 1) shader_addline(buffer
, "TEMP T%u;\n", i
);
869 if (!shader
->load_local_constsF
)
871 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
874 value
= (const float *)lconst
->value
;
875 shader_addline(buffer
, "PARAM C%u = ", lconst
->idx
);
876 shader_arb_append_imm_vec4(buffer
, value
);
877 shader_addline(buffer
, ";\n");
881 /* After subtracting privately used constants from the hardware limit(they are loaded as
882 * local constants), make sure the shader doesn't violate the env constant limit
886 max_constantsF
= min(max_constantsF
, gl_info
->limits
.arb_ps_float_constants
);
890 max_constantsF
= min(max_constantsF
, gl_info
->limits
.arb_vs_float_constants
);
893 /* Avoid declaring more constants than needed */
894 max_constantsF
= min(max_constantsF
, shader
->limits
->constant_float
);
896 /* we use the array-based constants array if the local constants are marked for loading,
897 * because then we use indirect addressing, or when the local constant list is empty,
898 * because then we don't know if we're using indirect addressing or not. If we're hardcoding
899 * local constants do not declare the loaded constants as an array because ARB compilers usually
900 * do not optimize unused constants away
902 if (reg_maps
->usesrelconstF
)
904 /* Need to PARAM the environment parameters (constants) so we can use relative addressing */
905 shader_addline(buffer
, "PARAM C[%d] = { program.env[0..%d] };\n",
906 max_constantsF
, max_constantsF
- 1);
910 for (i
= 0; i
< max_constantsF
; ++i
)
914 mask
= 1u << (i
& 0x1fu
);
915 if (!shader_constant_is_local(shader
, i
) && (reg_maps
->constf
[idx
] & mask
))
917 shader_addline(buffer
, "PARAM C%d = program.env[%d];\n",i
, i
);
923 static const char * const shift_tab
[] = {
924 "dummy", /* 0 (none) */
925 "coefmul.x", /* 1 (x2) */
926 "coefmul.y", /* 2 (x4) */
927 "coefmul.z", /* 3 (x8) */
928 "coefmul.w", /* 4 (x16) */
929 "dummy", /* 5 (x32) */
930 "dummy", /* 6 (x64) */
931 "dummy", /* 7 (x128) */
932 "dummy", /* 8 (d256) */
933 "dummy", /* 9 (d128) */
934 "dummy", /* 10 (d64) */
935 "dummy", /* 11 (d32) */
936 "coefdiv.w", /* 12 (d16) */
937 "coefdiv.z", /* 13 (d8) */
938 "coefdiv.y", /* 14 (d4) */
939 "coefdiv.x" /* 15 (d2) */
942 static void shader_arb_get_write_mask(const struct wined3d_shader_instruction
*ins
,
943 const struct wined3d_shader_dst_param
*dst
, char *write_mask
)
945 char *ptr
= write_mask
;
947 if (dst
->write_mask
!= WINED3DSP_WRITEMASK_ALL
)
950 if (dst
->write_mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
951 if (dst
->write_mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
952 if (dst
->write_mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
953 if (dst
->write_mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
959 static void shader_arb_get_swizzle(const struct wined3d_shader_src_param
*param
, BOOL fixup
, char *swizzle_str
)
961 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
962 * but addressed as "rgba". To fix this we need to swap the register's x
963 * and z components. */
964 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
965 char *ptr
= swizzle_str
;
967 /* swizzle bits fields: wwzzyyxx */
968 DWORD swizzle
= param
->swizzle
;
969 DWORD swizzle_x
= swizzle
& 0x03;
970 DWORD swizzle_y
= (swizzle
>> 2) & 0x03;
971 DWORD swizzle_z
= (swizzle
>> 4) & 0x03;
972 DWORD swizzle_w
= (swizzle
>> 6) & 0x03;
974 /* If the swizzle is the default swizzle (ie, "xyzw"), we don't need to
975 * generate a swizzle string. Unless we need to our own swizzling. */
976 if (swizzle
!= WINED3DSP_NOSWIZZLE
|| fixup
)
979 if (swizzle_x
== swizzle_y
&& swizzle_x
== swizzle_z
&& swizzle_x
== swizzle_w
) {
980 *ptr
++ = swizzle_chars
[swizzle_x
];
982 *ptr
++ = swizzle_chars
[swizzle_x
];
983 *ptr
++ = swizzle_chars
[swizzle_y
];
984 *ptr
++ = swizzle_chars
[swizzle_z
];
985 *ptr
++ = swizzle_chars
[swizzle_w
];
992 static void shader_arb_request_a0(const struct wined3d_shader_instruction
*ins
, const char *src
)
994 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
995 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
997 if (!strcmp(priv
->addr_reg
, src
)) return;
999 strcpy(priv
->addr_reg
, src
);
1000 shader_addline(buffer
, "ARL A0.x, %s;\n", src
);
1003 static void shader_arb_get_src_param(const struct wined3d_shader_instruction
*ins
,
1004 const struct wined3d_shader_src_param
*src
, unsigned int tmpreg
, char *outregstr
);
1006 static void shader_arb_get_register_name(const struct wined3d_shader_instruction
*ins
,
1007 const struct wined3d_shader_register
*reg
, char *register_name
, BOOL
*is_color
)
1009 /* oPos, oFog and oPts in D3D */
1010 static const char * const rastout_reg_names
[] = {"TMP_OUT", "TMP_FOGCOORD", "result.pointsize"};
1011 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
1012 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
1013 BOOL pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
1014 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1020 case WINED3DSPR_TEMP
:
1021 sprintf(register_name
, "R%u", reg
->idx
[0].offset
);
1024 case WINED3DSPR_INPUT
:
1027 if (reg_maps
->shader_version
.major
< 3)
1029 if (!reg
->idx
[0].offset
)
1030 strcpy(register_name
, "fragment.color.primary");
1032 strcpy(register_name
, "fragment.color.secondary");
1036 if (reg
->idx
[0].rel_addr
)
1039 shader_arb_get_src_param(ins
, reg
->idx
[0].rel_addr
, 0, rel_reg
);
1041 if (!strcmp(rel_reg
, "**aL_emul**"))
1043 DWORD idx
= ctx
->aL
+ reg
->idx
[0].offset
;
1044 if(idx
< MAX_REG_INPUT
)
1046 strcpy(register_name
, ctx
->ps_input
[idx
]);
1050 ERR("Pixel shader input register out of bounds: %u\n", idx
);
1051 sprintf(register_name
, "out_of_bounds_%u", idx
);
1054 else if (reg_maps
->input_registers
& 0x0300)
1056 /* There are two ways basically:
1058 * 1) Use the unrolling code that is used for loop emulation and unroll the loop.
1059 * That means trouble if the loop also contains a breakc or if the control values
1060 * aren't local constants.
1061 * 2) Generate an if block that checks if aL.y < 8, == 8 or == 9 and selects the
1062 * source dynamically. The trouble is that we cannot simply read aL.y because it
1063 * is an ADDRESS register. We could however push it, load .zw with a value and use
1064 * ADAC to load the condition code register and pop it again afterwards
1066 FIXME("Relative input register addressing with more than 8 registers\n");
1068 /* This is better than nothing for now */
1069 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1071 else if(ctx
->cur_ps_args
->super
.vp_mode
!= vertexshader
)
1073 /* This is problematic because we'd have to consult the ctx->ps_input strings
1074 * for where to find the varying. Some may be "0.0", others can be texcoords or
1075 * colors. This needs either a pipeline replacement to make the vertex shader feed
1076 * proper varyings, or loop unrolling
1078 * For now use the texcoords and hope for the best
1080 FIXME("Non-vertex shader varying input with indirect addressing\n");
1081 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1085 /* D3D supports indirect addressing only with aL in loop registers. The loop instruction
1086 * pulls GL_NV_fragment_program2 in
1088 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1093 if (reg
->idx
[0].offset
< MAX_REG_INPUT
)
1095 strcpy(register_name
, ctx
->ps_input
[reg
->idx
[0].offset
]);
1099 ERR("Pixel shader input register out of bounds: %u\n", reg
->idx
[0].offset
);
1100 sprintf(register_name
, "out_of_bounds_%u", reg
->idx
[0].offset
);
1107 if (ctx
->cur_vs_args
->super
.swizzle_map
& (1u << reg
->idx
[0].offset
))
1109 sprintf(register_name
, "vertex.attrib[%u]", reg
->idx
[0].offset
);
1113 case WINED3DSPR_CONST
:
1114 if (!pshader
&& reg
->idx
[0].rel_addr
)
1116 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1117 UINT rel_offset
= ctx
->target_version
== ARB
? shader_data
->rel_offset
: 0;
1120 if (reg_maps
->shader_version
.major
< 2)
1122 sprintf(rel_reg
, "A0.x");
1126 shader_arb_get_src_param(ins
, reg
->idx
[0].rel_addr
, 0, rel_reg
);
1127 if (ctx
->target_version
== ARB
)
1129 if (!strcmp(rel_reg
, "**aL_emul**"))
1133 shader_arb_request_a0(ins
, rel_reg
);
1134 sprintf(rel_reg
, "A0.x");
1139 sprintf(register_name
, "C[%u]", ctx
->aL
+ reg
->idx
[0].offset
);
1140 else if (reg
->idx
[0].offset
>= rel_offset
)
1141 sprintf(register_name
, "C[%s + %u]", rel_reg
, reg
->idx
[0].offset
- rel_offset
);
1143 sprintf(register_name
, "C[%s - %u]", rel_reg
, rel_offset
- reg
->idx
[0].offset
);
1147 if (reg_maps
->usesrelconstF
)
1148 sprintf(register_name
, "C[%u]", reg
->idx
[0].offset
);
1150 sprintf(register_name
, "C%u", reg
->idx
[0].offset
);
1154 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
1157 if (reg_maps
->shader_version
.major
== 1
1158 && reg_maps
->shader_version
.minor
<= 3)
1159 /* In ps <= 1.3, Tx is a temporary register as destination
1160 * to all instructions, and as source to most instructions.
1161 * For some instructions it is the texcoord input. Those
1162 * instructions know about the special use. */
1163 sprintf(register_name
, "T%u", reg
->idx
[0].offset
);
1165 /* In ps 1.4 and 2.x Tx is always a (read-only) varying. */
1166 sprintf(register_name
, "fragment.texcoord[%u]", reg
->idx
[0].offset
);
1170 if (reg_maps
->shader_version
.major
== 1 || ctx
->target_version
>= NV2
)
1171 sprintf(register_name
, "A%u", reg
->idx
[0].offset
);
1173 sprintf(register_name
, "A%u_SHADOW", reg
->idx
[0].offset
);
1177 case WINED3DSPR_COLOROUT
:
1178 if (ctx
->ps_post_process
&& !reg
->idx
[0].offset
)
1180 strcpy(register_name
, "TMP_COLOR");
1184 if (ctx
->cur_ps_args
->super
.srgb_correction
)
1185 FIXME("sRGB correction on higher render targets.\n");
1186 if (reg_maps
->rt_mask
> 1)
1187 sprintf(register_name
, "result.color[%u]", reg
->idx
[0].offset
);
1189 strcpy(register_name
, "result.color");
1193 case WINED3DSPR_RASTOUT
:
1194 if (reg
->idx
[0].offset
== 1)
1195 sprintf(register_name
, "%s", ctx
->fog_output
);
1197 sprintf(register_name
, "%s", rastout_reg_names
[reg
->idx
[0].offset
]);
1200 case WINED3DSPR_DEPTHOUT
:
1201 strcpy(register_name
, "result.depth");
1204 case WINED3DSPR_ATTROUT
:
1205 /* case WINED3DSPR_OUTPUT: */
1207 sprintf(register_name
, "oD[%u]", reg
->idx
[0].offset
);
1209 strcpy(register_name
, ctx
->color_output
[reg
->idx
[0].offset
]);
1212 case WINED3DSPR_TEXCRDOUT
:
1214 sprintf(register_name
, "oT[%u]", reg
->idx
[0].offset
);
1215 else if (reg_maps
->shader_version
.major
< 3)
1216 strcpy(register_name
, ctx
->texcrd_output
[reg
->idx
[0].offset
]);
1218 strcpy(register_name
, ctx
->vs_output
[reg
->idx
[0].offset
]);
1221 case WINED3DSPR_LOOP
:
1222 if(ctx
->target_version
>= NV2
)
1224 /* Pshader has an implicitly declared loop index counter A0.x that cannot be renamed */
1225 if(pshader
) sprintf(register_name
, "A0.x");
1226 else sprintf(register_name
, "aL.y");
1230 /* Unfortunately this code cannot return the value of ctx->aL here. An immediate value
1231 * would be valid, but if aL is used for indexing(its only use), there's likely an offset,
1232 * thus the result would be something like C[15 + 30], which is not valid in the ARB program
1233 * grammar. So return a marker for the emulated aL and intercept it in constant and varying
1236 sprintf(register_name
, "**aL_emul**");
1241 case WINED3DSPR_CONSTINT
:
1242 sprintf(register_name
, "I%u", reg
->idx
[0].offset
);
1245 case WINED3DSPR_MISCTYPE
:
1246 if (!reg
->idx
[0].offset
)
1247 sprintf(register_name
, "vpos");
1248 else if (reg
->idx
[0].offset
== 1)
1249 sprintf(register_name
, "fragment.facing.x");
1251 FIXME("Unknown MISCTYPE register index %u.\n", reg
->idx
[0].offset
);
1255 FIXME("Unhandled register type %#x[%u].\n", reg
->type
, reg
->idx
[0].offset
);
1256 sprintf(register_name
, "unrecognized_register[%u]", reg
->idx
[0].offset
);
1261 static void shader_arb_get_dst_param(const struct wined3d_shader_instruction
*ins
,
1262 const struct wined3d_shader_dst_param
*wined3d_dst
, char *str
)
1264 char register_name
[255];
1268 shader_arb_get_register_name(ins
, &wined3d_dst
->reg
, register_name
, &is_color
);
1269 strcpy(str
, register_name
);
1271 shader_arb_get_write_mask(ins
, wined3d_dst
, write_mask
);
1272 strcat(str
, write_mask
);
1275 static const char *shader_arb_get_fixup_swizzle(enum fixup_channel_source channel_source
)
1277 switch(channel_source
)
1279 case CHANNEL_SOURCE_ZERO
: return "0";
1280 case CHANNEL_SOURCE_ONE
: return "1";
1281 case CHANNEL_SOURCE_X
: return "x";
1282 case CHANNEL_SOURCE_Y
: return "y";
1283 case CHANNEL_SOURCE_Z
: return "z";
1284 case CHANNEL_SOURCE_W
: return "w";
1286 FIXME("Unhandled channel source %#x\n", channel_source
);
1291 struct color_fixup_masks
1297 static struct color_fixup_masks
calc_color_correction(struct color_fixup_desc fixup
, DWORD dst_mask
)
1299 struct color_fixup_masks masks
= {0, 0};
1301 if (is_complex_fixup(fixup
))
1303 enum complex_fixup complex_fixup
= get_complex_fixup(fixup
);
1304 FIXME("Complex fixup (%#x) not supported\n", complex_fixup
);
1308 if (fixup
.x_source
!= CHANNEL_SOURCE_X
)
1309 masks
.source
|= WINED3DSP_WRITEMASK_0
;
1310 if (fixup
.y_source
!= CHANNEL_SOURCE_Y
)
1311 masks
.source
|= WINED3DSP_WRITEMASK_1
;
1312 if (fixup
.z_source
!= CHANNEL_SOURCE_Z
)
1313 masks
.source
|= WINED3DSP_WRITEMASK_2
;
1314 if (fixup
.w_source
!= CHANNEL_SOURCE_W
)
1315 masks
.source
|= WINED3DSP_WRITEMASK_3
;
1316 masks
.source
&= dst_mask
;
1318 if (fixup
.x_sign_fixup
)
1319 masks
.sign
|= WINED3DSP_WRITEMASK_0
;
1320 if (fixup
.y_sign_fixup
)
1321 masks
.sign
|= WINED3DSP_WRITEMASK_1
;
1322 if (fixup
.z_sign_fixup
)
1323 masks
.sign
|= WINED3DSP_WRITEMASK_2
;
1324 if (fixup
.w_sign_fixup
)
1325 masks
.sign
|= WINED3DSP_WRITEMASK_3
;
1326 masks
.sign
&= dst_mask
;
1331 static void gen_color_correction(struct wined3d_string_buffer
*buffer
, const char *dst
,
1332 const char *src
, const char *one
, const char *two
,
1333 struct color_fixup_desc fixup
, struct color_fixup_masks masks
)
1335 const char *sign_fixup_src
= dst
;
1340 sign_fixup_src
= "TA";
1342 shader_addline(buffer
, "SWZ %s, %s, %s, %s, %s, %s;\n", sign_fixup_src
, src
,
1343 shader_arb_get_fixup_swizzle(fixup
.x_source
), shader_arb_get_fixup_swizzle(fixup
.y_source
),
1344 shader_arb_get_fixup_swizzle(fixup
.z_source
), shader_arb_get_fixup_swizzle(fixup
.w_source
));
1346 else if (masks
.sign
)
1348 sign_fixup_src
= src
;
1354 char *ptr
= reg_mask
;
1356 if (masks
.sign
!= WINED3DSP_WRITEMASK_ALL
)
1359 if (masks
.sign
& WINED3DSP_WRITEMASK_0
)
1361 if (masks
.sign
& WINED3DSP_WRITEMASK_1
)
1363 if (masks
.sign
& WINED3DSP_WRITEMASK_2
)
1365 if (masks
.sign
& WINED3DSP_WRITEMASK_3
)
1370 shader_addline(buffer
, "MAD %s%s, %s, %s, -%s;\n", dst
, reg_mask
, sign_fixup_src
, two
, one
);
1374 static const char *shader_arb_get_modifier(const struct wined3d_shader_instruction
*ins
)
1377 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1378 if (!ins
->dst_count
) return "";
1380 mod
= ins
->dst
[0].modifiers
;
1382 /* Silently ignore PARTIALPRECISION if it's not supported */
1383 if(priv
->target_version
== ARB
) mod
&= ~WINED3DSPDM_PARTIALPRECISION
;
1385 if(mod
& WINED3DSPDM_MSAMPCENTROID
)
1387 FIXME("Unhandled modifier WINED3DSPDM_MSAMPCENTROID\n");
1388 mod
&= ~WINED3DSPDM_MSAMPCENTROID
;
1393 case WINED3DSPDM_SATURATE
| WINED3DSPDM_PARTIALPRECISION
:
1396 case WINED3DSPDM_SATURATE
:
1399 case WINED3DSPDM_PARTIALPRECISION
:
1406 FIXME("Unknown modifiers 0x%08x\n", mod
);
1411 #define TEX_PROJ 0x1
1412 #define TEX_BIAS 0x2
1414 #define TEX_DERIV 0x10
1416 static void shader_hw_sample(const struct wined3d_shader_instruction
*ins
, DWORD sampler_idx
,
1417 const char *dst_str
, const char *coord_reg
, WORD flags
, const char *dsx
, const char *dsy
)
1419 enum wined3d_shader_resource_type resource_type
= ins
->ctx
->reg_maps
->resource_info
[sampler_idx
].type
;
1420 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1421 const char *tex_type
;
1422 BOOL np2_fixup
= FALSE
;
1423 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1425 BOOL pshader
= shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
1426 const struct wined3d_shader
*shader
;
1427 const struct wined3d_device
*device
;
1428 const struct wined3d_gl_info
*gl_info
;
1429 const char *tex_dst
= dst_str
;
1430 struct color_fixup_masks masks
;
1432 /* D3D vertex shader sampler IDs are vertex samplers(0-3), not global d3d samplers */
1433 if(!pshader
) sampler_idx
+= MAX_FRAGMENT_SAMPLERS
;
1435 switch (resource_type
)
1437 case WINED3D_SHADER_RESOURCE_TEXTURE_1D
:
1441 case WINED3D_SHADER_RESOURCE_TEXTURE_2D
:
1442 shader
= ins
->ctx
->shader
;
1443 device
= shader
->device
;
1444 gl_info
= &device
->adapter
->gl_info
;
1446 if (pshader
&& priv
->cur_ps_args
->super
.np2_fixup
& (1u << sampler_idx
)
1447 && gl_info
->supported
[ARB_TEXTURE_RECTANGLE
])
1451 if (shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
))
1453 if (priv
->cur_np2fixup_info
->super
.active
& (1u << sampler_idx
))
1455 if (flags
) FIXME("Only ordinary sampling from NP2 textures is supported.\n");
1456 else np2_fixup
= TRUE
;
1461 case WINED3D_SHADER_RESOURCE_TEXTURE_3D
:
1465 case WINED3D_SHADER_RESOURCE_TEXTURE_CUBE
:
1470 ERR("Unexpected resource type %#x.\n", resource_type
);
1474 /* TEX, TXL, TXD and TXP do not support the "H" modifier,
1475 * so don't use shader_arb_get_modifier
1477 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
) mod
= "_SAT";
1480 /* Fragment samplers always have indentity mapping */
1481 if(sampler_idx
>= MAX_FRAGMENT_SAMPLERS
)
1483 sampler_idx
= priv
->cur_vs_args
->vertex
.samplers
[sampler_idx
- MAX_FRAGMENT_SAMPLERS
];
1488 masks
= calc_color_correction(priv
->cur_ps_args
->super
.color_fixup
[sampler_idx
],
1489 ins
->dst
[0].write_mask
);
1491 if (masks
.source
|| masks
.sign
)
1495 if (flags
& TEX_DERIV
)
1497 if(flags
& TEX_PROJ
) FIXME("Projected texture sampling with custom derivatives\n");
1498 if(flags
& TEX_BIAS
) FIXME("Biased texture sampling with custom derivatives\n");
1499 shader_addline(buffer
, "TXD%s %s, %s, %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
,
1500 dsx
, dsy
, sampler_idx
, tex_type
);
1502 else if(flags
& TEX_LOD
)
1504 if(flags
& TEX_PROJ
) FIXME("Projected texture sampling with explicit lod\n");
1505 if(flags
& TEX_BIAS
) FIXME("Biased texture sampling with explicit lod\n");
1506 shader_addline(buffer
, "TXL%s %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
,
1507 sampler_idx
, tex_type
);
1509 else if (flags
& TEX_BIAS
)
1511 /* Shouldn't be possible, but let's check for it */
1512 if(flags
& TEX_PROJ
) FIXME("Biased and Projected texture sampling\n");
1513 /* TXB takes the 4th component of the source vector automatically, as d3d. Nothing more to do */
1514 shader_addline(buffer
, "TXB%s %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
, sampler_idx
, tex_type
);
1516 else if (flags
& TEX_PROJ
)
1518 shader_addline(buffer
, "TXP%s %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
, sampler_idx
, tex_type
);
1524 const unsigned char idx
= priv
->cur_np2fixup_info
->super
.idx
[sampler_idx
];
1525 shader_addline(buffer
, "MUL TA, np2fixup[%u].%s, %s;\n", idx
>> 1,
1526 (idx
% 2) ? "zwxy" : "xyzw", coord_reg
);
1528 shader_addline(buffer
, "TEX%s %s, TA, texture[%u], %s;\n", mod
, tex_dst
, sampler_idx
, tex_type
);
1531 shader_addline(buffer
, "TEX%s %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
, sampler_idx
, tex_type
);
1536 gen_color_correction(buffer
, dst_str
, tex_dst
,
1537 arb_get_helper_value(WINED3D_SHADER_TYPE_PIXEL
, ARB_ONE
),
1538 arb_get_helper_value(WINED3D_SHADER_TYPE_PIXEL
, ARB_TWO
),
1539 priv
->cur_ps_args
->super
.color_fixup
[sampler_idx
], masks
);
1543 static void shader_arb_get_src_param(const struct wined3d_shader_instruction
*ins
,
1544 const struct wined3d_shader_src_param
*src
, unsigned int tmpreg
, char *outregstr
)
1546 /* Generate a line that does the input modifier computation and return the input register to use */
1547 BOOL is_color
= FALSE
, insert_line
;
1550 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1551 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1552 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
1553 const char *two
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_TWO
);
1555 /* Assume a new line will be added */
1558 /* Get register name */
1559 shader_arb_get_register_name(ins
, &src
->reg
, regstr
, &is_color
);
1560 shader_arb_get_swizzle(src
, is_color
, swzstr
);
1562 switch (src
->modifiers
)
1564 case WINED3DSPSM_NONE
:
1565 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
1566 insert_line
= FALSE
;
1568 case WINED3DSPSM_NEG
:
1569 sprintf(outregstr
, "-%s%s", regstr
, swzstr
);
1570 insert_line
= FALSE
;
1572 case WINED3DSPSM_BIAS
:
1573 shader_addline(buffer
, "ADD T%c, %s, -coefdiv.x;\n", 'A' + tmpreg
, regstr
);
1575 case WINED3DSPSM_BIASNEG
:
1576 shader_addline(buffer
, "ADD T%c, -%s, coefdiv.x;\n", 'A' + tmpreg
, regstr
);
1578 case WINED3DSPSM_SIGN
:
1579 shader_addline(buffer
, "MAD T%c, %s, %s, -%s;\n", 'A' + tmpreg
, regstr
, two
, one
);
1581 case WINED3DSPSM_SIGNNEG
:
1582 shader_addline(buffer
, "MAD T%c, %s, -%s, %s;\n", 'A' + tmpreg
, regstr
, two
, one
);
1584 case WINED3DSPSM_COMP
:
1585 shader_addline(buffer
, "SUB T%c, %s, %s;\n", 'A' + tmpreg
, one
, regstr
);
1587 case WINED3DSPSM_X2
:
1588 shader_addline(buffer
, "ADD T%c, %s, %s;\n", 'A' + tmpreg
, regstr
, regstr
);
1590 case WINED3DSPSM_X2NEG
:
1591 shader_addline(buffer
, "ADD T%c, -%s, -%s;\n", 'A' + tmpreg
, regstr
, regstr
);
1593 case WINED3DSPSM_DZ
:
1594 shader_addline(buffer
, "RCP T%c, %s.z;\n", 'A' + tmpreg
, regstr
);
1595 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
1597 case WINED3DSPSM_DW
:
1598 shader_addline(buffer
, "RCP T%c, %s.w;\n", 'A' + tmpreg
, regstr
);
1599 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
1601 case WINED3DSPSM_ABS
:
1602 if(ctx
->target_version
>= NV2
) {
1603 sprintf(outregstr
, "|%s%s|", regstr
, swzstr
);
1604 insert_line
= FALSE
;
1606 shader_addline(buffer
, "ABS T%c, %s;\n", 'A' + tmpreg
, regstr
);
1609 case WINED3DSPSM_ABSNEG
:
1610 if(ctx
->target_version
>= NV2
) {
1611 sprintf(outregstr
, "-|%s%s|", regstr
, swzstr
);
1613 shader_addline(buffer
, "ABS T%c, %s;\n", 'A' + tmpreg
, regstr
);
1614 sprintf(outregstr
, "-T%c%s", 'A' + tmpreg
, swzstr
);
1616 insert_line
= FALSE
;
1619 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
1620 insert_line
= FALSE
;
1623 /* Return modified or original register, with swizzle */
1625 sprintf(outregstr
, "T%c%s", 'A' + tmpreg
, swzstr
);
1628 static void pshader_hw_bem(const struct wined3d_shader_instruction
*ins
)
1630 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1631 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1632 DWORD sampler_code
= dst
->reg
.idx
[0].offset
;
1634 char src_name
[2][50];
1636 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1638 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed
1640 * Keep in mind that src_name[1] can be "TB" and src_name[0] can be "TA" because modifiers like _x2 are valid
1641 * with bem. So delay loading the first parameter until after the perturbation calculation which needs two
1644 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1645 shader_addline(buffer
, "SWZ TA, bumpenvmat%d, x, z, 0, 0;\n", sampler_code
);
1646 shader_addline(buffer
, "DP3 TC.r, TA, %s;\n", src_name
[1]);
1647 shader_addline(buffer
, "SWZ TA, bumpenvmat%d, y, w, 0, 0;\n", sampler_code
);
1648 shader_addline(buffer
, "DP3 TC.g, TA, %s;\n", src_name
[1]);
1650 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1651 shader_addline(buffer
, "ADD %s, %s, TC;\n", dst_name
, src_name
[0]);
1654 static DWORD
negate_modifiers(DWORD mod
, char *extra_char
)
1659 case WINED3DSPSM_NONE
: return WINED3DSPSM_NEG
;
1660 case WINED3DSPSM_NEG
: return WINED3DSPSM_NONE
;
1661 case WINED3DSPSM_BIAS
: return WINED3DSPSM_BIASNEG
;
1662 case WINED3DSPSM_BIASNEG
: return WINED3DSPSM_BIAS
;
1663 case WINED3DSPSM_SIGN
: return WINED3DSPSM_SIGNNEG
;
1664 case WINED3DSPSM_SIGNNEG
: return WINED3DSPSM_SIGN
;
1665 case WINED3DSPSM_COMP
: *extra_char
= '-'; return WINED3DSPSM_COMP
;
1666 case WINED3DSPSM_X2
: return WINED3DSPSM_X2NEG
;
1667 case WINED3DSPSM_X2NEG
: return WINED3DSPSM_X2
;
1668 case WINED3DSPSM_DZ
: *extra_char
= '-'; return WINED3DSPSM_DZ
;
1669 case WINED3DSPSM_DW
: *extra_char
= '-'; return WINED3DSPSM_DW
;
1670 case WINED3DSPSM_ABS
: return WINED3DSPSM_ABSNEG
;
1671 case WINED3DSPSM_ABSNEG
: return WINED3DSPSM_ABS
;
1673 FIXME("Unknown modifier %u\n", mod
);
1677 static void pshader_hw_cnd(const struct wined3d_shader_instruction
*ins
)
1679 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1680 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1682 char src_name
[3][50];
1683 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1684 ins
->ctx
->reg_maps
->shader_version
.minor
);
1686 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1687 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1689 if (shader_version
<= WINED3D_SHADER_VERSION(1, 3) && ins
->coissue
1690 && ins
->dst
->write_mask
!= WINED3DSP_WRITEMASK_3
)
1692 shader_addline(buffer
, "MOV%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
[1]);
1696 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
1699 /* src0 may have a negate srcmod set, so we can't blindly add "-" to the name */
1700 src0_copy
.modifiers
= negate_modifiers(src0_copy
.modifiers
, &extra_neg
);
1702 shader_arb_get_src_param(ins
, &src0_copy
, 0, src_name
[0]);
1703 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1704 shader_addline(buffer
, "ADD TA, %c%s, coefdiv.x;\n", extra_neg
, src_name
[0]);
1705 shader_addline(buffer
, "CMP%s %s, TA, %s, %s;\n", shader_arb_get_modifier(ins
),
1706 dst_name
, src_name
[1], src_name
[2]);
1710 static void pshader_hw_cmp(const struct wined3d_shader_instruction
*ins
)
1712 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1713 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1715 char src_name
[3][50];
1717 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1719 /* Generate input register names (with modifiers) */
1720 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1721 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1722 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1724 shader_addline(buffer
, "CMP%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1725 dst_name
, src_name
[0], src_name
[2], src_name
[1]);
1728 /** Process the WINED3DSIO_DP2ADD instruction in ARB.
1729 * dst = dot2(src0, src1) + src2 */
1730 static void pshader_hw_dp2add(const struct wined3d_shader_instruction
*ins
)
1732 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1733 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1735 char src_name
[3][50];
1736 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1738 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1739 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1740 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1742 if(ctx
->target_version
>= NV3
)
1744 /* GL_NV_fragment_program2 has a 1:1 matching instruction */
1745 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1746 shader_addline(buffer
, "DP2A%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1747 dst_name
, src_name
[0], src_name
[1], src_name
[2]);
1749 else if(ctx
->target_version
>= NV2
)
1751 /* dst.x = src2.?, src0.x, src1.x + src0.y * src1.y
1752 * dst.y = src2.?, src0.x, src1.z + src0.y * src1.w
1753 * dst.z = src2.?, src0.x, src1.x + src0.y * src1.y
1754 * dst.z = src2.?, src0.x, src1.z + src0.y * src1.w
1756 * Make sure that src1.zw = src1.xy, then we get a classic dp2add
1758 * .xyxy and other swizzles that we could get with this are not valid in
1759 * plain ARBfp, but luckily the NV extension grammar lifts this limitation.
1761 struct wined3d_shader_src_param tmp_param
= ins
->src
[1];
1762 DWORD swizzle
= tmp_param
.swizzle
& 0xf; /* Selects .xy */
1763 tmp_param
.swizzle
= swizzle
| (swizzle
<< 4); /* Creates .xyxy */
1765 shader_arb_get_src_param(ins
, &tmp_param
, 1, src_name
[1]);
1767 shader_addline(buffer
, "X2D%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1768 dst_name
, src_name
[2], src_name
[0], src_name
[1]);
1772 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1773 /* 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]
1774 * src_name[0] can be TA, but TA is a private temp for modifiers, so it is save to overwrite
1776 shader_addline(buffer
, "MOV TA, %s;\n", src_name
[0]);
1777 shader_addline(buffer
, "MOV TA.z, 0.0;\n");
1778 shader_addline(buffer
, "DP3 TA, TA, %s;\n", src_name
[1]);
1779 shader_addline(buffer
, "ADD%s %s, TA, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
[2]);
1783 /* Map the opcode 1-to-1 to the GL code */
1784 static void shader_hw_map2gl(const struct wined3d_shader_instruction
*ins
)
1786 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1787 const char *instruction
;
1788 char arguments
[256], dst_str
[50];
1790 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1792 switch (ins
->handler_idx
)
1794 case WINED3DSIH_ABS
: instruction
= "ABS"; break;
1795 case WINED3DSIH_ADD
: instruction
= "ADD"; break;
1796 case WINED3DSIH_CRS
: instruction
= "XPD"; break;
1797 case WINED3DSIH_DP3
: instruction
= "DP3"; break;
1798 case WINED3DSIH_DP4
: instruction
= "DP4"; break;
1799 case WINED3DSIH_DST
: instruction
= "DST"; break;
1800 case WINED3DSIH_FRC
: instruction
= "FRC"; break;
1801 case WINED3DSIH_LIT
: instruction
= "LIT"; break;
1802 case WINED3DSIH_LRP
: instruction
= "LRP"; break;
1803 case WINED3DSIH_MAD
: instruction
= "MAD"; break;
1804 case WINED3DSIH_MAX
: instruction
= "MAX"; break;
1805 case WINED3DSIH_MIN
: instruction
= "MIN"; break;
1806 case WINED3DSIH_MOV
: instruction
= "MOV"; break;
1807 case WINED3DSIH_MUL
: instruction
= "MUL"; break;
1808 case WINED3DSIH_SGE
: instruction
= "SGE"; break;
1809 case WINED3DSIH_SLT
: instruction
= "SLT"; break;
1810 case WINED3DSIH_SUB
: instruction
= "SUB"; break;
1811 case WINED3DSIH_MOVA
:instruction
= "ARR"; break;
1812 case WINED3DSIH_DSX
: instruction
= "DDX"; break;
1813 default: instruction
= "";
1814 FIXME("Unhandled opcode %s.\n", debug_d3dshaderinstructionhandler(ins
->handler_idx
));
1818 /* Note that shader_arb_add_dst_param() adds spaces. */
1819 arguments
[0] = '\0';
1820 shader_arb_get_dst_param(ins
, dst
, dst_str
);
1821 for (i
= 0; i
< ins
->src_count
; ++i
)
1824 strcat(arguments
, ", ");
1825 shader_arb_get_src_param(ins
, &ins
->src
[i
], i
, operand
);
1826 strcat(arguments
, operand
);
1828 shader_addline(buffer
, "%s%s %s%s;\n", instruction
, shader_arb_get_modifier(ins
), dst_str
, arguments
);
1831 static void shader_hw_nop(const struct wined3d_shader_instruction
*ins
) {}
1833 static DWORD
shader_arb_select_component(DWORD swizzle
, DWORD component
)
1835 return ((swizzle
>> 2 * component
) & 0x3) * 0x55;
1838 static void shader_hw_mov(const struct wined3d_shader_instruction
*ins
)
1840 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
1841 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
1842 BOOL pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
1843 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1844 const char *zero
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ZERO
);
1845 const char *one
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ONE
);
1846 const char *two
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_TWO
);
1848 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1849 char src0_param
[256];
1851 if (ins
->handler_idx
== WINED3DSIH_MOVA
)
1853 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1855 const char *offset
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_VS_REL_OFFSET
);
1857 if(ctx
->target_version
>= NV2
) {
1858 shader_hw_map2gl(ins
);
1861 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_param
);
1862 shader_arb_get_write_mask(ins
, &ins
->dst
[0], write_mask
);
1864 /* This implements the mova formula used in GLSL. The first two instructions
1865 * prepare the sign() part. Note that it is fine to have my_sign(0.0) = 1.0
1869 * A0.x = arl(floor(abs(0.0) + 0.5) * 1.0) = floor(0.5) = 0.0 since arl does a floor
1871 * The ARL is performed when A0 is used - the requested component is read from A0_SHADOW into
1872 * A0.x. We can use the overwritten component of A0_shadow as temporary storage for the sign.
1874 shader_addline(buffer
, "SGE A0_SHADOW%s, %s, %s;\n", write_mask
, src0_param
, zero
);
1875 shader_addline(buffer
, "MAD A0_SHADOW%s, A0_SHADOW, %s, -%s;\n", write_mask
, two
, one
);
1877 shader_addline(buffer
, "ABS TA%s, %s;\n", write_mask
, src0_param
);
1878 shader_addline(buffer
, "ADD TA%s, TA, rel_addr_const.x;\n", write_mask
);
1879 shader_addline(buffer
, "FLR TA%s, TA;\n", write_mask
);
1880 if (shader_data
->rel_offset
)
1882 shader_addline(buffer
, "ADD TA%s, TA, %s;\n", write_mask
, offset
);
1884 shader_addline(buffer
, "MUL A0_SHADOW%s, TA, A0_SHADOW;\n", write_mask
);
1886 ((struct shader_arb_ctx_priv
*)ins
->ctx
->backend_data
)->addr_reg
[0] = '\0';
1888 else if (reg_maps
->shader_version
.major
== 1
1889 && !shader_is_pshader_version(reg_maps
->shader_version
.type
)
1890 && ins
->dst
[0].reg
.type
== WINED3DSPR_ADDR
)
1892 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1893 src0_param
[0] = '\0';
1895 if (shader_data
->rel_offset
&& ctx
->target_version
== ARB
)
1897 const char *offset
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_VS_REL_OFFSET
);
1898 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_param
);
1899 shader_addline(buffer
, "ADD TA.x, %s, %s;\n", src0_param
, offset
);
1900 shader_addline(buffer
, "ARL A0.x, TA.x;\n");
1904 /* Apple's ARB_vertex_program implementation does not accept an ARL source argument
1905 * with more than one component. Thus replicate the first source argument over all
1906 * 4 components. For example, .xyzw -> .x (or better: .xxxx), .zwxy -> .z, etc) */
1907 struct wined3d_shader_src_param tmp_src
= ins
->src
[0];
1908 tmp_src
.swizzle
= shader_arb_select_component(tmp_src
.swizzle
, 0);
1909 shader_arb_get_src_param(ins
, &tmp_src
, 0, src0_param
);
1910 shader_addline(buffer
, "ARL A0.x, %s;\n", src0_param
);
1913 else if (ins
->dst
[0].reg
.type
== WINED3DSPR_COLOROUT
&& !ins
->dst
[0].reg
.idx
[0].offset
&& pshader
)
1915 if (ctx
->ps_post_process
&& shader
->u
.ps
.color0_mov
)
1917 shader_addline(buffer
, "#mov handled in srgb write or fog code\n");
1920 shader_hw_map2gl(ins
);
1924 shader_hw_map2gl(ins
);
1928 static void pshader_hw_texkill(const struct wined3d_shader_instruction
*ins
)
1930 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1931 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1934 /* No swizzles are allowed in d3d's texkill. PS 1.x ignores the 4th component as documented,
1935 * but >= 2.0 honors it (undocumented, but tested by the d3d9 testsuite)
1937 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1939 if (ins
->ctx
->reg_maps
->shader_version
.major
>= 2)
1941 const char *kilsrc
= "TA";
1944 shader_arb_get_register_name(ins
, &dst
->reg
, reg_dest
, &is_color
);
1945 if(dst
->write_mask
== WINED3DSP_WRITEMASK_ALL
)
1951 /* Sigh. KIL doesn't support swizzles/writemasks. KIL passes a writemask, but ".xy" for example
1952 * is not valid as a swizzle in ARB (needs ".xyyy"). Use SWZ to load the register properly, and set
1953 * masked out components to 0(won't kill)
1955 char x
= '0', y
= '0', z
= '0', w
= '0';
1956 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
) x
= 'x';
1957 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
) y
= 'y';
1958 if(dst
->write_mask
& WINED3DSP_WRITEMASK_2
) z
= 'z';
1959 if(dst
->write_mask
& WINED3DSP_WRITEMASK_3
) w
= 'w';
1960 shader_addline(buffer
, "SWZ TA, %s, %c, %c, %c, %c;\n", reg_dest
, x
, y
, z
, w
);
1962 shader_addline(buffer
, "KIL %s;\n", kilsrc
);
1966 /* ARB fp doesn't like swizzles on the parameter of the KIL instruction. To mask the 4th component,
1967 * copy the register into our general purpose TMP variable, overwrite .w and pass TMP to KIL
1969 * ps_1_3 shaders use the texcoord incarnation of the Tx register. ps_1_4 shaders can use the same,
1970 * or pass in any temporary register(in shader phase 2)
1972 if (ins
->ctx
->reg_maps
->shader_version
.minor
<= 3)
1973 sprintf(reg_dest
, "fragment.texcoord[%u]", dst
->reg
.idx
[0].offset
);
1975 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1976 shader_addline(buffer
, "SWZ TA, %s, x, y, z, 1;\n", reg_dest
);
1977 shader_addline(buffer
, "KIL TA;\n");
1981 static void pshader_hw_tex(const struct wined3d_shader_instruction
*ins
)
1983 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1984 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1985 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1986 ins
->ctx
->reg_maps
->shader_version
.minor
);
1987 struct wined3d_shader_src_param src
;
1991 DWORD reg_sampler_code
;
1993 BOOL swizzle_coord
= FALSE
;
1995 /* All versions have a destination register */
1996 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1998 /* 1.0-1.4: Use destination register number as texture code.
1999 2.0+: Use provided sampler number as texture code. */
2000 if (shader_version
< WINED3D_SHADER_VERSION(2,0))
2001 reg_sampler_code
= dst
->reg
.idx
[0].offset
;
2003 reg_sampler_code
= ins
->src
[1].reg
.idx
[0].offset
;
2005 /* 1.0-1.3: Use the texcoord varying.
2006 1.4+: Use provided coordinate source register. */
2007 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
2008 sprintf(reg_coord
, "fragment.texcoord[%u]", reg_sampler_code
);
2010 /* TEX is the only instruction that can handle DW and DZ natively */
2012 if(src
.modifiers
== WINED3DSPSM_DW
) src
.modifiers
= WINED3DSPSM_NONE
;
2013 if(src
.modifiers
== WINED3DSPSM_DZ
) src
.modifiers
= WINED3DSPSM_NONE
;
2014 shader_arb_get_src_param(ins
, &src
, 0, reg_coord
);
2018 * 1.1, 1.2, 1.3: Use WINED3D_TSS_TEXTURETRANSFORMFLAGS
2019 * 1.4: Use WINED3DSPSM_DZ or WINED3DSPSM_DW on src[0]
2020 * 2.0+: Use WINED3DSI_TEXLD_PROJECT on the opcode
2022 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
2025 if (reg_sampler_code
< MAX_TEXTURES
)
2026 flags
= priv
->cur_ps_args
->super
.tex_transform
>> reg_sampler_code
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
;
2027 if (flags
& WINED3D_PSARGS_PROJECTED
)
2029 myflags
|= TEX_PROJ
;
2030 if ((flags
& ~WINED3D_PSARGS_PROJECTED
) == WINED3D_TTFF_COUNT3
)
2031 swizzle_coord
= TRUE
;
2034 else if (shader_version
< WINED3D_SHADER_VERSION(2,0))
2036 enum wined3d_shader_src_modifier src_mod
= ins
->src
[0].modifiers
;
2037 if (src_mod
== WINED3DSPSM_DZ
)
2039 swizzle_coord
= TRUE
;
2040 myflags
|= TEX_PROJ
;
2041 } else if(src_mod
== WINED3DSPSM_DW
) {
2042 myflags
|= TEX_PROJ
;
2045 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) myflags
|= TEX_PROJ
;
2046 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) myflags
|= TEX_BIAS
;
2051 /* TXP cannot handle DZ natively, so move the z coordinate to .w.
2052 * reg_coord is a read-only varying register, so we need a temp reg */
2053 shader_addline(ins
->ctx
->buffer
, "SWZ TA, %s, x, y, z, z;\n", reg_coord
);
2054 strcpy(reg_coord
, "TA");
2057 shader_hw_sample(ins
, reg_sampler_code
, reg_dest
, reg_coord
, myflags
, NULL
, NULL
);
2060 static void pshader_hw_texcoord(const struct wined3d_shader_instruction
*ins
)
2062 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2063 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2064 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
2065 ins
->ctx
->reg_maps
->shader_version
.minor
);
2068 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
2070 DWORD reg
= dst
->reg
.idx
[0].offset
;
2072 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2073 shader_addline(buffer
, "MOV_SAT %s, fragment.texcoord[%u];\n", dst_str
, reg
);
2077 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_src
);
2078 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2079 shader_addline(buffer
, "MOV %s, %s;\n", dst_str
, reg_src
);
2083 static void pshader_hw_texreg2ar(const struct wined3d_shader_instruction
*ins
)
2085 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2088 DWORD reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2092 /* Note that texreg2ar treats Tx as a temporary register, not as a varying */
2093 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2094 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2095 /* Move .x first in case src_str is "TA" */
2096 shader_addline(buffer
, "MOV TA.y, %s.x;\n", src_str
);
2097 shader_addline(buffer
, "MOV TA.x, %s.w;\n", src_str
);
2098 if (reg1
< MAX_TEXTURES
)
2100 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2101 flags
= priv
->cur_ps_args
->super
.tex_transform
>> reg1
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
;
2103 shader_hw_sample(ins
, reg1
, dst_str
, "TA", flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2106 static void pshader_hw_texreg2gb(const struct wined3d_shader_instruction
*ins
)
2108 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2110 DWORD reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2114 /* Note that texreg2gb treats Tx as a temporary register, not as a varying */
2115 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2116 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2117 shader_addline(buffer
, "MOV TA.x, %s.y;\n", src_str
);
2118 shader_addline(buffer
, "MOV TA.y, %s.z;\n", src_str
);
2119 shader_hw_sample(ins
, reg1
, dst_str
, "TA", 0, NULL
, NULL
);
2122 static void pshader_hw_texreg2rgb(const struct wined3d_shader_instruction
*ins
)
2124 DWORD reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2128 /* Note that texreg2rg treats Tx as a temporary register, not as a varying */
2129 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2130 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2131 shader_hw_sample(ins
, reg1
, dst_str
, src_str
, 0, NULL
, NULL
);
2134 static void pshader_hw_texbem(const struct wined3d_shader_instruction
*ins
)
2136 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2137 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2138 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2139 char reg_coord
[40], dst_reg
[50], src_reg
[50];
2140 DWORD reg_dest_code
;
2142 /* All versions have a destination register. The Tx where the texture coordinates come
2143 * from is the varying incarnation of the texture register
2145 reg_dest_code
= dst
->reg
.idx
[0].offset
;
2146 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_reg
);
2147 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_reg
);
2148 sprintf(reg_coord
, "fragment.texcoord[%u]", reg_dest_code
);
2150 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed
2151 * The Tx in which the perturbation map is stored is the tempreg incarnation of the texture register
2153 * GL_NV_fragment_program_option could handle this in one instruction via X2D:
2154 * X2D TA.xy, fragment.texcoord, T%u, bumpenvmat%u.xzyw
2156 * However, the NV extensions are never enabled for <= 2.0 shaders because of the performance penalty that
2157 * comes with it, and texbem is an 1.x only instruction. No 1.x instruction forces us to enable the NV
2160 shader_addline(buffer
, "SWZ TB, bumpenvmat%d, x, z, 0, 0;\n", reg_dest_code
);
2161 shader_addline(buffer
, "DP3 TA.x, TB, %s;\n", src_reg
);
2162 shader_addline(buffer
, "SWZ TB, bumpenvmat%d, y, w, 0, 0;\n", reg_dest_code
);
2163 shader_addline(buffer
, "DP3 TA.y, TB, %s;\n", src_reg
);
2165 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2166 * so we can't let the GL handle this.
2168 if ((priv
->cur_ps_args
->super
.tex_transform
>> reg_dest_code
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
)
2169 & WINED3D_PSARGS_PROJECTED
)
2171 shader_addline(buffer
, "RCP TB.w, %s.w;\n", reg_coord
);
2172 shader_addline(buffer
, "MUL TB.xy, %s, TB.w;\n", reg_coord
);
2173 shader_addline(buffer
, "ADD TA.xy, TA, TB;\n");
2175 shader_addline(buffer
, "ADD TA.xy, TA, %s;\n", reg_coord
);
2178 shader_hw_sample(ins
, reg_dest_code
, dst_reg
, "TA", 0, NULL
, NULL
);
2180 if (ins
->handler_idx
== WINED3DSIH_TEXBEML
)
2182 /* No src swizzles are allowed, so this is ok */
2183 shader_addline(buffer
, "MAD TA, %s.z, luminance%d.x, luminance%d.y;\n",
2184 src_reg
, reg_dest_code
, reg_dest_code
);
2185 shader_addline(buffer
, "MUL %s, %s, TA;\n", dst_reg
, dst_reg
);
2189 static void pshader_hw_texm3x2pad(const struct wined3d_shader_instruction
*ins
)
2191 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2192 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2193 char src0_name
[50], dst_name
[50];
2195 struct wined3d_shader_register tmp_reg
= ins
->dst
[0].reg
;
2197 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2198 /* The next instruction will be a texm3x2tex or texm3x2depth that writes to the uninitialized
2199 * T<reg+1> register. Use this register to store the calculated vector
2201 tmp_reg
.idx
[0].offset
= reg
+ 1;
2202 shader_arb_get_register_name(ins
, &tmp_reg
, dst_name
, &is_color
);
2203 shader_addline(buffer
, "DP3 %s.x, fragment.texcoord[%u], %s;\n", dst_name
, reg
, src0_name
);
2206 static void pshader_hw_texm3x2tex(const struct wined3d_shader_instruction
*ins
)
2208 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2210 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2211 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2217 /* We know that we're writing to the uninitialized T<reg> register, so use it for temporary storage */
2218 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2220 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2221 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2222 shader_addline(buffer
, "DP3 %s.y, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2223 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2224 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2227 static void pshader_hw_texm3x3pad(const struct wined3d_shader_instruction
*ins
)
2229 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2230 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2231 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2232 char src0_name
[50], dst_name
[50];
2233 struct wined3d_shader_register tmp_reg
= ins
->dst
[0].reg
;
2236 /* There are always 2 texm3x3pad instructions followed by one texm3x3[tex,vspec, ...] instruction, with
2237 * incrementing ins->dst[0].register_idx numbers. So the pad instruction already knows the final destination
2238 * register, and this register is uninitialized(otherwise the assembler complains that it is 'redeclared')
2240 tmp_reg
.idx
[0].offset
= reg
+ 2 - tex_mx
->current_row
;
2241 shader_arb_get_register_name(ins
, &tmp_reg
, dst_name
, &is_color
);
2243 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2244 shader_addline(buffer
, "DP3 %s.%c, fragment.texcoord[%u], %s;\n",
2245 dst_name
, 'x' + tex_mx
->current_row
, reg
, src0_name
);
2246 tex_mx
->texcoord_w
[tex_mx
->current_row
++] = reg
;
2249 static void pshader_hw_texm3x3tex(const struct wined3d_shader_instruction
*ins
)
2251 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2252 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2254 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2255 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2257 char src0_name
[50], dst_name
[50];
2260 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2261 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2262 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_name
, reg
, src0_name
);
2264 /* Sample the texture using the calculated coordinates */
2265 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2266 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2267 shader_hw_sample(ins
, reg
, dst_str
, dst_name
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2268 tex_mx
->current_row
= 0;
2271 static void pshader_hw_texm3x3vspec(const struct wined3d_shader_instruction
*ins
)
2273 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2274 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2276 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2277 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2283 /* Get the dst reg without writemask strings. We know this register is uninitialized, so we can use all
2284 * components for temporary data storage
2286 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2287 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2288 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2290 /* Construct the eye-ray vector from w coordinates */
2291 shader_addline(buffer
, "MOV TB.x, fragment.texcoord[%u].w;\n", tex_mx
->texcoord_w
[0]);
2292 shader_addline(buffer
, "MOV TB.y, fragment.texcoord[%u].w;\n", tex_mx
->texcoord_w
[1]);
2293 shader_addline(buffer
, "MOV TB.z, fragment.texcoord[%u].w;\n", reg
);
2295 /* Calculate reflection vector
2297 shader_addline(buffer
, "DP3 %s.w, %s, TB;\n", dst_reg
, dst_reg
);
2298 /* The .w is ignored when sampling, so I can use TB.w to calculate dot(N, N) */
2299 shader_addline(buffer
, "DP3 TB.w, %s, %s;\n", dst_reg
, dst_reg
);
2300 shader_addline(buffer
, "RCP TB.w, TB.w;\n");
2301 shader_addline(buffer
, "MUL %s.w, %s.w, TB.w;\n", dst_reg
, dst_reg
);
2302 shader_addline(buffer
, "MUL %s, %s.w, %s;\n", dst_reg
, dst_reg
, dst_reg
);
2303 shader_addline(buffer
, "MAD %s, coefmul.x, %s, -TB;\n", dst_reg
, dst_reg
);
2305 /* Sample the texture using the calculated coordinates */
2306 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2307 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2308 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2309 tex_mx
->current_row
= 0;
2312 static void pshader_hw_texm3x3spec(const struct wined3d_shader_instruction
*ins
)
2314 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2315 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2317 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2318 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2325 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2326 shader_arb_get_src_param(ins
, &ins
->src
[0], 1, src1_name
);
2327 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2328 /* Note: dst_reg.xy is input here, generated by two texm3x3pad instructions */
2329 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2331 /* Calculate reflection vector.
2334 * dst_reg.xyz = 2 * --------- * N - E
2337 * Which normalizes the normal vector
2339 shader_addline(buffer
, "DP3 %s.w, %s, %s;\n", dst_reg
, dst_reg
, src1_name
);
2340 shader_addline(buffer
, "DP3 TC.w, %s, %s;\n", dst_reg
, dst_reg
);
2341 shader_addline(buffer
, "RCP TC.w, TC.w;\n");
2342 shader_addline(buffer
, "MUL %s.w, %s.w, TC.w;\n", dst_reg
, dst_reg
);
2343 shader_addline(buffer
, "MUL %s, %s.w, %s;\n", dst_reg
, dst_reg
, dst_reg
);
2344 shader_addline(buffer
, "MAD %s, coefmul.x, %s, -%s;\n", dst_reg
, dst_reg
, src1_name
);
2346 /* Sample the texture using the calculated coordinates */
2347 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2348 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2349 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2350 tex_mx
->current_row
= 0;
2353 static void pshader_hw_texdepth(const struct wined3d_shader_instruction
*ins
)
2355 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2356 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2358 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2359 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2361 /* texdepth has an implicit destination, the fragment depth value. It's only parameter,
2362 * which is essentially an input, is the destination register because it is the first
2363 * parameter. According to the msdn, this must be register r5, but let's keep it more flexible
2364 * here(writemasks/swizzles are not valid on texdepth)
2366 shader_arb_get_dst_param(ins
, dst
, dst_name
);
2368 /* According to the msdn, the source register(must be r5) is unusable after
2369 * the texdepth instruction, so we're free to modify it
2371 shader_addline(buffer
, "MIN %s.y, %s.y, %s;\n", dst_name
, dst_name
, one
);
2373 /* How to deal with the special case dst_name.g == 0? if r != 0, then
2374 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
2375 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
2377 shader_addline(buffer
, "RCP %s.y, %s.y;\n", dst_name
, dst_name
);
2378 shader_addline(buffer
, "MUL TA.x, %s.x, %s.y;\n", dst_name
, dst_name
);
2379 shader_addline(buffer
, "MIN TA.x, TA.x, %s;\n", one
);
2380 shader_addline(buffer
, "MAX result.depth, TA.x, %s;\n", zero
);
2383 /** Process the WINED3DSIO_TEXDP3TEX instruction in ARB:
2384 * Take a 3-component dot product of the TexCoord[dstreg] and src,
2385 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
2386 static void pshader_hw_texdp3tex(const struct wined3d_shader_instruction
*ins
)
2388 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2389 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
[0].offset
;
2393 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2394 shader_addline(buffer
, "MOV TB, 0.0;\n");
2395 shader_addline(buffer
, "DP3 TB.x, fragment.texcoord[%u], %s;\n", sampler_idx
, src0
);
2397 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2398 shader_hw_sample(ins
, sampler_idx
, dst_str
, "TB", 0 /* Only one coord, can't be projected */, NULL
, NULL
);
2401 /** Process the WINED3DSIO_TEXDP3 instruction in ARB:
2402 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
2403 static void pshader_hw_texdp3(const struct wined3d_shader_instruction
*ins
)
2405 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2408 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2410 /* Handle output register */
2411 shader_arb_get_dst_param(ins
, dst
, dst_str
);
2412 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2413 shader_addline(buffer
, "DP3 %s, fragment.texcoord[%u], %s;\n", dst_str
, dst
->reg
.idx
[0].offset
, src0
);
2416 /** Process the WINED3DSIO_TEXM3X3 instruction in ARB
2417 * Perform the 3rd row of a 3x3 matrix multiply */
2418 static void pshader_hw_texm3x3(const struct wined3d_shader_instruction
*ins
)
2420 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2421 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2422 char dst_str
[50], dst_name
[50];
2426 shader_arb_get_dst_param(ins
, dst
, dst_str
);
2427 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2428 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2429 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_name
, dst
->reg
.idx
[0].offset
, src0
);
2430 shader_addline(buffer
, "MOV %s, %s;\n", dst_str
, dst_name
);
2433 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in ARB:
2434 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
2435 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
2436 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
2438 static void pshader_hw_texm3x2depth(const struct wined3d_shader_instruction
*ins
)
2440 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2441 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2442 char src0
[50], dst_name
[50];
2444 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2445 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2447 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2448 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2449 shader_addline(buffer
, "DP3 %s.y, fragment.texcoord[%u], %s;\n", dst_name
, dst
->reg
.idx
[0].offset
, src0
);
2451 /* How to deal with the special case dst_name.g == 0? if r != 0, then
2452 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
2453 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
2455 shader_addline(buffer
, "RCP %s.y, %s.y;\n", dst_name
, dst_name
);
2456 shader_addline(buffer
, "MUL %s.x, %s.x, %s.y;\n", dst_name
, dst_name
, dst_name
);
2457 shader_addline(buffer
, "MIN %s.x, %s.x, %s;\n", dst_name
, dst_name
, one
);
2458 shader_addline(buffer
, "MAX result.depth, %s.x, %s;\n", dst_name
, zero
);
2461 /** Handles transforming all WINED3DSIO_M?x? opcodes for
2462 Vertex/Pixel shaders to ARB_vertex_program codes */
2463 static void shader_hw_mnxn(const struct wined3d_shader_instruction
*ins
)
2466 int nComponents
= 0;
2467 struct wined3d_shader_dst_param tmp_dst
= {{0}};
2468 struct wined3d_shader_src_param tmp_src
[2] = {{{0}}};
2469 struct wined3d_shader_instruction tmp_ins
;
2471 memset(&tmp_ins
, 0, sizeof(tmp_ins
));
2473 /* Set constants for the temporary argument */
2474 tmp_ins
.ctx
= ins
->ctx
;
2475 tmp_ins
.dst_count
= 1;
2476 tmp_ins
.dst
= &tmp_dst
;
2477 tmp_ins
.src_count
= 2;
2478 tmp_ins
.src
= tmp_src
;
2480 switch(ins
->handler_idx
)
2482 case WINED3DSIH_M4x4
:
2484 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2486 case WINED3DSIH_M4x3
:
2488 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2490 case WINED3DSIH_M3x4
:
2492 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2494 case WINED3DSIH_M3x3
:
2496 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2498 case WINED3DSIH_M3x2
:
2500 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2503 FIXME("Unhandled opcode %s.\n", debug_d3dshaderinstructionhandler(ins
->handler_idx
));
2507 tmp_dst
= ins
->dst
[0];
2508 tmp_src
[0] = ins
->src
[0];
2509 tmp_src
[1] = ins
->src
[1];
2510 for (i
= 0; i
< nComponents
; ++i
)
2512 tmp_dst
.write_mask
= WINED3DSP_WRITEMASK_0
<< i
;
2513 shader_hw_map2gl(&tmp_ins
);
2514 ++tmp_src
[1].reg
.idx
[0].offset
;
2518 static DWORD
abs_modifier(DWORD mod
, BOOL
*need_abs
)
2524 case WINED3DSPSM_NONE
: return WINED3DSPSM_ABS
;
2525 case WINED3DSPSM_NEG
: return WINED3DSPSM_ABS
;
2526 case WINED3DSPSM_BIAS
: *need_abs
= TRUE
; return WINED3DSPSM_BIAS
;
2527 case WINED3DSPSM_BIASNEG
: *need_abs
= TRUE
; return WINED3DSPSM_BIASNEG
;
2528 case WINED3DSPSM_SIGN
: *need_abs
= TRUE
; return WINED3DSPSM_SIGN
;
2529 case WINED3DSPSM_SIGNNEG
: *need_abs
= TRUE
; return WINED3DSPSM_SIGNNEG
;
2530 case WINED3DSPSM_COMP
: *need_abs
= TRUE
; return WINED3DSPSM_COMP
;
2531 case WINED3DSPSM_X2
: *need_abs
= TRUE
; return WINED3DSPSM_X2
;
2532 case WINED3DSPSM_X2NEG
: *need_abs
= TRUE
; return WINED3DSPSM_X2NEG
;
2533 case WINED3DSPSM_DZ
: *need_abs
= TRUE
; return WINED3DSPSM_DZ
;
2534 case WINED3DSPSM_DW
: *need_abs
= TRUE
; return WINED3DSPSM_DW
;
2535 case WINED3DSPSM_ABS
: return WINED3DSPSM_ABS
;
2536 case WINED3DSPSM_ABSNEG
: return WINED3DSPSM_ABS
;
2538 FIXME("Unknown modifier %u\n", mod
);
2542 static void shader_hw_scalar_op(const struct wined3d_shader_instruction
*ins
)
2544 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2545 const char *instruction
;
2546 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
2547 BOOL need_abs
= FALSE
;
2552 switch(ins
->handler_idx
)
2554 case WINED3DSIH_RSQ
: instruction
= "RSQ"; break;
2555 case WINED3DSIH_RCP
: instruction
= "RCP"; break;
2556 case WINED3DSIH_EXPP
:
2557 if (ins
->ctx
->reg_maps
->shader_version
.major
< 2)
2559 instruction
= "EXP";
2563 case WINED3DSIH_EXP
:
2564 instruction
= "EX2";
2566 case WINED3DSIH_LOG
:
2567 case WINED3DSIH_LOGP
:
2568 /* The precision requirements suggest that LOGP matches ARBvp's LOG
2569 * instruction, but notice that the output of those instructions is
2571 src0_copy
.modifiers
= abs_modifier(src0_copy
.modifiers
, &need_abs
);
2572 instruction
= "LG2";
2574 default: instruction
= "";
2575 FIXME("Unhandled opcode %s.\n", debug_d3dshaderinstructionhandler(ins
->handler_idx
));
2579 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
2581 src0_copy
.swizzle
= shader_arb_select_component(src0_copy
.swizzle
, 3);
2583 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
); /* Destination */
2584 shader_arb_get_src_param(ins
, &src0_copy
, 0, src
);
2588 shader_addline(buffer
, "ABS TA.w, %s;\n", src
);
2589 shader_addline(buffer
, "%s%s %s, TA.w;\n", instruction
, shader_arb_get_modifier(ins
), dst
);
2593 shader_addline(buffer
, "%s%s %s, %s;\n", instruction
, shader_arb_get_modifier(ins
), dst
, src
);
2598 static void shader_hw_nrm(const struct wined3d_shader_instruction
*ins
)
2600 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2603 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2604 BOOL pshader
= shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2605 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2607 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2608 shader_arb_get_src_param(ins
, &ins
->src
[0], 1 /* Use TB */, src_name
);
2610 /* In D3D, NRM of a vector with length zero returns zero. Catch this situation, as
2611 * otherwise NRM or RSQ would return NaN */
2612 if(pshader
&& priv
->target_version
>= NV3
)
2614 /* GL_NV_fragment_program2's NRM needs protection against length zero vectors too
2616 * TODO: Find out if DP3+NRM+MOV is really faster than DP3+RSQ+MUL
2618 shader_addline(buffer
, "DP3C TA, %s, %s;\n", src_name
, src_name
);
2619 shader_addline(buffer
, "NRM%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
);
2620 shader_addline(buffer
, "MOV %s (EQ), %s;\n", dst_name
, zero
);
2622 else if(priv
->target_version
>= NV2
)
2624 shader_addline(buffer
, "DP3C TA.x, %s, %s;\n", src_name
, src_name
);
2625 shader_addline(buffer
, "RSQ TA.x (NE), TA.x;\n");
2626 shader_addline(buffer
, "MUL%s %s, %s, TA.x;\n", shader_arb_get_modifier(ins
), dst_name
,
2631 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2633 shader_addline(buffer
, "DP3 TA.x, %s, %s;\n", src_name
, src_name
);
2634 /* Pass any non-zero value to RSQ if the input vector has a length of zero. The
2635 * RSQ result doesn't matter, as long as multiplying it by 0 returns 0.
2637 shader_addline(buffer
, "SGE TA.y, -TA.x, %s;\n", zero
);
2638 shader_addline(buffer
, "MAD TA.x, %s, TA.y, TA.x;\n", one
);
2640 shader_addline(buffer
, "RSQ TA.x, TA.x;\n");
2641 /* dst.w = src[0].w * 1 / (src.x^2 + src.y^2 + src.z^2)^(1/2) according to msdn*/
2642 shader_addline(buffer
, "MUL%s %s, %s, TA.x;\n", shader_arb_get_modifier(ins
), dst_name
,
2647 static void shader_hw_lrp(const struct wined3d_shader_instruction
*ins
)
2649 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2651 char src_name
[3][50];
2653 /* ARB_fragment_program has a convenient LRP instruction */
2654 if(shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)) {
2655 shader_hw_map2gl(ins
);
2659 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2660 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
2661 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
2662 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
2664 shader_addline(buffer
, "SUB TA, %s, %s;\n", src_name
[1], src_name
[2]);
2665 shader_addline(buffer
, "MAD%s %s, %s, TA, %s;\n", shader_arb_get_modifier(ins
),
2666 dst_name
, src_name
[0], src_name
[2]);
2669 static void shader_hw_sincos(const struct wined3d_shader_instruction
*ins
)
2671 /* This instruction exists in ARB, but the d3d instruction takes two extra parameters which
2672 * must contain fixed constants. So we need a separate function to filter those constants and
2675 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2676 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2677 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2679 char src_name0
[50], src_name1
[50], src_name2
[50];
2682 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
2683 if(shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)) {
2684 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2685 /* No modifiers are supported on SCS */
2686 shader_addline(buffer
, "SCS %s, %s;\n", dst_name
, src_name0
);
2688 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
)
2690 shader_arb_get_register_name(ins
, &dst
->reg
, src_name0
, &is_color
);
2691 shader_addline(buffer
, "MOV_SAT %s, %s;\n", dst_name
, src_name0
);
2693 } else if(priv
->target_version
>= NV2
) {
2694 shader_arb_get_register_name(ins
, &dst
->reg
, dst_name
, &is_color
);
2696 /* Sincos writemask must be .x, .y or .xy */
2697 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
)
2698 shader_addline(buffer
, "COS%s %s.x, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name0
);
2699 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
)
2700 shader_addline(buffer
, "SIN%s %s.y, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name0
);
2702 /* Approximate sine and cosine with a taylor series, as per math textbook. The application passes 8
2703 * helper constants(D3DSINCOSCONST1 and D3DSINCOSCONST2) in src1 and src2.
2705 * sin(x) = x - x^3/3! + x^5/5! - x^7/7! + ...
2706 * cos(x) = 1 - x^2/2! + x^4/4! - x^6/6! + ...
2708 * The constants we get are:
2710 * +1 +1, -1 -1 +1 +1 -1 -1
2711 * ---- , ---- , ---- , ----- , ----- , ----- , ------
2712 * 1!*2 2!*4 3!*8 4!*16 5!*32 6!*64 7!*128
2714 * If used with x^2, x^3, x^4 etc they calculate sin(x/2) and cos(x/2):
2718 * (x/2)^4 = x^4 / 16
2719 * (x/2)^5 = x^5 / 32
2722 * To get the final result:
2723 * sin(x) = 2 * sin(x/2) * cos(x/2)
2724 * cos(x) = cos(x/2)^2 - sin(x/2)^2
2725 * (from sin(x+y) and cos(x+y) rules)
2727 * As per MSDN, dst.z is undefined after the operation, and so is
2728 * dst.x and dst.y if they're masked out by the writemask. Ie
2729 * sincos dst.y, src1, c0, c1
2730 * returns the sine in dst.y. dst.x and dst.z are undefined, dst.w is not touched. The assembler
2731 * vsa.exe also stops with an error if the dest register is the same register as the source
2732 * register. This means we can use dest.xyz as temporary storage. The assembler vsa.exe output also
2733 * indicates that sincos consumes 8 instruction slots in vs_2_0(and, strangely, in vs_3_0).
2735 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
2736 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name2
);
2737 shader_arb_get_register_name(ins
, &dst
->reg
, dst_name
, &is_color
);
2739 shader_addline(buffer
, "MUL %s.x, %s, %s;\n", dst_name
, src_name0
, src_name0
); /* x ^ 2 */
2740 shader_addline(buffer
, "MUL TA.y, %s.x, %s;\n", dst_name
, src_name0
); /* x ^ 3 */
2741 shader_addline(buffer
, "MUL %s.y, TA.y, %s;\n", dst_name
, src_name0
); /* x ^ 4 */
2742 shader_addline(buffer
, "MUL TA.z, %s.y, %s;\n", dst_name
, src_name0
); /* x ^ 5 */
2743 shader_addline(buffer
, "MUL %s.z, TA.z, %s;\n", dst_name
, src_name0
); /* x ^ 6 */
2744 shader_addline(buffer
, "MUL TA.w, %s.z, %s;\n", dst_name
, src_name0
); /* x ^ 7 */
2748 * Unfortunately we don't get the constants in a DP4-capable form. Is there a way to
2749 * properly merge that with MULs in the code above?
2750 * The swizzles .yz and xw however fit into the .yzxw swizzle added to ps_2_0. Maybe
2751 * we can merge the sine and cosine MAD rows to calculate them together.
2753 shader_addline(buffer
, "MUL TA.x, %s, %s.w;\n", src_name0
, src_name2
); /* x^1, +1/(1!*2) */
2754 shader_addline(buffer
, "MAD TA.x, TA.y, %s.x, TA.x;\n", src_name2
); /* -1/(3!*8) */
2755 shader_addline(buffer
, "MAD TA.x, TA.z, %s.w, TA.x;\n", src_name1
); /* +1/(5!*32) */
2756 shader_addline(buffer
, "MAD TA.x, TA.w, %s.x, TA.x;\n", src_name1
); /* -1/(7!*128) */
2759 shader_addline(buffer
, "MAD TA.y, %s.x, %s.y, %s.z;\n", dst_name
, src_name2
, src_name2
); /* -1/(2!*4), +1.0 */
2760 shader_addline(buffer
, "MAD TA.y, %s.y, %s.z, TA.y;\n", dst_name
, src_name1
); /* +1/(4!*16) */
2761 shader_addline(buffer
, "MAD TA.y, %s.z, %s.y, TA.y;\n", dst_name
, src_name1
); /* -1/(6!*64) */
2763 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
) {
2765 shader_addline(buffer
, "MUL TA.z, TA.y, TA.y;\n");
2766 shader_addline(buffer
, "MAD %s.x, -TA.x, TA.x, TA.z;\n", dst_name
);
2768 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
) {
2770 shader_addline(buffer
, "MUL %s.y, TA.x, TA.y;\n", dst_name
);
2771 shader_addline(buffer
, "ADD %s.y, %s.y, %s.y;\n", dst_name
, dst_name
, dst_name
);
2776 static void shader_hw_sgn(const struct wined3d_shader_instruction
*ins
)
2778 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2781 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
2783 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2784 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
);
2786 /* SGN is only valid in vertex shaders */
2787 if(ctx
->target_version
>= NV2
) {
2788 shader_addline(buffer
, "SSG%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
);
2792 /* If SRC > 0.0, -SRC < SRC = TRUE, otherwise false.
2793 * if SRC < 0.0, SRC < -SRC = TRUE. If neither is true, src = 0.0
2795 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
) {
2796 shader_addline(buffer
, "SLT %s, -%s, %s;\n", dst_name
, src_name
, src_name
);
2798 /* src contains TA? Write to the dest first. This won't overwrite our destination.
2799 * Then use TA, and calculate the final result
2801 * Not reading from TA? Store the first result in TA to avoid overwriting the
2802 * destination if src reg = dst reg
2804 if(strstr(src_name
, "TA"))
2806 shader_addline(buffer
, "SLT %s, %s, -%s;\n", dst_name
, src_name
, src_name
);
2807 shader_addline(buffer
, "SLT TA, -%s, %s;\n", src_name
, src_name
);
2808 shader_addline(buffer
, "ADD %s, %s, -TA;\n", dst_name
, dst_name
);
2812 shader_addline(buffer
, "SLT TA, -%s, %s;\n", src_name
, src_name
);
2813 shader_addline(buffer
, "SLT %s, %s, -%s;\n", dst_name
, src_name
, src_name
);
2814 shader_addline(buffer
, "ADD %s, TA, -%s;\n", dst_name
, dst_name
);
2819 static void shader_hw_dsy(const struct wined3d_shader_instruction
*ins
)
2821 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2827 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
);
2828 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src
);
2829 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2831 shader_addline(buffer
, "DDY %s, %s;\n", dst
, src
);
2832 shader_addline(buffer
, "MUL%s %s, %s, ycorrection.y;\n", shader_arb_get_modifier(ins
), dst
, dst_name
);
2835 static void shader_hw_pow(const struct wined3d_shader_instruction
*ins
)
2837 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2838 char src0
[50], src1
[50], dst
[50];
2839 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
2840 BOOL need_abs
= FALSE
;
2841 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2842 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2844 /* POW operates on the absolute value of the input */
2845 src0_copy
.modifiers
= abs_modifier(src0_copy
.modifiers
, &need_abs
);
2847 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
);
2848 shader_arb_get_src_param(ins
, &src0_copy
, 0, src0
);
2849 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src1
);
2852 shader_addline(buffer
, "ABS TA.x, %s;\n", src0
);
2854 shader_addline(buffer
, "MOV TA.x, %s;\n", src0
);
2856 if (priv
->target_version
>= NV2
)
2858 shader_addline(buffer
, "MOVC TA.y, %s;\n", src1
);
2859 shader_addline(buffer
, "POW%s %s, TA.x, TA.y;\n", shader_arb_get_modifier(ins
), dst
);
2860 shader_addline(buffer
, "MOV %s (EQ.y), %s;\n", dst
, one
);
2864 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2865 const char *flt_eps
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_EPS
);
2867 shader_addline(buffer
, "ABS TA.y, %s;\n", src1
);
2868 shader_addline(buffer
, "SGE TA.y, -TA.y, %s;\n", zero
);
2869 /* Possibly add flt_eps to avoid getting float special values */
2870 shader_addline(buffer
, "MAD TA.z, TA.y, %s, %s;\n", flt_eps
, src1
);
2871 shader_addline(buffer
, "POW%s TA.x, TA.x, TA.z;\n", shader_arb_get_modifier(ins
));
2872 shader_addline(buffer
, "MAD TA.x, -TA.x, TA.y, TA.x;\n");
2873 shader_addline(buffer
, "MAD %s, TA.y, %s, TA.x;\n", dst
, one
);
2877 static void shader_hw_loop(const struct wined3d_shader_instruction
*ins
)
2879 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2881 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2884 shader_arb_get_src_param(ins
, &ins
->src
[1], 0, src_name
);
2888 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2889 struct list
*e
= list_head(&priv
->control_frames
);
2890 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2892 if(priv
->loop_depth
> 1) shader_addline(buffer
, "PUSHA aL;\n");
2893 /* The constant loader makes sure to load -1 into iX.w */
2894 shader_addline(buffer
, "ARLC aL, %s.xywz;\n", src_name
);
2895 shader_addline(buffer
, "BRA loop_%u_end (LE.x);\n", control_frame
->no
.loop
);
2896 shader_addline(buffer
, "loop_%u_start:\n", control_frame
->no
.loop
);
2900 shader_addline(buffer
, "LOOP %s;\n", src_name
);
2904 static void shader_hw_rep(const struct wined3d_shader_instruction
*ins
)
2906 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2908 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2910 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
);
2912 /* The constant loader makes sure to load -1 into iX.w */
2915 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2916 struct list
*e
= list_head(&priv
->control_frames
);
2917 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2919 if(priv
->loop_depth
> 1) shader_addline(buffer
, "PUSHA aL;\n");
2921 shader_addline(buffer
, "ARLC aL, %s.xywz;\n", src_name
);
2922 shader_addline(buffer
, "BRA loop_%u_end (LE.x);\n", control_frame
->no
.loop
);
2923 shader_addline(buffer
, "loop_%u_start:\n", control_frame
->no
.loop
);
2927 shader_addline(buffer
, "REP %s;\n", src_name
);
2931 static void shader_hw_endloop(const struct wined3d_shader_instruction
*ins
)
2933 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2934 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2938 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2939 struct list
*e
= list_head(&priv
->control_frames
);
2940 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2942 shader_addline(buffer
, "ARAC aL.xy, aL;\n");
2943 shader_addline(buffer
, "BRA loop_%u_start (GT.x);\n", control_frame
->no
.loop
);
2944 shader_addline(buffer
, "loop_%u_end:\n", control_frame
->no
.loop
);
2946 if(priv
->loop_depth
> 1) shader_addline(buffer
, "POPA aL;\n");
2950 shader_addline(buffer
, "ENDLOOP;\n");
2954 static void shader_hw_endrep(const struct wined3d_shader_instruction
*ins
)
2956 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2957 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2961 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2962 struct list
*e
= list_head(&priv
->control_frames
);
2963 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2965 shader_addline(buffer
, "ARAC aL.xy, aL;\n");
2966 shader_addline(buffer
, "BRA loop_%u_start (GT.x);\n", control_frame
->no
.loop
);
2967 shader_addline(buffer
, "loop_%u_end:\n", control_frame
->no
.loop
);
2969 if(priv
->loop_depth
> 1) shader_addline(buffer
, "POPA aL;\n");
2973 shader_addline(buffer
, "ENDREP;\n");
2977 static const struct control_frame
*find_last_loop(const struct shader_arb_ctx_priv
*priv
)
2979 struct control_frame
*control_frame
;
2981 LIST_FOR_EACH_ENTRY(control_frame
, &priv
->control_frames
, struct control_frame
, entry
)
2983 if(control_frame
->type
== LOOP
|| control_frame
->type
== REP
) return control_frame
;
2985 ERR("Could not find loop for break\n");
2989 static void shader_hw_break(const struct wined3d_shader_instruction
*ins
)
2991 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2992 const struct control_frame
*control_frame
= find_last_loop(ins
->ctx
->backend_data
);
2993 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2997 shader_addline(buffer
, "BRA loop_%u_end;\n", control_frame
->no
.loop
);
3001 shader_addline(buffer
, "BRK;\n");
3005 static const char *get_compare(enum wined3d_shader_rel_op op
)
3009 case WINED3D_SHADER_REL_OP_GT
: return "GT";
3010 case WINED3D_SHADER_REL_OP_EQ
: return "EQ";
3011 case WINED3D_SHADER_REL_OP_GE
: return "GE";
3012 case WINED3D_SHADER_REL_OP_LT
: return "LT";
3013 case WINED3D_SHADER_REL_OP_NE
: return "NE";
3014 case WINED3D_SHADER_REL_OP_LE
: return "LE";
3016 FIXME("Unrecognized operator %#x.\n", op
);
3021 static enum wined3d_shader_rel_op
invert_compare(enum wined3d_shader_rel_op op
)
3025 case WINED3D_SHADER_REL_OP_GT
: return WINED3D_SHADER_REL_OP_LE
;
3026 case WINED3D_SHADER_REL_OP_EQ
: return WINED3D_SHADER_REL_OP_NE
;
3027 case WINED3D_SHADER_REL_OP_GE
: return WINED3D_SHADER_REL_OP_LT
;
3028 case WINED3D_SHADER_REL_OP_LT
: return WINED3D_SHADER_REL_OP_GE
;
3029 case WINED3D_SHADER_REL_OP_NE
: return WINED3D_SHADER_REL_OP_EQ
;
3030 case WINED3D_SHADER_REL_OP_LE
: return WINED3D_SHADER_REL_OP_GT
;
3032 FIXME("Unrecognized operator %#x.\n", op
);
3037 static void shader_hw_breakc(const struct wined3d_shader_instruction
*ins
)
3039 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3040 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3041 const struct control_frame
*control_frame
= find_last_loop(ins
->ctx
->backend_data
);
3044 const char *comp
= get_compare(ins
->flags
);
3046 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
3047 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
3051 /* SUBC CC, src0, src1" works only in pixel shaders, so use TA to throw
3052 * away the subtraction result
3054 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3055 shader_addline(buffer
, "BRA loop_%u_end (%s.x);\n", control_frame
->no
.loop
, comp
);
3059 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3060 shader_addline(buffer
, "BRK (%s.x);\n", comp
);
3064 static void shader_hw_ifc(const struct wined3d_shader_instruction
*ins
)
3066 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3067 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3068 struct list
*e
= list_head(&priv
->control_frames
);
3069 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3073 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3075 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
3076 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
3080 /* Invert the flag. We jump to the else label if the condition is NOT true */
3081 comp
= get_compare(invert_compare(ins
->flags
));
3082 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3083 shader_addline(buffer
, "BRA ifc_%u_else (%s.x);\n", control_frame
->no
.ifc
, comp
);
3087 comp
= get_compare(ins
->flags
);
3088 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3089 shader_addline(buffer
, "IF %s.x;\n", comp
);
3093 static void shader_hw_else(const struct wined3d_shader_instruction
*ins
)
3095 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3096 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3097 struct list
*e
= list_head(&priv
->control_frames
);
3098 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3099 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3103 shader_addline(buffer
, "BRA ifc_%u_endif;\n", control_frame
->no
.ifc
);
3104 shader_addline(buffer
, "ifc_%u_else:\n", control_frame
->no
.ifc
);
3105 control_frame
->had_else
= TRUE
;
3109 shader_addline(buffer
, "ELSE;\n");
3113 static void shader_hw_endif(const struct wined3d_shader_instruction
*ins
)
3115 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3116 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3117 struct list
*e
= list_head(&priv
->control_frames
);
3118 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3119 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3123 if(control_frame
->had_else
)
3125 shader_addline(buffer
, "ifc_%u_endif:\n", control_frame
->no
.ifc
);
3129 shader_addline(buffer
, "#No else branch. else is endif\n");
3130 shader_addline(buffer
, "ifc_%u_else:\n", control_frame
->no
.ifc
);
3135 shader_addline(buffer
, "ENDIF;\n");
3139 static void shader_hw_texldd(const struct wined3d_shader_instruction
*ins
)
3141 DWORD sampler_idx
= ins
->src
[1].reg
.idx
[0].offset
;
3143 char reg_src
[3][40];
3144 WORD flags
= TEX_DERIV
;
3146 shader_arb_get_dst_param(ins
, &ins
->dst
[0], reg_dest
);
3147 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_src
[0]);
3148 shader_arb_get_src_param(ins
, &ins
->src
[2], 1, reg_src
[1]);
3149 shader_arb_get_src_param(ins
, &ins
->src
[3], 2, reg_src
[2]);
3151 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) flags
|= TEX_PROJ
;
3152 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) flags
|= TEX_BIAS
;
3154 shader_hw_sample(ins
, sampler_idx
, reg_dest
, reg_src
[0], flags
, reg_src
[1], reg_src
[2]);
3157 static void shader_hw_texldl(const struct wined3d_shader_instruction
*ins
)
3159 DWORD sampler_idx
= ins
->src
[1].reg
.idx
[0].offset
;
3162 WORD flags
= TEX_LOD
;
3164 shader_arb_get_dst_param(ins
, &ins
->dst
[0], reg_dest
);
3165 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_coord
);
3167 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) flags
|= TEX_PROJ
;
3168 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) flags
|= TEX_BIAS
;
3170 shader_hw_sample(ins
, sampler_idx
, reg_dest
, reg_coord
, flags
, NULL
, NULL
);
3173 static void shader_hw_label(const struct wined3d_shader_instruction
*ins
)
3175 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3176 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3178 priv
->in_main_func
= FALSE
;
3179 /* Call instructions activate the NV extensions, not labels and rets. If there is an uncalled
3180 * subroutine, don't generate a label that will make GL complain
3182 if(priv
->target_version
== ARB
) return;
3184 shader_addline(buffer
, "l%u:\n", ins
->src
[0].reg
.idx
[0].offset
);
3187 static void vshader_add_footer(struct shader_arb_ctx_priv
*priv_ctx
,
3188 const struct arb_vshader_private
*shader_data
, const struct arb_vs_compile_args
*args
,
3189 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
,
3190 struct wined3d_string_buffer
*buffer
)
3194 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
3195 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
3196 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
3197 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
3199 if (args
->super
.fog_src
== VS_FOG_Z
)
3201 shader_addline(buffer
, "MOV result.fogcoord, TMP_OUT.z;\n");
3207 /* posFixup.x is always 1.0, so we can safely use it */
3208 shader_addline(buffer
, "ADD result.fogcoord, posFixup.x, -posFixup.x;\n");
3212 /* Clamp fogcoord */
3213 const char *zero
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ZERO
);
3214 const char *one
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ONE
);
3216 shader_addline(buffer
, "MIN TMP_FOGCOORD.x, TMP_FOGCOORD.x, %s;\n", one
);
3217 shader_addline(buffer
, "MAX result.fogcoord.x, TMP_FOGCOORD.x, %s;\n", zero
);
3221 /* Clipplanes are always stored without y inversion */
3222 if (use_nv_clip(gl_info
) && priv_ctx
->target_version
>= NV2
)
3224 if (args
->super
.clip_enabled
)
3226 for (i
= 0; i
< priv_ctx
->vs_clipplanes
; i
++)
3228 shader_addline(buffer
, "DP4 result.clip[%u].x, TMP_OUT, state.clip[%u].plane;\n", i
, i
);
3232 else if (args
->clip
.boolclip
.clip_texcoord
)
3234 static const char component
[4] = {'x', 'y', 'z', 'w'};
3235 unsigned int cur_clip
= 0;
3236 const char *zero
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_ZERO
);
3238 for (i
= 0; i
< gl_info
->limits
.user_clip_distances
; ++i
)
3240 if (args
->clip
.boolclip
.clipplane_mask
& (1u << i
))
3242 shader_addline(buffer
, "DP4 TA.%c, TMP_OUT, state.clip[%u].plane;\n",
3243 component
[cur_clip
++], i
);
3249 shader_addline(buffer
, "MOV TA, %s;\n", zero
);
3252 shader_addline(buffer
, "MOV TA.yzw, %s;\n", zero
);
3255 shader_addline(buffer
, "MOV TA.zw, %s;\n", zero
);
3258 shader_addline(buffer
, "MOV TA.w, %s;\n", zero
);
3261 shader_addline(buffer
, "MOV result.texcoord[%u], TA;\n",
3262 args
->clip
.boolclip
.clip_texcoord
- 1);
3265 /* Write the final position.
3267 * OpenGL coordinates specify the center of the pixel while d3d coords specify
3268 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
3269 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
3270 * contains 1.0 to allow a mad, but arb vs swizzles are too restricted for that.
3272 if (!gl_info
->supported
[ARB_CLIP_CONTROL
])
3274 shader_addline(buffer
, "MUL TA, posFixup, TMP_OUT.w;\n");
3275 shader_addline(buffer
, "ADD TMP_OUT.x, TMP_OUT.x, TA.z;\n");
3276 shader_addline(buffer
, "MAD TMP_OUT.y, TMP_OUT.y, posFixup.y, TA.w;\n");
3278 /* Z coord [0;1]->[-1;1] mapping, see comment in
3279 * get_projection_matrix() in utils.c. */
3280 if (need_helper_const(shader_data
, reg_maps
, gl_info
))
3282 const char *two
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_TWO
);
3283 shader_addline(buffer
, "MAD TMP_OUT.z, TMP_OUT.z, %s, -TMP_OUT.w;\n", two
);
3287 shader_addline(buffer
, "ADD TMP_OUT.z, TMP_OUT.z, TMP_OUT.z;\n");
3288 shader_addline(buffer
, "ADD TMP_OUT.z, TMP_OUT.z, -TMP_OUT.w;\n");
3292 shader_addline(buffer
, "MOV result.position, TMP_OUT;\n");
3294 priv_ctx
->footer_written
= TRUE
;
3297 static void shader_hw_ret(const struct wined3d_shader_instruction
*ins
)
3299 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3300 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3301 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
3302 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3304 if(priv
->target_version
== ARB
) return;
3308 if (priv
->in_main_func
) vshader_add_footer(priv
, shader
->backend_data
,
3309 priv
->cur_vs_args
, ins
->ctx
->reg_maps
, ins
->ctx
->gl_info
, buffer
);
3312 shader_addline(buffer
, "RET;\n");
3315 static void shader_hw_call(const struct wined3d_shader_instruction
*ins
)
3317 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3318 shader_addline(buffer
, "CAL l%u;\n", ins
->src
[0].reg
.idx
[0].offset
);
3321 static BOOL
shader_arb_compile(const struct wined3d_gl_info
*gl_info
, GLenum target
, const char *src
)
3323 const char *ptr
, *line
;
3326 if (TRACE_ON(d3d_shader
))
3329 while ((line
= get_line(&ptr
))) TRACE_(d3d_shader
)(" %.*s", (int)(ptr
- line
), line
);
3332 GL_EXTCALL(glProgramStringARB(target
, GL_PROGRAM_FORMAT_ASCII_ARB
, strlen(src
), src
));
3333 checkGLcall("glProgramStringARB()");
3335 if (FIXME_ON(d3d_shader
))
3337 gl_info
->gl_ops
.gl
.p_glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
3340 FIXME_(d3d_shader
)("Program error at position %d: %s\n\n", pos
,
3341 debugstr_a((const char *)gl_info
->gl_ops
.gl
.p_glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
3343 while ((line
= get_line(&ptr
))) FIXME_(d3d_shader
)(" %.*s", (int)(ptr
- line
), line
);
3344 FIXME_(d3d_shader
)("\n");
3350 if (WARN_ON(d3d_perf
))
3352 GL_EXTCALL(glGetProgramivARB(target
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
3353 checkGLcall("glGetProgramivARB()");
3355 WARN_(d3d_perf
)("Program exceeds native resource limits.\n");
3361 /* Context activation is done by the caller. */
3362 static GLuint
create_arb_blt_vertex_program(const struct wined3d_gl_info
*gl_info
)
3364 GLuint program_id
= 0;
3366 static const char blt_vprogram
[] =
3368 "PARAM c[1] = { { 1, 0.5 } };\n"
3369 "MOV result.position, vertex.position;\n"
3370 "MOV result.color, c[0].x;\n"
3371 "MOV result.texcoord[0], vertex.texcoord[0];\n"
3374 GL_EXTCALL(glGenProgramsARB(1, &program_id
));
3375 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, program_id
));
3376 shader_arb_compile(gl_info
, GL_VERTEX_PROGRAM_ARB
, blt_vprogram
);
3381 /* Context activation is done by the caller. */
3382 static GLuint
create_arb_blt_fragment_program(const struct wined3d_gl_info
*gl_info
,
3383 enum wined3d_gl_resource_type tex_type
, BOOL masked
)
3385 GLuint program_id
= 0;
3386 const char *fprogram
;
3388 static const char * const blt_fprograms_full
[WINED3D_GL_RES_TYPE_COUNT
] =
3390 /* WINED3D_GL_RES_TYPE_TEX_1D */
3392 /* WINED3D_GL_RES_TYPE_TEX_2D */
3395 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
3396 "MOV result.depth.z, R0.x;\n"
3398 /* WINED3D_GL_RES_TYPE_TEX_3D */
3400 /* WINED3D_GL_RES_TYPE_TEX_CUBE */
3403 "TEX R0.x, fragment.texcoord[0], texture[0], CUBE;\n"
3404 "MOV result.depth.z, R0.x;\n"
3406 /* WINED3D_GL_RES_TYPE_TEX_RECT */
3409 "TEX R0.x, fragment.texcoord[0], texture[0], RECT;\n"
3410 "MOV result.depth.z, R0.x;\n"
3412 /* WINED3D_GL_RES_TYPE_BUFFER */
3414 /* WINED3D_GL_RES_TYPE_RB */
3418 static const char * const blt_fprograms_masked
[WINED3D_GL_RES_TYPE_COUNT
] =
3420 /* WINED3D_GL_RES_TYPE_TEX_1D */
3422 /* WINED3D_GL_RES_TYPE_TEX_2D */
3424 "PARAM mask = program.local[0];\n"
3426 "SLT R0.xy, fragment.position, mask.zwzw;\n"
3427 "MUL R0.x, R0.x, R0.y;\n"
3429 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
3430 "MOV result.depth.z, R0.x;\n"
3432 /* WINED3D_GL_RES_TYPE_TEX_3D */
3434 /* WINED3D_GL_RES_TYPE_TEX_CUBE */
3436 "PARAM mask = program.local[0];\n"
3438 "SLT R0.xy, fragment.position, mask.zwzw;\n"
3439 "MUL R0.x, R0.x, R0.y;\n"
3441 "TEX R0.x, fragment.texcoord[0], texture[0], CUBE;\n"
3442 "MOV result.depth.z, R0.x;\n"
3444 /* WINED3D_GL_RES_TYPE_TEX_RECT */
3446 "PARAM mask = program.local[0];\n"
3448 "SLT R0.xy, fragment.position, mask.zwzw;\n"
3449 "MUL R0.x, R0.x, R0.y;\n"
3451 "TEX R0.x, fragment.texcoord[0], texture[0], RECT;\n"
3452 "MOV result.depth.z, R0.x;\n"
3454 /* WINED3D_GL_RES_TYPE_BUFFER */
3456 /* WINED3D_GL_RES_TYPE_RB */
3460 fprogram
= masked
? blt_fprograms_masked
[tex_type
] : blt_fprograms_full
[tex_type
];
3463 FIXME("tex_type %#x not supported, falling back to 2D\n", tex_type
);
3464 tex_type
= WINED3D_GL_RES_TYPE_TEX_2D
;
3465 fprogram
= masked
? blt_fprograms_masked
[tex_type
] : blt_fprograms_full
[tex_type
];
3468 GL_EXTCALL(glGenProgramsARB(1, &program_id
));
3469 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, program_id
));
3470 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, fprogram
);
3475 static void arbfp_add_sRGB_correction(struct wined3d_string_buffer
*buffer
, const char *fragcolor
,
3476 const char *tmp1
, const char *tmp2
, const char *tmp3
, const char *tmp4
, BOOL condcode
)
3478 /* Perform sRGB write correction. See GLX_EXT_framebuffer_sRGB */
3482 /* Sigh. MOVC CC doesn't work, so use one of the temps as dummy dest */
3483 shader_addline(buffer
, "SUBC %s, %s.x, srgb_consts1.x;\n", tmp1
, fragcolor
);
3484 /* Calculate the > 0.0031308 case */
3485 shader_addline(buffer
, "POW %s.x (GE), %s.x, srgb_consts0.x;\n", fragcolor
, fragcolor
);
3486 shader_addline(buffer
, "POW %s.y (GE), %s.y, srgb_consts0.x;\n", fragcolor
, fragcolor
);
3487 shader_addline(buffer
, "POW %s.z (GE), %s.z, srgb_consts0.x;\n", fragcolor
, fragcolor
);
3488 shader_addline(buffer
, "MUL %s.xyz (GE), %s, srgb_consts0.y;\n", fragcolor
, fragcolor
);
3489 shader_addline(buffer
, "SUB %s.xyz (GE), %s, srgb_consts0.z;\n", fragcolor
, fragcolor
);
3490 /* Calculate the < case */
3491 shader_addline(buffer
, "MUL %s.xyz (LT), srgb_consts0.w, %s;\n", fragcolor
, fragcolor
);
3495 /* Calculate the > 0.0031308 case */
3496 shader_addline(buffer
, "POW %s.x, %s.x, srgb_consts0.x;\n", tmp1
, fragcolor
);
3497 shader_addline(buffer
, "POW %s.y, %s.y, srgb_consts0.x;\n", tmp1
, fragcolor
);
3498 shader_addline(buffer
, "POW %s.z, %s.z, srgb_consts0.x;\n", tmp1
, fragcolor
);
3499 shader_addline(buffer
, "MUL %s, %s, srgb_consts0.y;\n", tmp1
, tmp1
);
3500 shader_addline(buffer
, "SUB %s, %s, srgb_consts0.z;\n", tmp1
, tmp1
);
3501 /* Calculate the < case */
3502 shader_addline(buffer
, "MUL %s, srgb_consts0.w, %s;\n", tmp2
, fragcolor
);
3503 /* Get 1.0 / 0.0 masks for > 0.0031308 and < 0.0031308 */
3504 shader_addline(buffer
, "SLT %s, srgb_consts1.x, %s;\n", tmp3
, fragcolor
);
3505 shader_addline(buffer
, "SGE %s, srgb_consts1.x, %s;\n", tmp4
, fragcolor
);
3506 /* Store the components > 0.0031308 in the destination */
3507 shader_addline(buffer
, "MUL %s.xyz, %s, %s;\n", fragcolor
, tmp1
, tmp3
);
3508 /* Add the components that are < 0.0031308 */
3509 shader_addline(buffer
, "MAD %s.xyz, %s, %s, %s;\n", fragcolor
, tmp2
, tmp4
, fragcolor
);
3510 /* Move everything into result.color at once. Nvidia hardware cannot handle partial
3511 * result.color writes(.rgb first, then .a), or handle overwriting already written
3512 * components. The assembler uses a temporary register in this case, which is usually
3513 * not allocated from one of our registers that were used earlier.
3516 /* [0.0;1.0] clamping. Not needed, this is done implicitly */
3519 static const DWORD
*find_loop_control_values(const struct wined3d_shader
*shader
, DWORD idx
)
3521 const struct wined3d_shader_lconst
*constant
;
3523 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsI
, struct wined3d_shader_lconst
, entry
)
3525 if (constant
->idx
== idx
)
3527 return constant
->value
;
3533 static void init_ps_input(const struct wined3d_shader
*shader
,
3534 const struct arb_ps_compile_args
*args
, struct shader_arb_ctx_priv
*priv
)
3536 static const char * const texcoords
[8] =
3538 "fragment.texcoord[0]", "fragment.texcoord[1]", "fragment.texcoord[2]", "fragment.texcoord[3]",
3539 "fragment.texcoord[4]", "fragment.texcoord[5]", "fragment.texcoord[6]", "fragment.texcoord[7]"
3542 const struct wined3d_shader_signature_element
*input
;
3543 const char *semantic_name
;
3546 switch(args
->super
.vp_mode
)
3548 case pretransformed
:
3550 /* The pixelshader has to collect the varyings on its own. In any case properly load
3551 * color0 and color1. In the case of pretransformed vertices also load texcoords. Set
3552 * other attribs to 0.0.
3554 * For fixedfunction this behavior is correct, according to the tests. For pretransformed
3555 * we'd either need a replacement shader that can load other attribs like BINORMAL, or
3556 * load the texcoord attrib pointers to match the pixel shader signature
3558 for (i
= 0; i
< shader
->input_signature
.element_count
; ++i
)
3560 input
= &shader
->input_signature
.elements
[i
];
3561 if (!(semantic_name
= input
->semantic_name
))
3563 semantic_idx
= input
->semantic_idx
;
3565 if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_COLOR
))
3568 priv
->ps_input
[input
->register_idx
] = "fragment.color.primary";
3569 else if (semantic_idx
== 1)
3570 priv
->ps_input
[input
->register_idx
] = "fragment.color.secondary";
3572 priv
->ps_input
[input
->register_idx
] = "0.0";
3574 else if (args
->super
.vp_mode
== fixedfunction
)
3576 priv
->ps_input
[input
->register_idx
] = "0.0";
3578 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
3580 if (semantic_idx
< 8)
3581 priv
->ps_input
[input
->register_idx
] = texcoords
[semantic_idx
];
3583 priv
->ps_input
[input
->register_idx
] = "0.0";
3585 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_FOG
))
3588 priv
->ps_input
[input
->register_idx
] = "fragment.fogcoord";
3590 priv
->ps_input
[input
->register_idx
] = "0.0";
3594 priv
->ps_input
[input
->register_idx
] = "0.0";
3597 TRACE("v%u, semantic %s%u is %s\n", input
->register_idx
,
3598 semantic_name
, semantic_idx
, priv
->ps_input
[input
->register_idx
]);
3603 /* That one is easy. The vertex shaders provide v0-v7 in fragment.texcoord and v8 and v9 in
3606 for(i
= 0; i
< 8; i
++)
3608 priv
->ps_input
[i
] = texcoords
[i
];
3610 priv
->ps_input
[8] = "fragment.color.primary";
3611 priv
->ps_input
[9] = "fragment.color.secondary";
3616 static void arbfp_add_linear_fog(struct wined3d_string_buffer
*buffer
,
3617 const char *fragcolor
, const char *tmp
)
3619 shader_addline(buffer
, "SUB %s.x, state.fog.params.z, fragment.fogcoord.x;\n", tmp
);
3620 shader_addline(buffer
, "MUL_SAT %s.x, %s.x, state.fog.params.w;\n", tmp
, tmp
);
3621 shader_addline(buffer
, "LRP %s.rgb, %s.x, %s, state.fog.color;\n", fragcolor
, tmp
, fragcolor
);
3624 /* Context activation is done by the caller. */
3625 static GLuint
shader_arb_generate_pshader(const struct wined3d_shader
*shader
,
3626 const struct wined3d_gl_info
*gl_info
, struct wined3d_string_buffer
*buffer
,
3627 const struct arb_ps_compile_args
*args
, struct arb_ps_compiled_shader
*compiled
)
3629 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
3630 const DWORD
*function
= shader
->function
;
3633 DWORD next_local
= 0;
3634 struct shader_arb_ctx_priv priv_ctx
;
3635 BOOL dcl_td
= FALSE
;
3636 BOOL want_nv_prog
= FALSE
;
3637 struct arb_pshader_private
*shader_priv
= shader
->backend_data
;
3639 BOOL custom_linear_fog
= FALSE
;
3643 unsigned int i
, found
= 0;
3645 for (i
= 0, map
= reg_maps
->temporary
; map
; map
>>= 1, ++i
)
3648 || (shader
->u
.ps
.color0_mov
&& i
== shader
->u
.ps
.color0_reg
)
3649 || (reg_maps
->shader_version
.major
< 2 && !i
))
3652 sprintf(srgbtmp
[found
], "R%u", i
);
3654 if (found
== 4) break;
3659 sprintf(srgbtmp
[0], "TA");
3660 sprintf(srgbtmp
[1], "TB");
3661 sprintf(srgbtmp
[2], "TC");
3662 sprintf(srgbtmp
[3], "TD");
3666 sprintf(srgbtmp
[1], "TA");
3667 sprintf(srgbtmp
[2], "TB");
3668 sprintf(srgbtmp
[3], "TC");
3671 sprintf(srgbtmp
[2], "TA");
3672 sprintf(srgbtmp
[3], "TB");
3675 sprintf(srgbtmp
[3], "TA");
3681 /* Create the hw ARB shader */
3682 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
3683 priv_ctx
.cur_ps_args
= args
;
3684 priv_ctx
.compiled_fprog
= compiled
;
3685 priv_ctx
.cur_np2fixup_info
= &compiled
->np2fixup_info
;
3686 init_ps_input(shader
, args
, &priv_ctx
);
3687 list_init(&priv_ctx
.control_frames
);
3688 priv_ctx
.ps_post_process
= args
->super
.srgb_correction
;
3690 /* Avoid enabling NV_fragment_program* if we do not need it.
3692 * Enabling GL_NV_fragment_program_option causes the driver to occupy a temporary register,
3693 * and it slows down the shader execution noticeably(about 5%). Usually our instruction emulation
3694 * is faster than what we gain from using higher native instructions. There are some things though
3695 * that cannot be emulated. In that case enable the extensions.
3696 * If the extension is enabled, instruction handlers that support both ways will use it.
3698 * Testing shows no performance difference between OPTION NV_fragment_program2 and NV_fragment_program.
3699 * So enable the best we can get.
3701 if(reg_maps
->usesdsx
|| reg_maps
->usesdsy
|| reg_maps
->loop_depth
> 0 || reg_maps
->usestexldd
||
3702 reg_maps
->usestexldl
|| reg_maps
->usesfacing
|| reg_maps
->usesifc
|| reg_maps
->usescall
)
3704 want_nv_prog
= TRUE
;
3707 shader_addline(buffer
, "!!ARBfp1.0\n");
3708 if (want_nv_prog
&& gl_info
->supported
[NV_FRAGMENT_PROGRAM2
])
3710 shader_addline(buffer
, "OPTION NV_fragment_program2;\n");
3711 priv_ctx
.target_version
= NV3
;
3713 else if (want_nv_prog
&& gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
])
3715 shader_addline(buffer
, "OPTION NV_fragment_program;\n");
3716 priv_ctx
.target_version
= NV2
;
3720 /* This is an error - either we're advertising the wrong shader version, or aren't enforcing some
3723 ERR("The shader requires instructions that are not available in plain GL_ARB_fragment_program\n");
3726 priv_ctx
.target_version
= ARB
;
3729 if (reg_maps
->rt_mask
> 1)
3731 shader_addline(buffer
, "OPTION ARB_draw_buffers;\n");
3734 if (reg_maps
->shader_version
.major
< 3)
3736 switch (args
->super
.fog
)
3738 case WINED3D_FFP_PS_FOG_OFF
:
3740 case WINED3D_FFP_PS_FOG_LINEAR
:
3741 if (gl_info
->quirks
& WINED3D_QUIRK_BROKEN_ARB_FOG
)
3743 custom_linear_fog
= TRUE
;
3744 priv_ctx
.ps_post_process
= TRUE
;
3747 shader_addline(buffer
, "OPTION ARB_fog_linear;\n");
3749 case WINED3D_FFP_PS_FOG_EXP
:
3750 shader_addline(buffer
, "OPTION ARB_fog_exp;\n");
3752 case WINED3D_FFP_PS_FOG_EXP2
:
3753 shader_addline(buffer
, "OPTION ARB_fog_exp2;\n");
3758 /* For now always declare the temps. At least the Nvidia assembler optimizes completely
3759 * unused temps away(but occupies them for the whole shader if they're used once). Always
3760 * declaring them avoids tricky bookkeeping work
3762 shader_addline(buffer
, "TEMP TA;\n"); /* Used for modifiers */
3763 shader_addline(buffer
, "TEMP TB;\n"); /* Used for modifiers */
3764 shader_addline(buffer
, "TEMP TC;\n"); /* Used for modifiers */
3765 if(dcl_td
) shader_addline(buffer
, "TEMP TD;\n"); /* Used for sRGB writing */
3766 shader_addline(buffer
, "PARAM coefdiv = { 0.5, 0.25, 0.125, 0.0625 };\n");
3767 shader_addline(buffer
, "PARAM coefmul = { 2, 4, 8, 16 };\n");
3768 wined3d_ftoa(eps
, ftoa_tmp
);
3769 shader_addline(buffer
, "PARAM ps_helper_const = { 0.0, 1.0, %s, 0.0 };\n", ftoa_tmp
);
3771 if (reg_maps
->shader_version
.major
< 2)
3773 strcpy(fragcolor
, "R0");
3777 if (priv_ctx
.ps_post_process
)
3779 if (shader
->u
.ps
.color0_mov
)
3781 sprintf(fragcolor
, "R%u", shader
->u
.ps
.color0_reg
);
3785 shader_addline(buffer
, "TEMP TMP_COLOR;\n");
3786 strcpy(fragcolor
, "TMP_COLOR");
3789 strcpy(fragcolor
, "result.color");
3793 if (args
->super
.srgb_correction
)
3795 shader_addline(buffer
, "PARAM srgb_consts0 = ");
3796 shader_arb_append_imm_vec4(buffer
, wined3d_srgb_const0
);
3797 shader_addline(buffer
, ";\n");
3798 shader_addline(buffer
, "PARAM srgb_consts1 = ");
3799 shader_arb_append_imm_vec4(buffer
, wined3d_srgb_const1
);
3800 shader_addline(buffer
, ";\n");
3803 /* Base Declarations */
3804 shader_generate_arb_declarations(shader
, reg_maps
, buffer
, gl_info
, NULL
, &priv_ctx
);
3806 for (i
= 0, map
= reg_maps
->bumpmat
; map
; map
>>= 1, ++i
)
3808 unsigned char bump_const
;
3810 if (!(map
& 1)) continue;
3812 bump_const
= compiled
->numbumpenvmatconsts
;
3813 compiled
->bumpenvmatconst
[bump_const
].const_num
= WINED3D_CONST_NUM_UNUSED
;
3814 compiled
->bumpenvmatconst
[bump_const
].texunit
= i
;
3815 compiled
->luminanceconst
[bump_const
].const_num
= WINED3D_CONST_NUM_UNUSED
;
3816 compiled
->luminanceconst
[bump_const
].texunit
= i
;
3818 /* We can fit the constants into the constant limit for sure because texbem, texbeml, bem and beml are only supported
3819 * in 1.x shaders, and GL_ARB_fragment_program has a constant limit of 24 constants. So in the worst case we're loading
3820 * 8 shader constants, 8 bump matrices and 8 luminance parameters and are perfectly fine. (No NP2 fixup on bumpmapped
3821 * textures due to conditional NP2 restrictions)
3823 * Use local constants to load the bump env parameters, not program.env. This avoids collisions with d3d constants of
3824 * shaders in newer shader models. Since the bump env parameters have to share their space with NP2 fixup constants,
3825 * their location is shader dependent anyway and they cannot be loaded globally.
3827 compiled
->bumpenvmatconst
[bump_const
].const_num
= next_local
++;
3828 shader_addline(buffer
, "PARAM bumpenvmat%d = program.local[%d];\n",
3829 i
, compiled
->bumpenvmatconst
[bump_const
].const_num
);
3830 compiled
->numbumpenvmatconsts
= bump_const
+ 1;
3832 if (!(reg_maps
->luminanceparams
& (1u << i
)))
3835 compiled
->luminanceconst
[bump_const
].const_num
= next_local
++;
3836 shader_addline(buffer
, "PARAM luminance%d = program.local[%d];\n",
3837 i
, compiled
->luminanceconst
[bump_const
].const_num
);
3840 for (i
= 0; i
< WINED3D_MAX_CONSTS_I
; ++i
)
3842 compiled
->int_consts
[i
] = WINED3D_CONST_NUM_UNUSED
;
3843 if (reg_maps
->integer_constants
& (1u << i
) && priv_ctx
.target_version
>= NV2
)
3845 const DWORD
*control_values
= find_loop_control_values(shader
, i
);
3849 shader_addline(buffer
, "PARAM I%u = {%u, %u, %u, -1};\n", i
,
3850 control_values
[0], control_values
[1], control_values
[2]);
3854 compiled
->int_consts
[i
] = next_local
;
3855 compiled
->num_int_consts
++;
3856 shader_addline(buffer
, "PARAM I%u = program.local[%u];\n", i
, next_local
++);
3861 if(reg_maps
->vpos
|| reg_maps
->usesdsy
)
3863 compiled
->ycorrection
= next_local
;
3864 shader_addline(buffer
, "PARAM ycorrection = program.local[%u];\n", next_local
++);
3868 shader_addline(buffer
, "TEMP vpos;\n");
3869 /* ycorrection.x: Backbuffer height(onscreen) or 0(offscreen).
3870 * ycorrection.y: -1.0(onscreen), 1.0(offscreen)
3871 * ycorrection.z: 1.0
3872 * ycorrection.w: 0.0
3874 shader_addline(buffer
, "MAD vpos, fragment.position, ycorrection.zyww, ycorrection.wxww;\n");
3875 shader_addline(buffer
, "FLR vpos.xy, vpos;\n");
3880 compiled
->ycorrection
= WINED3D_CONST_NUM_UNUSED
;
3883 /* Load constants to fixup NP2 texcoords if there are still free constants left:
3884 * Constants (texture dimensions) for the NP2 fixup are loaded as local program parameters. This will consume
3885 * at most 8 (MAX_FRAGMENT_SAMPLERS / 2) parameters, which is highly unlikely, since the application had to
3886 * use 16 NP2 textures at the same time. In case that we run out of constants the fixup is simply not
3887 * applied / activated. This will probably result in wrong rendering of the texture, but will save us from
3888 * shader compilation errors and the subsequent errors when drawing with this shader. */
3889 if (priv_ctx
.cur_ps_args
->super
.np2_fixup
) {
3890 unsigned char cur_fixup_sampler
= 0;
3892 struct arb_ps_np2fixup_info
* const fixup
= priv_ctx
.cur_np2fixup_info
;
3893 const WORD map
= priv_ctx
.cur_ps_args
->super
.np2_fixup
;
3894 const UINT max_lconsts
= gl_info
->limits
.arb_ps_local_constants
;
3896 fixup
->offset
= next_local
;
3897 fixup
->super
.active
= 0;
3899 for (i
= 0; i
< MAX_FRAGMENT_SAMPLERS
; ++i
)
3901 if (!(map
& (1u << i
)))
3904 if (fixup
->offset
+ (cur_fixup_sampler
>> 1) < max_lconsts
)
3906 fixup
->super
.active
|= (1u << i
);
3907 fixup
->super
.idx
[i
] = cur_fixup_sampler
++;
3911 FIXME("No free constant found to load NP2 fixup data into shader. "
3912 "Sampling from this texture will probably look wrong.\n");
3917 fixup
->super
.num_consts
= (cur_fixup_sampler
+ 1) >> 1;
3918 if (fixup
->super
.num_consts
) {
3919 shader_addline(buffer
, "PARAM np2fixup[%u] = { program.env[%u..%u] };\n",
3920 fixup
->super
.num_consts
, fixup
->offset
, fixup
->super
.num_consts
+ fixup
->offset
- 1);
3924 if (shader_priv
->clipplane_emulation
!= ~0U && args
->clip
)
3926 shader_addline(buffer
, "KIL fragment.texcoord[%u];\n", shader_priv
->clipplane_emulation
);
3929 /* Base Shader Body */
3930 shader_generate_main(shader
, buffer
, reg_maps
, function
, &priv_ctx
);
3932 if(args
->super
.srgb_correction
) {
3933 arbfp_add_sRGB_correction(buffer
, fragcolor
, srgbtmp
[0], srgbtmp
[1], srgbtmp
[2], srgbtmp
[3],
3934 priv_ctx
.target_version
>= NV2
);
3937 if (custom_linear_fog
)
3938 arbfp_add_linear_fog(buffer
, fragcolor
, "TA");
3940 if(strcmp(fragcolor
, "result.color")) {
3941 shader_addline(buffer
, "MOV result.color, %s;\n", fragcolor
);
3943 shader_addline(buffer
, "END\n");
3945 /* TODO: change to resource.glObjectHandle or something like that */
3946 GL_EXTCALL(glGenProgramsARB(1, &retval
));
3948 TRACE("Creating a hw pixel shader, prg=%d\n", retval
);
3949 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, retval
));
3951 TRACE("Created hw pixel shader, prg=%d\n", retval
);
3952 if (!shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
->buffer
))
3958 static int compare_sig(const struct wined3d_shader_signature
*sig1
, const struct wined3d_shader_signature
*sig2
)
3963 if (sig1
->element_count
!= sig2
->element_count
)
3964 return sig1
->element_count
< sig2
->element_count
? -1 : 1;
3966 for (i
= 0; i
< sig1
->element_count
; ++i
)
3968 const struct wined3d_shader_signature_element
*e1
, *e2
;
3970 e1
= &sig1
->elements
[i
];
3971 e2
= &sig2
->elements
[i
];
3973 if (!e1
->semantic_name
|| !e2
->semantic_name
)
3975 /* Compare pointers, not contents. One string is NULL (element
3976 * does not exist), the other one is not NULL. */
3977 if (e1
->semantic_name
!= e2
->semantic_name
)
3978 return e1
->semantic_name
< e2
->semantic_name
? -1 : 1;
3982 if ((ret
= strcmp(e1
->semantic_name
, e2
->semantic_name
)))
3984 if (e1
->semantic_idx
!= e2
->semantic_idx
)
3985 return e1
->semantic_idx
< e2
->semantic_idx
? -1 : 1;
3986 if (e1
->sysval_semantic
!= e2
->sysval_semantic
)
3987 return e1
->sysval_semantic
< e2
->sysval_semantic
? -1 : 1;
3988 if (e1
->component_type
!= e2
->component_type
)
3989 return e1
->component_type
< e2
->component_type
? -1 : 1;
3990 if (e1
->register_idx
!= e2
->register_idx
)
3991 return e1
->register_idx
< e2
->register_idx
? -1 : 1;
3992 if (e1
->mask
!= e2
->mask
)
3993 return e1
->mask
< e2
->mask
? -1 : 1;
3998 static void clone_sig(struct wined3d_shader_signature
*new, const struct wined3d_shader_signature
*sig
)
4003 new->element_count
= sig
->element_count
;
4004 new->elements
= wined3d_calloc(new->element_count
, sizeof(*new->elements
));
4005 for (i
= 0; i
< sig
->element_count
; ++i
)
4007 new->elements
[i
] = sig
->elements
[i
];
4009 if (!new->elements
[i
].semantic_name
)
4012 /* Clone the semantic string */
4013 name
= HeapAlloc(GetProcessHeap(), 0, strlen(sig
->elements
[i
].semantic_name
) + 1);
4014 strcpy(name
, sig
->elements
[i
].semantic_name
);
4015 new->elements
[i
].semantic_name
= name
;
4019 static DWORD
find_input_signature(struct shader_arb_priv
*priv
, const struct wined3d_shader_signature
*sig
)
4021 struct wine_rb_entry
*entry
= wine_rb_get(&priv
->signature_tree
, sig
);
4022 struct ps_signature
*found_sig
;
4026 found_sig
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
4027 TRACE("Found existing signature %u\n", found_sig
->idx
);
4028 return found_sig
->idx
;
4030 found_sig
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*found_sig
));
4031 clone_sig(&found_sig
->sig
, sig
);
4032 found_sig
->idx
= priv
->ps_sig_number
++;
4033 TRACE("New signature stored and assigned number %u\n", found_sig
->idx
);
4034 if(wine_rb_put(&priv
->signature_tree
, sig
, &found_sig
->entry
) == -1)
4036 ERR("Failed to insert program entry.\n");
4038 return found_sig
->idx
;
4041 static void init_output_registers(const struct wined3d_shader
*shader
,
4042 const struct wined3d_shader_signature
*ps_input_sig
,
4043 struct shader_arb_ctx_priv
*priv_ctx
, struct arb_vs_compiled_shader
*compiled
)
4046 static const char * const texcoords
[8] =
4048 "result.texcoord[0]", "result.texcoord[1]", "result.texcoord[2]", "result.texcoord[3]",
4049 "result.texcoord[4]", "result.texcoord[5]", "result.texcoord[6]", "result.texcoord[7]"
4051 /* Write generic input varyings 0 to 7 to result.texcoord[], varying 8 to result.color.primary
4052 * and varying 9 to result.color.secondary
4054 static const char * const decl_idx_to_string
[MAX_REG_INPUT
] =
4056 "result.texcoord[0]", "result.texcoord[1]", "result.texcoord[2]", "result.texcoord[3]",
4057 "result.texcoord[4]", "result.texcoord[5]", "result.texcoord[6]", "result.texcoord[7]",
4058 "result.color.primary", "result.color.secondary"
4063 TRACE("Pixel shader uses builtin varyings\n");
4064 /* Map builtins to builtins */
4065 for(i
= 0; i
< 8; i
++)
4067 priv_ctx
->texcrd_output
[i
] = texcoords
[i
];
4069 priv_ctx
->color_output
[0] = "result.color.primary";
4070 priv_ctx
->color_output
[1] = "result.color.secondary";
4071 priv_ctx
->fog_output
= "TMP_FOGCOORD";
4073 /* Map declared regs to builtins. Use "TA" to /dev/null unread output */
4074 for (i
= 0; i
< shader
->output_signature
.element_count
; ++i
)
4076 const struct wined3d_shader_signature_element
*output
= &shader
->output_signature
.elements
[i
];
4078 if (!output
->semantic_name
)
4081 if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_POSITION
))
4083 TRACE("o%u is TMP_OUT\n", output
->register_idx
);
4084 if (!output
->semantic_idx
)
4085 priv_ctx
->vs_output
[output
->register_idx
] = "TMP_OUT";
4087 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4089 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_PSIZE
))
4091 TRACE("o%u is result.pointsize\n", output
->register_idx
);
4092 if (!output
->semantic_idx
)
4093 priv_ctx
->vs_output
[output
->register_idx
] = "result.pointsize";
4095 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4097 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_COLOR
))
4099 TRACE("o%u is result.color.?, idx %u\n", output
->register_idx
, output
->semantic_idx
);
4100 if (!output
->semantic_idx
)
4101 priv_ctx
->vs_output
[output
->register_idx
] = "result.color.primary";
4102 else if (output
->semantic_idx
== 1)
4103 priv_ctx
->vs_output
[output
->register_idx
] = "result.color.secondary";
4104 else priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4106 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
4108 TRACE("o%u is result.texcoord[%u]\n", output
->register_idx
, output
->semantic_idx
);
4109 if (output
->semantic_idx
>= 8)
4110 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4112 priv_ctx
->vs_output
[output
->register_idx
] = texcoords
[output
->semantic_idx
];
4114 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_FOG
))
4116 TRACE("o%u is result.fogcoord\n", output
->register_idx
);
4117 if (output
->semantic_idx
> 0)
4118 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4120 priv_ctx
->vs_output
[output
->register_idx
] = "result.fogcoord";
4124 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4130 TRACE("Pixel shader uses declared varyings\n");
4132 /* Map builtin to declared. /dev/null the results by default to the TA temp reg */
4133 for(i
= 0; i
< 8; i
++)
4135 priv_ctx
->texcrd_output
[i
] = "TA";
4137 priv_ctx
->color_output
[0] = "TA";
4138 priv_ctx
->color_output
[1] = "TA";
4139 priv_ctx
->fog_output
= "TA";
4141 for (i
= 0; i
< ps_input_sig
->element_count
; ++i
)
4143 const struct wined3d_shader_signature_element
*input
= &ps_input_sig
->elements
[i
];
4145 if (!input
->semantic_name
)
4148 /* If a declared input register is not written by builtin arguments, don't write to it.
4149 * GL_NV_vertex_program makes sure the input defaults to 0.0, which is correct with D3D
4151 * Don't care about POSITION and PSIZE here - this is a builtin vertex shader, position goes
4152 * to TMP_OUT in any case
4154 if (shader_match_semantic(input
->semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
4156 if (input
->semantic_idx
< 8)
4157 priv_ctx
->texcrd_output
[input
->semantic_idx
] = decl_idx_to_string
[input
->register_idx
];
4159 else if (shader_match_semantic(input
->semantic_name
, WINED3D_DECL_USAGE_COLOR
))
4161 if (input
->semantic_idx
< 2)
4162 priv_ctx
->color_output
[input
->semantic_idx
] = decl_idx_to_string
[input
->register_idx
];
4164 else if (shader_match_semantic(input
->semantic_name
, WINED3D_DECL_USAGE_FOG
))
4166 if (!input
->semantic_idx
)
4167 priv_ctx
->fog_output
= decl_idx_to_string
[input
->register_idx
];
4174 if (!strcmp(decl_idx_to_string
[input
->register_idx
], "result.color.primary")
4175 || !strcmp(decl_idx_to_string
[input
->register_idx
], "result.color.secondary"))
4177 compiled
->need_color_unclamp
= TRUE
;
4181 /* Map declared to declared */
4182 for (i
= 0; i
< shader
->output_signature
.element_count
; ++i
)
4184 const struct wined3d_shader_signature_element
*output
= &shader
->output_signature
.elements
[i
];
4186 /* Write unread output to TA to throw them away */
4187 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4189 if (!output
->semantic_name
)
4192 if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_POSITION
) && !output
->semantic_idx
)
4194 priv_ctx
->vs_output
[output
->register_idx
] = "TMP_OUT";
4197 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_PSIZE
) && !output
->semantic_idx
)
4199 priv_ctx
->vs_output
[output
->register_idx
] = "result.pointsize";
4203 for (j
= 0; j
< ps_input_sig
->element_count
; ++j
)
4205 const struct wined3d_shader_signature_element
*input
= &ps_input_sig
->elements
[j
];
4207 if (!input
->semantic_name
)
4210 if (!strcmp(input
->semantic_name
, output
->semantic_name
)
4211 && input
->semantic_idx
== output
->semantic_idx
)
4213 priv_ctx
->vs_output
[output
->register_idx
] = decl_idx_to_string
[input
->register_idx
];
4215 if (!strcmp(priv_ctx
->vs_output
[output
->register_idx
], "result.color.primary")
4216 || !strcmp(priv_ctx
->vs_output
[output
->register_idx
], "result.color.secondary"))
4218 compiled
->need_color_unclamp
= TRUE
;
4225 /* Context activation is done by the caller. */
4226 static GLuint
shader_arb_generate_vshader(const struct wined3d_shader
*shader
,
4227 const struct wined3d_gl_info
*gl_info
, struct wined3d_string_buffer
*buffer
,
4228 const struct arb_vs_compile_args
*args
, struct arb_vs_compiled_shader
*compiled
,
4229 const struct wined3d_shader_signature
*ps_input_sig
)
4231 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
4232 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
4233 struct shader_arb_priv
*priv
= shader
->device
->shader_priv
;
4234 const DWORD
*function
= shader
->function
;
4236 DWORD next_local
= 0;
4237 struct shader_arb_ctx_priv priv_ctx
;
4240 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
4241 priv_ctx
.cur_vs_args
= args
;
4242 list_init(&priv_ctx
.control_frames
);
4243 init_output_registers(shader
, ps_input_sig
, &priv_ctx
, compiled
);
4245 /* Create the hw ARB shader */
4246 shader_addline(buffer
, "!!ARBvp1.0\n");
4248 /* Always enable the NV extension if available. Unlike fragment shaders, there is no
4249 * mesurable performance penalty, and we can always make use of it for clipplanes.
4251 if (gl_info
->supported
[NV_VERTEX_PROGRAM3
])
4253 shader_addline(buffer
, "OPTION NV_vertex_program3;\n");
4254 priv_ctx
.target_version
= NV3
;
4255 shader_addline(buffer
, "ADDRESS aL;\n");
4257 else if (gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
])
4259 shader_addline(buffer
, "OPTION NV_vertex_program2;\n");
4260 priv_ctx
.target_version
= NV2
;
4261 shader_addline(buffer
, "ADDRESS aL;\n");
4263 priv_ctx
.target_version
= ARB
;
4266 shader_addline(buffer
, "TEMP TMP_OUT;\n");
4268 shader_addline(buffer
, "TEMP TMP_FOGCOORD;\n");
4269 if (need_helper_const(shader_data
, reg_maps
, gl_info
))
4272 wined3d_ftoa(eps
, ftoa_tmp
);
4273 shader_addline(buffer
, "PARAM helper_const = { 0.0, 1.0, 2.0, %s};\n", ftoa_tmp
);
4275 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
))
4277 shader_addline(buffer
, "PARAM rel_addr_const = { 0.5, %d.0, 0.0, 0.0 };\n", shader_data
->rel_offset
);
4278 shader_addline(buffer
, "TEMP A0_SHADOW;\n");
4281 shader_addline(buffer
, "TEMP TA;\n");
4282 shader_addline(buffer
, "TEMP TB;\n");
4284 /* Base Declarations */
4285 shader_generate_arb_declarations(shader
, reg_maps
, buffer
, gl_info
,
4286 &priv_ctx
.vs_clipplanes
, &priv_ctx
);
4288 for (i
= 0; i
< WINED3D_MAX_CONSTS_I
; ++i
)
4290 compiled
->int_consts
[i
] = WINED3D_CONST_NUM_UNUSED
;
4291 if (reg_maps
->integer_constants
& (1u << i
) && priv_ctx
.target_version
>= NV2
)
4293 const DWORD
*control_values
= find_loop_control_values(shader
, i
);
4297 shader_addline(buffer
, "PARAM I%u = {%u, %u, %u, -1};\n", i
,
4298 control_values
[0], control_values
[1], control_values
[2]);
4302 compiled
->int_consts
[i
] = next_local
;
4303 compiled
->num_int_consts
++;
4304 shader_addline(buffer
, "PARAM I%u = program.local[%u];\n", i
, next_local
++);
4309 /* We need a constant to fixup the final position */
4310 shader_addline(buffer
, "PARAM posFixup = program.local[%u];\n", next_local
);
4311 compiled
->pos_fixup
= next_local
++;
4313 /* Initialize output parameters. GL_ARB_vertex_program does not require special initialization values
4314 * for output parameters. D3D in theory does not do that either, but some applications depend on a
4315 * proper initialization of the secondary color, and programs using the fixed function pipeline without
4316 * a replacement shader depend on the texcoord.w being set properly.
4318 * GL_NV_vertex_program defines that all output values are initialized to {0.0, 0.0, 0.0, 1.0}. This
4319 * assertion is in effect even when using GL_ARB_vertex_program without any NV specific additions. So
4320 * skip this if NV_vertex_program is supported. Otherwise, initialize the secondary color. For the tex-
4321 * coords, we have a flag in the opengl caps. Many cards do not require the texcoord being set, and
4322 * this can eat a number of instructions, so skip it unless this cap is set as well
4324 if (!gl_info
->supported
[NV_VERTEX_PROGRAM
])
4326 const char *color_init
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_0001
);
4327 shader_addline(buffer
, "MOV result.color.secondary, %s;\n", color_init
);
4329 if (gl_info
->quirks
& WINED3D_QUIRK_SET_TEXCOORD_W
&& !priv
->ffp_proj_control
)
4332 const char *one
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_ONE
);
4333 for(i
= 0; i
< MAX_REG_TEXCRD
; i
++)
4335 if (reg_maps
->u
.texcoord_mask
[i
] && reg_maps
->u
.texcoord_mask
[i
] != WINED3DSP_WRITEMASK_ALL
)
4336 shader_addline(buffer
, "MOV result.texcoord[%u].w, %s\n", i
, one
);
4341 /* The shader starts with the main function */
4342 priv_ctx
.in_main_func
= TRUE
;
4343 /* Base Shader Body */
4344 shader_generate_main(shader
, buffer
, reg_maps
, function
, &priv_ctx
);
4346 if (!priv_ctx
.footer_written
) vshader_add_footer(&priv_ctx
,
4347 shader_data
, args
, reg_maps
, gl_info
, buffer
);
4349 shader_addline(buffer
, "END\n");
4351 /* TODO: change to resource.glObjectHandle or something like that */
4352 GL_EXTCALL(glGenProgramsARB(1, &ret
));
4354 TRACE("Creating a hw vertex shader, prg=%d\n", ret
);
4355 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, ret
));
4357 TRACE("Created hw vertex shader, prg=%d\n", ret
);
4358 if (!shader_arb_compile(gl_info
, GL_VERTEX_PROGRAM_ARB
, buffer
->buffer
))
4364 /* Context activation is done by the caller. */
4365 static struct arb_ps_compiled_shader
*find_arb_pshader(struct wined3d_shader
*shader
,
4366 const struct arb_ps_compile_args
*args
)
4368 struct wined3d_device
*device
= shader
->device
;
4369 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
4370 const struct wined3d_d3d_info
*d3d_info
= &device
->adapter
->d3d_info
;
4373 struct arb_ps_compiled_shader
*new_array
;
4374 struct wined3d_string_buffer buffer
;
4375 struct arb_pshader_private
*shader_data
;
4378 if (!shader
->backend_data
)
4380 struct shader_arb_priv
*priv
= device
->shader_priv
;
4382 shader
->backend_data
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
4383 shader_data
= shader
->backend_data
;
4384 shader_data
->clamp_consts
= shader
->reg_maps
.shader_version
.major
== 1;
4386 if (shader
->reg_maps
.shader_version
.major
< 3)
4387 shader_data
->input_signature_idx
= ~0U;
4389 shader_data
->input_signature_idx
= find_input_signature(priv
, &shader
->input_signature
);
4391 TRACE("Shader got assigned input signature index %u\n", shader_data
->input_signature_idx
);
4393 if (!d3d_info
->vs_clipping
)
4394 shader_data
->clipplane_emulation
= shader_find_free_input_register(&shader
->reg_maps
,
4395 d3d_info
->limits
.ffp_blend_stages
- 1);
4397 shader_data
->clipplane_emulation
= ~0U;
4399 shader_data
= shader
->backend_data
;
4401 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4402 * so a linear search is more performant than a hashmap or a binary search
4403 * (cache coherency etc)
4405 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4407 if (!memcmp(&shader_data
->gl_shaders
[i
].args
, args
, sizeof(*args
)))
4408 return &shader_data
->gl_shaders
[i
];
4411 TRACE("No matching GL shader found, compiling a new shader\n");
4412 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4413 if (shader_data
->num_gl_shaders
)
4415 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4416 new_array
= HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, shader_data
->gl_shaders
,
4417 new_size
* sizeof(*shader_data
->gl_shaders
));
4419 new_array
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
->gl_shaders
));
4424 ERR("Out of memory\n");
4427 shader_data
->gl_shaders
= new_array
;
4428 shader_data
->shader_array_size
= new_size
;
4431 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4433 pixelshader_update_resource_types(shader
, args
->super
.tex_types
);
4435 if (!string_buffer_init(&buffer
))
4437 ERR("Failed to initialize shader buffer.\n");
4441 ret
= shader_arb_generate_pshader(shader
, gl_info
, &buffer
, args
,
4442 &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
]);
4443 string_buffer_free(&buffer
);
4444 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].prgId
= ret
;
4446 return &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++];
4449 static inline BOOL
vs_args_equal(const struct arb_vs_compile_args
*stored
, const struct arb_vs_compile_args
*new,
4450 const DWORD use_map
, BOOL skip_int
) {
4451 if((stored
->super
.swizzle_map
& use_map
) != new->super
.swizzle_map
) return FALSE
;
4452 if(stored
->super
.clip_enabled
!= new->super
.clip_enabled
) return FALSE
;
4453 if(stored
->super
.fog_src
!= new->super
.fog_src
) return FALSE
;
4454 if(stored
->clip
.boolclip_compare
!= new->clip
.boolclip_compare
) return FALSE
;
4455 if(stored
->ps_signature
!= new->ps_signature
) return FALSE
;
4456 if(stored
->vertex
.samplers_compare
!= new->vertex
.samplers_compare
) return FALSE
;
4457 if(skip_int
) return TRUE
;
4459 return !memcmp(stored
->loop_ctrl
, new->loop_ctrl
, sizeof(stored
->loop_ctrl
));
4462 static struct arb_vs_compiled_shader
*find_arb_vshader(struct wined3d_shader
*shader
,
4463 const struct wined3d_gl_info
*gl_info
, DWORD use_map
, const struct arb_vs_compile_args
*args
,
4464 const struct wined3d_shader_signature
*ps_input_sig
)
4468 struct arb_vs_compiled_shader
*new_array
;
4469 struct wined3d_string_buffer buffer
;
4470 struct arb_vshader_private
*shader_data
;
4473 if (!shader
->backend_data
)
4475 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
4477 shader
->backend_data
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
4478 shader_data
= shader
->backend_data
;
4480 if ((gl_info
->quirks
& WINED3D_QUIRK_ARB_VS_OFFSET_LIMIT
)
4481 && reg_maps
->min_rel_offset
<= reg_maps
->max_rel_offset
)
4483 if (reg_maps
->max_rel_offset
- reg_maps
->min_rel_offset
> 127)
4485 FIXME("The difference between the minimum and maximum relative offset is > 127.\n");
4486 FIXME("Which this OpenGL implementation does not support. Try using GLSL.\n");
4487 FIXME("Min: %u, Max: %u.\n", reg_maps
->min_rel_offset
, reg_maps
->max_rel_offset
);
4489 else if (reg_maps
->max_rel_offset
- reg_maps
->min_rel_offset
> 63)
4490 shader_data
->rel_offset
= reg_maps
->min_rel_offset
+ 63;
4491 else if (reg_maps
->max_rel_offset
> 63)
4492 shader_data
->rel_offset
= reg_maps
->min_rel_offset
;
4495 shader_data
= shader
->backend_data
;
4497 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4498 * so a linear search is more performant than a hashmap or a binary search
4499 * (cache coherency etc)
4501 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
4502 if (vs_args_equal(&shader_data
->gl_shaders
[i
].args
, args
,
4503 use_map
, gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]))
4505 return &shader_data
->gl_shaders
[i
];
4509 TRACE("No matching GL shader found, compiling a new shader\n");
4511 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4512 if (shader_data
->num_gl_shaders
)
4514 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4515 new_array
= HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, shader_data
->gl_shaders
,
4516 new_size
* sizeof(*shader_data
->gl_shaders
));
4518 new_array
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
->gl_shaders
));
4523 ERR("Out of memory\n");
4526 shader_data
->gl_shaders
= new_array
;
4527 shader_data
->shader_array_size
= new_size
;
4530 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4532 if (!string_buffer_init(&buffer
))
4534 ERR("Failed to initialize shader buffer.\n");
4538 ret
= shader_arb_generate_vshader(shader
, gl_info
, &buffer
, args
,
4539 &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
],
4541 string_buffer_free(&buffer
);
4542 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].prgId
= ret
;
4544 return &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++];
4547 static void find_arb_ps_compile_args(const struct wined3d_state
*state
,
4548 const struct wined3d_context
*context
, const struct wined3d_shader
*shader
,
4549 struct arb_ps_compile_args
*args
)
4551 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4552 const struct wined3d_d3d_info
*d3d_info
= context
->d3d_info
;
4556 find_ps_compile_args(state
, shader
, context
->stream_info
.position_transformed
, &args
->super
, context
);
4558 /* This forces all local boolean constants to 1 to make them stateblock independent */
4559 args
->bools
= shader
->reg_maps
.local_bool_consts
;
4561 for (i
= 0; i
< WINED3D_MAX_CONSTS_B
; ++i
)
4563 if (state
->ps_consts_b
[i
])
4564 args
->bools
|= ( 1u << i
);
4567 /* Only enable the clip plane emulation KIL if at least one clipplane is enabled. The KIL instruction
4568 * is quite expensive because it forces the driver to disable early Z discards. It is cheaper to
4569 * duplicate the shader than have a no-op KIL instruction in every shader
4571 if (!d3d_info
->vs_clipping
&& use_vs(state
)
4572 && state
->render_states
[WINED3D_RS_CLIPPING
]
4573 && state
->render_states
[WINED3D_RS_CLIPPLANEENABLE
])
4578 /* Skip if unused or local, or supported natively */
4579 int_skip
= ~shader
->reg_maps
.integer_constants
| shader
->reg_maps
.local_int_consts
;
4580 if (int_skip
== 0xffff || gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
])
4582 memset(args
->loop_ctrl
, 0, sizeof(args
->loop_ctrl
));
4586 for (i
= 0; i
< WINED3D_MAX_CONSTS_I
; ++i
)
4588 if (int_skip
& (1u << i
))
4590 args
->loop_ctrl
[i
][0] = 0;
4591 args
->loop_ctrl
[i
][1] = 0;
4592 args
->loop_ctrl
[i
][2] = 0;
4596 args
->loop_ctrl
[i
][0] = state
->ps_consts_i
[i
].x
;
4597 args
->loop_ctrl
[i
][1] = state
->ps_consts_i
[i
].y
;
4598 args
->loop_ctrl
[i
][2] = state
->ps_consts_i
[i
].z
;
4603 static void find_arb_vs_compile_args(const struct wined3d_state
*state
,
4604 const struct wined3d_context
*context
, const struct wined3d_shader
*shader
,
4605 struct arb_vs_compile_args
*args
)
4607 const struct wined3d_device
*device
= shader
->device
;
4608 const struct wined3d_adapter
*adapter
= device
->adapter
;
4609 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4610 const struct wined3d_d3d_info
*d3d_info
= context
->d3d_info
;
4614 find_vs_compile_args(state
, shader
, context
->stream_info
.swizzle_map
, &args
->super
, d3d_info
);
4616 args
->clip
.boolclip_compare
= 0;
4619 const struct wined3d_shader
*ps
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
4620 const struct arb_pshader_private
*shader_priv
= ps
->backend_data
;
4621 args
->ps_signature
= shader_priv
->input_signature_idx
;
4623 args
->clip
.boolclip
.clip_texcoord
= shader_priv
->clipplane_emulation
+ 1;
4627 args
->ps_signature
= ~0;
4628 if (!d3d_info
->vs_clipping
&& adapter
->fragment_pipe
== &arbfp_fragment_pipeline
)
4629 args
->clip
.boolclip
.clip_texcoord
= ffp_clip_emul(context
) ? d3d_info
->limits
.ffp_blend_stages
: 0;
4630 /* Otherwise: Setting boolclip_compare set clip_texcoord to 0 */
4633 if (args
->clip
.boolclip
.clip_texcoord
)
4635 if (state
->render_states
[WINED3D_RS_CLIPPING
])
4636 args
->clip
.boolclip
.clipplane_mask
= (unsigned char)state
->render_states
[WINED3D_RS_CLIPPLANEENABLE
];
4637 /* clipplane_mask was set to 0 by setting boolclip_compare to 0 */
4640 /* This forces all local boolean constants to 1 to make them stateblock independent */
4641 args
->clip
.boolclip
.bools
= shader
->reg_maps
.local_bool_consts
;
4642 /* TODO: Figure out if it would be better to store bool constants as bitmasks in the stateblock */
4643 for (i
= 0; i
< WINED3D_MAX_CONSTS_B
; ++i
)
4645 if (state
->vs_consts_b
[i
])
4646 args
->clip
.boolclip
.bools
|= (1u << i
);
4649 args
->vertex
.samplers
[0] = context
->tex_unit_map
[MAX_FRAGMENT_SAMPLERS
+ 0];
4650 args
->vertex
.samplers
[1] = context
->tex_unit_map
[MAX_FRAGMENT_SAMPLERS
+ 1];
4651 args
->vertex
.samplers
[2] = context
->tex_unit_map
[MAX_FRAGMENT_SAMPLERS
+ 2];
4652 args
->vertex
.samplers
[3] = 0;
4654 /* Skip if unused or local */
4655 int_skip
= ~shader
->reg_maps
.integer_constants
| shader
->reg_maps
.local_int_consts
;
4656 /* This is about flow control, not clipping. */
4657 if (int_skip
== 0xffff || gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
])
4659 memset(args
->loop_ctrl
, 0, sizeof(args
->loop_ctrl
));
4663 for (i
= 0; i
< WINED3D_MAX_CONSTS_I
; ++i
)
4665 if (int_skip
& (1u << i
))
4667 args
->loop_ctrl
[i
][0] = 0;
4668 args
->loop_ctrl
[i
][1] = 0;
4669 args
->loop_ctrl
[i
][2] = 0;
4673 args
->loop_ctrl
[i
][0] = state
->vs_consts_i
[i
].x
;
4674 args
->loop_ctrl
[i
][1] = state
->vs_consts_i
[i
].y
;
4675 args
->loop_ctrl
[i
][2] = state
->vs_consts_i
[i
].z
;
4680 /* Context activation is done by the caller. */
4681 static void shader_arb_select(void *shader_priv
, struct wined3d_context
*context
,
4682 const struct wined3d_state
*state
)
4684 struct shader_arb_priv
*priv
= shader_priv
;
4685 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4688 /* Deal with pixel shaders first so the vertex shader arg function has the input signature ready */
4691 struct wined3d_shader
*ps
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
4692 struct arb_ps_compile_args compile_args
;
4693 struct arb_ps_compiled_shader
*compiled
;
4695 TRACE("Using pixel shader %p.\n", ps
);
4696 find_arb_ps_compile_args(state
, context
, ps
, &compile_args
);
4697 compiled
= find_arb_pshader(ps
, &compile_args
);
4698 priv
->current_fprogram_id
= compiled
->prgId
;
4699 priv
->compiled_fprog
= compiled
;
4701 /* Bind the fragment program */
4702 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, priv
->current_fprogram_id
));
4703 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->current_fprogram_id);");
4705 if (!priv
->use_arbfp_fixed_func
)
4706 priv
->fragment_pipe
->enable_extension(gl_info
, FALSE
);
4708 /* Enable OpenGL fragment programs. */
4709 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
4710 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
4712 TRACE("Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", priv
->current_fprogram_id
);
4714 /* Pixel Shader 1.x constants are clamped to [-1;1], Pixel Shader 2.0 constants are not. If switching between
4715 * a 1.x and newer shader, reload the first 8 constants
4717 if (priv
->last_ps_const_clamped
!= ((struct arb_pshader_private
*)ps
->backend_data
)->clamp_consts
)
4719 priv
->last_ps_const_clamped
= ((struct arb_pshader_private
*)ps
->backend_data
)->clamp_consts
;
4720 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, 8);
4721 for(i
= 0; i
< 8; i
++)
4723 priv
->pshader_const_dirty
[i
] = 1;
4725 /* Also takes care of loading local constants */
4726 shader_arb_load_constants_internal(shader_priv
, context
, state
, TRUE
, FALSE
, TRUE
);
4730 UINT rt_height
= state
->fb
->render_targets
[0]->height
;
4731 shader_arb_ps_local_constants(compiled
, context
, state
, rt_height
);
4734 /* Force constant reloading for the NP2 fixup (see comment in shader_glsl_select for more info) */
4735 if (compiled
->np2fixup_info
.super
.active
)
4736 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_NP2_FIXUP
;
4738 if (ps
->load_local_constsF
)
4739 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_F
;
4743 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
] && !priv
->use_arbfp_fixed_func
)
4745 /* Disable only if we're not using arbfp fixed function fragment
4746 * processing. If this is used, keep GL_FRAGMENT_PROGRAM_ARB
4747 * enabled, and the fixed function pipeline will bind the fixed
4748 * function replacement shader. */
4749 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4750 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4751 priv
->current_fprogram_id
= 0;
4753 priv
->fragment_pipe
->enable_extension(gl_info
, TRUE
);
4758 struct wined3d_shader
*vs
= state
->shader
[WINED3D_SHADER_TYPE_VERTEX
];
4759 struct arb_vs_compile_args compile_args
;
4760 struct arb_vs_compiled_shader
*compiled
;
4761 const struct wined3d_shader_signature
*ps_input_sig
;
4763 TRACE("Using vertex shader %p\n", vs
);
4764 find_arb_vs_compile_args(state
, context
, vs
, &compile_args
);
4766 /* Instead of searching for the signature in the signature list, read the one from the
4767 * current pixel shader. It's maybe not the shader where the signature came from, but it
4768 * is the same signature and faster to find. */
4769 if (compile_args
.ps_signature
== ~0U)
4770 ps_input_sig
= NULL
;
4772 ps_input_sig
= &state
->shader
[WINED3D_SHADER_TYPE_PIXEL
]->input_signature
;
4774 compiled
= find_arb_vshader(vs
, context
->gl_info
, context
->stream_info
.use_map
,
4775 &compile_args
, ps_input_sig
);
4776 priv
->current_vprogram_id
= compiled
->prgId
;
4777 priv
->compiled_vprog
= compiled
;
4779 /* Bind the vertex program */
4780 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->current_vprogram_id
));
4781 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->current_vprogram_id);");
4783 priv
->vertex_pipe
->vp_enable(gl_info
, FALSE
);
4785 /* Enable OpenGL vertex programs */
4786 gl_info
->gl_ops
.gl
.p_glEnable(GL_VERTEX_PROGRAM_ARB
);
4787 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
4788 TRACE("Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", priv
->current_vprogram_id
);
4789 shader_arb_vs_local_constants(compiled
, context
, state
);
4791 if(priv
->last_vs_color_unclamp
!= compiled
->need_color_unclamp
) {
4792 priv
->last_vs_color_unclamp
= compiled
->need_color_unclamp
;
4794 if (gl_info
->supported
[ARB_COLOR_BUFFER_FLOAT
])
4796 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, !compiled
->need_color_unclamp
));
4797 checkGLcall("glClampColorARB");
4799 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
4803 if (vs
->load_local_constsF
)
4804 context
->constant_update_mask
|= WINED3D_SHADER_CONST_VS_F
;
4808 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
4810 priv
->current_vprogram_id
= 0;
4811 gl_info
->gl_ops
.gl
.p_glDisable(GL_VERTEX_PROGRAM_ARB
);
4812 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4814 priv
->vertex_pipe
->vp_enable(gl_info
, TRUE
);
4819 /* Context activation is done by the caller. */
4820 static void shader_arb_disable(void *shader_priv
, struct wined3d_context
*context
)
4822 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4823 struct shader_arb_priv
*priv
= shader_priv
;
4825 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
4827 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4828 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4829 priv
->current_fprogram_id
= 0;
4831 priv
->fragment_pipe
->enable_extension(gl_info
, FALSE
);
4833 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
4835 priv
->current_vprogram_id
= 0;
4836 gl_info
->gl_ops
.gl
.p_glDisable(GL_VERTEX_PROGRAM_ARB
);
4837 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4839 priv
->vertex_pipe
->vp_enable(gl_info
, FALSE
);
4841 if (gl_info
->supported
[ARB_COLOR_BUFFER_FLOAT
] && priv
->last_vs_color_unclamp
)
4843 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, GL_FIXED_ONLY_ARB
));
4844 checkGLcall("glClampColorARB");
4845 priv
->last_vs_color_unclamp
= FALSE
;
4848 context
->shader_update_mask
= (1u << WINED3D_SHADER_TYPE_PIXEL
)
4849 | (1u << WINED3D_SHADER_TYPE_VERTEX
)
4850 | (1u << WINED3D_SHADER_TYPE_GEOMETRY
)
4851 | (1u << WINED3D_SHADER_TYPE_HULL
)
4852 | (1u << WINED3D_SHADER_TYPE_DOMAIN
)
4853 | (1u << WINED3D_SHADER_TYPE_COMPUTE
);
4856 /* Context activation is done by the caller. */
4857 static void shader_arb_select_depth_blt(void *shader_priv
, const struct wined3d_gl_info
*gl_info
,
4858 enum wined3d_gl_resource_type tex_type
, const SIZE
*ds_mask_size
)
4860 const float mask
[] = {0.0f
, 0.0f
, (float)ds_mask_size
->cx
, (float)ds_mask_size
->cy
};
4861 BOOL masked
= ds_mask_size
->cx
&& ds_mask_size
->cy
;
4862 struct shader_arb_priv
*priv
= shader_priv
;
4863 GLuint
*blt_fprogram
;
4865 if (!priv
->depth_blt_vprogram_id
) priv
->depth_blt_vprogram_id
= create_arb_blt_vertex_program(gl_info
);
4866 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->depth_blt_vprogram_id
));
4867 gl_info
->gl_ops
.gl
.p_glEnable(GL_VERTEX_PROGRAM_ARB
);
4869 blt_fprogram
= masked
? &priv
->depth_blt_fprogram_id_masked
[tex_type
] : &priv
->depth_blt_fprogram_id_full
[tex_type
];
4870 if (!*blt_fprogram
) *blt_fprogram
= create_arb_blt_fragment_program(gl_info
, tex_type
, masked
);
4871 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, *blt_fprogram
));
4872 if (masked
) GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, 0, mask
));
4873 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
4876 /* Context activation is done by the caller. */
4877 static void shader_arb_deselect_depth_blt(void *shader_priv
, const struct wined3d_gl_info
*gl_info
)
4879 struct shader_arb_priv
*priv
= shader_priv
;
4881 if (priv
->current_vprogram_id
) {
4882 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->current_vprogram_id
));
4883 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
4885 TRACE("Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB.\n", priv
->current_vprogram_id
);
4889 gl_info
->gl_ops
.gl
.p_glDisable(GL_VERTEX_PROGRAM_ARB
);
4890 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4893 if (priv
->current_fprogram_id
) {
4894 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, priv
->current_fprogram_id
));
4895 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
4897 TRACE("Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB.\n", priv
->current_fprogram_id
);
4899 else if(!priv
->use_arbfp_fixed_func
)
4901 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4902 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4906 static void shader_arb_destroy(struct wined3d_shader
*shader
)
4908 struct wined3d_device
*device
= shader
->device
;
4909 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
4911 if (shader_is_pshader_version(shader
->reg_maps
.shader_version
.type
))
4913 struct arb_pshader_private
*shader_data
= shader
->backend_data
;
4916 if(!shader_data
) return; /* This can happen if a shader was never compiled */
4918 if (shader_data
->num_gl_shaders
)
4920 struct wined3d_context
*context
= context_acquire(device
, NULL
);
4922 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4924 GL_EXTCALL(glDeleteProgramsARB(1, &shader_data
->gl_shaders
[i
].prgId
));
4925 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader_data->gl_shaders[i].prgId))");
4928 context_release(context
);
4931 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
);
4932 HeapFree(GetProcessHeap(), 0, shader_data
);
4933 shader
->backend_data
= NULL
;
4937 struct arb_vshader_private
*shader_data
= shader
->backend_data
;
4940 if(!shader_data
) return; /* This can happen if a shader was never compiled */
4942 if (shader_data
->num_gl_shaders
)
4944 struct wined3d_context
*context
= context_acquire(device
, NULL
);
4946 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4948 GL_EXTCALL(glDeleteProgramsARB(1, &shader_data
->gl_shaders
[i
].prgId
));
4949 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader_data->gl_shaders[i].prgId))");
4952 context_release(context
);
4955 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
);
4956 HeapFree(GetProcessHeap(), 0, shader_data
);
4957 shader
->backend_data
= NULL
;
4961 static int sig_tree_compare(const void *key
, const struct wine_rb_entry
*entry
)
4963 struct ps_signature
*e
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
4964 return compare_sig(key
, &e
->sig
);
4967 static HRESULT
shader_arb_alloc(struct wined3d_device
*device
, const struct wined3d_vertex_pipe_ops
*vertex_pipe
,
4968 const struct fragment_pipeline
*fragment_pipe
)
4970 struct shader_arb_priv
*priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*priv
));
4971 struct fragment_caps fragment_caps
;
4972 void *vertex_priv
, *fragment_priv
;
4973 const struct wined3d_d3d_info
*d3d_info
= &device
->adapter
->d3d_info
;
4975 if (!(vertex_priv
= vertex_pipe
->vp_alloc(&arb_program_shader_backend
, priv
)))
4977 ERR("Failed to initialize vertex pipe.\n");
4978 HeapFree(GetProcessHeap(), 0, priv
);
4982 if (!(fragment_priv
= fragment_pipe
->alloc_private(&arb_program_shader_backend
, priv
)))
4984 ERR("Failed to initialize fragment pipe.\n");
4985 vertex_pipe
->vp_free(device
);
4986 HeapFree(GetProcessHeap(), 0, priv
);
4990 memset(priv
->vshader_const_dirty
, 1,
4991 sizeof(*priv
->vshader_const_dirty
) * d3d_info
->limits
.vs_uniform_count
);
4992 memset(priv
->pshader_const_dirty
, 1,
4993 sizeof(*priv
->pshader_const_dirty
) * d3d_info
->limits
.ps_uniform_count
);
4995 wine_rb_init(&priv
->signature_tree
, sig_tree_compare
);
4997 priv
->vertex_pipe
= vertex_pipe
;
4998 priv
->fragment_pipe
= fragment_pipe
;
4999 fragment_pipe
->get_caps(&device
->adapter
->gl_info
, &fragment_caps
);
5000 priv
->ffp_proj_control
= fragment_caps
.wined3d_caps
& WINED3D_FRAGMENT_CAP_PROJ_CONTROL
;
5002 device
->vertex_priv
= vertex_priv
;
5003 device
->fragment_priv
= fragment_priv
;
5004 device
->shader_priv
= priv
;
5009 static void release_signature(struct wine_rb_entry
*entry
, void *context
)
5011 struct ps_signature
*sig
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
5014 for (i
= 0; i
< sig
->sig
.element_count
; ++i
)
5016 HeapFree(GetProcessHeap(), 0, (char *)sig
->sig
.elements
[i
].semantic_name
);
5018 HeapFree(GetProcessHeap(), 0, sig
->sig
.elements
);
5019 HeapFree(GetProcessHeap(), 0, sig
);
5022 /* Context activation is done by the caller. */
5023 static void shader_arb_free(struct wined3d_device
*device
)
5025 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
5026 struct shader_arb_priv
*priv
= device
->shader_priv
;
5029 if (priv
->depth_blt_vprogram_id
)
5030 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_vprogram_id
));
5032 for (i
= 0; i
< WINED3D_GL_RES_TYPE_COUNT
; ++i
)
5034 if (priv
->depth_blt_fprogram_id_full
[i
])
5036 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_fprogram_id_full
[i
]));
5038 if (priv
->depth_blt_fprogram_id_masked
[i
])
5040 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_fprogram_id_masked
[i
]));
5044 wine_rb_destroy(&priv
->signature_tree
, release_signature
, NULL
);
5045 priv
->fragment_pipe
->free_private(device
);
5046 priv
->vertex_pipe
->vp_free(device
);
5047 HeapFree(GetProcessHeap(), 0, device
->shader_priv
);
5050 static BOOL
shader_arb_allocate_context_data(struct wined3d_context
*context
)
5055 static void shader_arb_free_context_data(struct wined3d_context
*context
)
5057 struct shader_arb_priv
*priv
;
5059 priv
= context
->device
->shader_priv
;
5060 if (priv
->last_context
== context
)
5061 priv
->last_context
= NULL
;
5064 static void shader_arb_init_context_state(struct wined3d_context
*context
) {}
5066 static void shader_arb_get_caps(const struct wined3d_gl_info
*gl_info
, struct shader_caps
*caps
)
5068 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
5073 /* 96 is the minimum allowed value of MAX_PROGRAM_ENV_PARAMETERS_ARB
5074 * for vertex programs. If the native limit is less than that it's
5075 * not very useful, and e.g. Mesa swrast returns 0, probably to
5076 * indicate it's a software implementation. */
5077 if (gl_info
->limits
.arb_vs_native_constants
< 96)
5078 vs_consts
= gl_info
->limits
.arb_vs_float_constants
;
5080 vs_consts
= min(gl_info
->limits
.arb_vs_float_constants
, gl_info
->limits
.arb_vs_native_constants
);
5082 if (gl_info
->supported
[NV_VERTEX_PROGRAM3
])
5085 TRACE("Hardware vertex shader version 3.0 enabled (NV_VERTEX_PROGRAM3)\n");
5087 else if (vs_consts
>= 256)
5089 /* Shader Model 2.0 requires at least 256 vertex shader constants */
5091 TRACE("Hardware vertex shader version 2.0 enabled (ARB_PROGRAM)\n");
5096 TRACE("Hardware vertex shader version 1.1 enabled (ARB_PROGRAM)\n");
5098 caps
->vs_version
= min(wined3d_settings
.max_sm_vs
, vs_version
);
5099 caps
->vs_uniform_count
= min(WINED3D_MAX_VS_CONSTS_F
, vs_consts
);
5103 caps
->vs_version
= 0;
5104 caps
->vs_uniform_count
= 0;
5107 caps
->hs_version
= 0;
5108 caps
->ds_version
= 0;
5109 caps
->gs_version
= 0;
5110 caps
->cs_version
= 0;
5112 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
5117 /* Similar as above for vertex programs, but the minimum for fragment
5118 * programs is 24. */
5119 if (gl_info
->limits
.arb_ps_native_constants
< 24)
5120 ps_consts
= gl_info
->limits
.arb_ps_float_constants
;
5122 ps_consts
= min(gl_info
->limits
.arb_ps_float_constants
, gl_info
->limits
.arb_ps_native_constants
);
5124 if (gl_info
->supported
[NV_FRAGMENT_PROGRAM2
])
5127 TRACE("Hardware pixel shader version 3.0 enabled (NV_FRAGMENT_PROGRAM2)\n");
5129 else if (ps_consts
>= 32)
5131 /* Shader Model 2.0 requires at least 32 pixel shader constants */
5133 TRACE("Hardware pixel shader version 2.0 enabled (ARB_PROGRAM)\n");
5138 TRACE("Hardware pixel shader version 1.4 enabled (ARB_PROGRAM)\n");
5140 caps
->ps_version
= min(wined3d_settings
.max_sm_ps
, ps_version
);
5141 caps
->ps_uniform_count
= min(WINED3D_MAX_PS_CONSTS_F
, ps_consts
);
5142 caps
->ps_1x_max_value
= 8.0f
;
5146 caps
->ps_version
= 0;
5147 caps
->ps_uniform_count
= 0;
5148 caps
->ps_1x_max_value
= 0.0f
;
5151 caps
->varying_count
= 0;
5152 caps
->wined3d_caps
= WINED3D_SHADER_CAP_SRGB_WRITE
;
5153 if (use_nv_clip(gl_info
))
5154 caps
->wined3d_caps
|= WINED3D_SHADER_CAP_VS_CLIPPING
;
5157 static BOOL
shader_arb_color_fixup_supported(struct color_fixup_desc fixup
)
5159 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
5161 TRACE("Checking support for color_fixup:\n");
5162 dump_color_fixup_desc(fixup
);
5165 /* We support everything except complex conversions. */
5166 if (!is_complex_fixup(fixup
))
5172 TRACE("[FAILED]\n");
5176 static void shader_arb_add_instruction_modifiers(const struct wined3d_shader_instruction
*ins
) {
5178 char write_mask
[20], regstr
[50];
5179 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
5180 BOOL is_color
= FALSE
;
5181 const struct wined3d_shader_dst_param
*dst
;
5183 if (!ins
->dst_count
) return;
5187 if (!shift
) return; /* Saturate alone is handled by the instructions */
5189 shader_arb_get_write_mask(ins
, dst
, write_mask
);
5190 shader_arb_get_register_name(ins
, &dst
->reg
, regstr
, &is_color
);
5192 /* Generate a line that does the output modifier computation
5193 * FIXME: _SAT vs shift? _SAT alone is already handled in the instructions, if this
5194 * maps problems in e.g. _d4_sat modify shader_arb_get_modifier
5196 shader_addline(buffer
, "MUL%s %s%s, %s, %s;\n", shader_arb_get_modifier(ins
),
5197 regstr
, write_mask
, regstr
, shift_tab
[shift
]);
5200 static const SHADER_HANDLER shader_arb_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
5202 /* WINED3DSIH_ABS */ shader_hw_map2gl
,
5203 /* WINED3DSIH_ADD */ shader_hw_map2gl
,
5204 /* WINED3DSIH_AND */ NULL
,
5205 /* WINED3DSIH_ATOMIC_AND */ NULL
,
5206 /* WINED3DSIH_ATOMIC_IADD */ NULL
,
5207 /* WINED3DSIH_ATOMIC_OR */ NULL
,
5208 /* WINED3DSIH_ATOMIC_XOR */ NULL
,
5209 /* WINED3DSIH_BEM */ pshader_hw_bem
,
5210 /* WINED3DSIH_BFI */ NULL
,
5211 /* WINED3DSIH_BFREV */ NULL
,
5212 /* WINED3DSIH_BREAK */ shader_hw_break
,
5213 /* WINED3DSIH_BREAKC */ shader_hw_breakc
,
5214 /* WINED3DSIH_BREAKP */ NULL
,
5215 /* WINED3DSIH_BUFINFO */ NULL
,
5216 /* WINED3DSIH_CALL */ shader_hw_call
,
5217 /* WINED3DSIH_CALLNZ */ NULL
,
5218 /* WINED3DSIH_CASE */ NULL
,
5219 /* WINED3DSIH_CMP */ pshader_hw_cmp
,
5220 /* WINED3DSIH_CND */ pshader_hw_cnd
,
5221 /* WINED3DSIH_CONTINUE */ NULL
,
5222 /* WINED3DSIH_CRS */ shader_hw_map2gl
,
5223 /* WINED3DSIH_CUT */ NULL
,
5224 /* WINED3DSIH_CUT_STREAM */ NULL
,
5225 /* WINED3DSIH_DCL */ shader_hw_nop
,
5226 /* WINED3DSIH_DCL_CONSTANT_BUFFER */ shader_hw_nop
,
5227 /* WINED3DSIH_DCL_GLOBAL_FLAGS */ NULL
,
5228 /* WINED3DSIH_DCL_HS_FORK_PHASE_INSTANCE_COUNT */ NULL
,
5229 /* WINED3DSIH_DCL_HS_MAX_TESSFACTOR */ NULL
,
5230 /* WINED3DSIH_DCL_IMMEDIATE_CONSTANT_BUFFER */ NULL
,
5231 /* WINED3DSIH_DCL_INDEXABLE_TEMP */ NULL
,
5232 /* WINED3DSIH_DCL_INPUT */ NULL
,
5233 /* WINED3DSIH_DCL_INPUT_CONTROL_POINT_COUNT */ NULL
,
5234 /* WINED3DSIH_DCL_INPUT_PRIMITIVE */ shader_hw_nop
,
5235 /* WINED3DSIH_DCL_INPUT_PS */ NULL
,
5236 /* WINED3DSIH_DCL_INPUT_PS_SGV */ NULL
,
5237 /* WINED3DSIH_DCL_INPUT_PS_SIV */ NULL
,
5238 /* WINED3DSIH_DCL_INPUT_SGV */ NULL
,
5239 /* WINED3DSIH_DCL_INPUT_SIV */ NULL
,
5240 /* WINED3DSIH_DCL_OUTPUT */ NULL
,
5241 /* WINED3DSIH_DCL_OUTPUT_CONTROL_POINT_COUNT */ NULL
,
5242 /* WINED3DSIH_DCL_OUTPUT_SIV */ NULL
,
5243 /* WINED3DSIH_DCL_OUTPUT_TOPOLOGY */ shader_hw_nop
,
5244 /* WINED3DSIH_DCL_RESOURCE_STRUCTURED */ NULL
,
5245 /* WINED3DSIH_DCL_SAMPLER */ NULL
,
5246 /* WINED3DSIH_DCL_STREAM */ NULL
,
5247 /* WINED3DSIH_DCL_TEMPS */ NULL
,
5248 /* WINED3DSIH_DCL_TESSELLATOR_DOMAIN */ NULL
,
5249 /* WINED3DSIH_DCL_TESSELLATOR_OUTPUT_PRIMITIVE */ NULL
,
5250 /* WINED3DSIH_DCL_TESSELLATOR_PARTITIONING */ NULL
,
5251 /* WINED3DSIH_DCL_TGSM_RAW */ NULL
,
5252 /* WINED3DSIH_DCL_TGSM_STRUCTURED */ NULL
,
5253 /* WINED3DSIH_DCL_THREAD_GROUP */ NULL
,
5254 /* WINED3DSIH_DCL_UAV_RAW */ NULL
,
5255 /* WINED3DSIH_DCL_UAV_STRUCTURED */ NULL
,
5256 /* WINED3DSIH_DCL_UAV_TYPED */ NULL
,
5257 /* WINED3DSIH_DCL_VERTICES_OUT */ shader_hw_nop
,
5258 /* WINED3DSIH_DEF */ shader_hw_nop
,
5259 /* WINED3DSIH_DEFAULT */ NULL
,
5260 /* WINED3DSIH_DEFB */ shader_hw_nop
,
5261 /* WINED3DSIH_DEFI */ shader_hw_nop
,
5262 /* WINED3DSIH_DIV */ NULL
,
5263 /* WINED3DSIH_DP2 */ NULL
,
5264 /* WINED3DSIH_DP2ADD */ pshader_hw_dp2add
,
5265 /* WINED3DSIH_DP3 */ shader_hw_map2gl
,
5266 /* WINED3DSIH_DP4 */ shader_hw_map2gl
,
5267 /* WINED3DSIH_DST */ shader_hw_map2gl
,
5268 /* WINED3DSIH_DSX */ shader_hw_map2gl
,
5269 /* WINED3DSIH_DSX_COARSE */ NULL
,
5270 /* WINED3DSIH_DSX_FINE */ NULL
,
5271 /* WINED3DSIH_DSY */ shader_hw_dsy
,
5272 /* WINED3DSIH_DSY_COARSE */ NULL
,
5273 /* WINED3DSIH_DSY_FINE */ NULL
,
5274 /* WINED3DSIH_ELSE */ shader_hw_else
,
5275 /* WINED3DSIH_EMIT */ NULL
,
5276 /* WINED3DSIH_EMIT_STREAM */ NULL
,
5277 /* WINED3DSIH_ENDIF */ shader_hw_endif
,
5278 /* WINED3DSIH_ENDLOOP */ shader_hw_endloop
,
5279 /* WINED3DSIH_ENDREP */ shader_hw_endrep
,
5280 /* WINED3DSIH_ENDSWITCH */ NULL
,
5281 /* WINED3DSIH_EQ */ NULL
,
5282 /* WINED3DSIH_EXP */ shader_hw_scalar_op
,
5283 /* WINED3DSIH_EXPP */ shader_hw_scalar_op
,
5284 /* WINED3DSIH_FRC */ shader_hw_map2gl
,
5285 /* WINED3DSIH_FTOI */ NULL
,
5286 /* WINED3DSIH_FTOU */ NULL
,
5287 /* WINED3DSIH_GATHER4 */ NULL
,
5288 /* WINED3DSIH_GATHER4_C */ NULL
,
5289 /* WINED3DSIH_GE */ NULL
,
5290 /* WINED3DSIH_HS_CONTROL_POINT_PHASE */ NULL
,
5291 /* WINED3DSIH_HS_DECLS */ NULL
,
5292 /* WINED3DSIH_HS_FORK_PHASE */ NULL
,
5293 /* WINED3DSIH_HS_JOIN_PHASE */ NULL
,
5294 /* WINED3DSIH_IADD */ NULL
,
5295 /* WINED3DSIH_IEQ */ NULL
,
5296 /* WINED3DSIH_IF */ NULL
/* Hardcoded into the shader */,
5297 /* WINED3DSIH_IFC */ shader_hw_ifc
,
5298 /* WINED3DSIH_IGE */ NULL
,
5299 /* WINED3DSIH_ILT */ NULL
,
5300 /* WINED3DSIH_IMAD */ NULL
,
5301 /* WINED3DSIH_IMAX */ NULL
,
5302 /* WINED3DSIH_IMIN */ NULL
,
5303 /* WINED3DSIH_IMM_ATOMIC_ALLOC */ NULL
,
5304 /* WINED3DSIH_IMM_ATOMIC_AND */ NULL
,
5305 /* WINED3DSIH_IMM_ATOMIC_CONSUME */ NULL
,
5306 /* WINED3DSIH_IMM_ATOMIC_OR */ NULL
,
5307 /* WINED3DSIH_IMM_ATOMIC_XOR */ NULL
,
5308 /* WINED3DSIH_IMUL */ NULL
,
5309 /* WINED3DSIH_INE */ NULL
,
5310 /* WINED3DSIH_INEG */ NULL
,
5311 /* WINED3DSIH_ISHL */ NULL
,
5312 /* WINED3DSIH_ISHR */ NULL
,
5313 /* WINED3DSIH_ITOF */ NULL
,
5314 /* WINED3DSIH_LABEL */ shader_hw_label
,
5315 /* WINED3DSIH_LD */ NULL
,
5316 /* WINED3DSIH_LD2DMS */ NULL
,
5317 /* WINED3DSIH_LD_RAW */ NULL
,
5318 /* WINED3DSIH_LD_STRUCTURED */ NULL
,
5319 /* WINED3DSIH_LD_UAV_TYPED */ NULL
,
5320 /* WINED3DSIH_LIT */ shader_hw_map2gl
,
5321 /* WINED3DSIH_LOD */ NULL
,
5322 /* WINED3DSIH_LOG */ shader_hw_scalar_op
,
5323 /* WINED3DSIH_LOGP */ shader_hw_scalar_op
,
5324 /* WINED3DSIH_LOOP */ shader_hw_loop
,
5325 /* WINED3DSIH_LRP */ shader_hw_lrp
,
5326 /* WINED3DSIH_LT */ NULL
,
5327 /* WINED3DSIH_M3x2 */ shader_hw_mnxn
,
5328 /* WINED3DSIH_M3x3 */ shader_hw_mnxn
,
5329 /* WINED3DSIH_M3x4 */ shader_hw_mnxn
,
5330 /* WINED3DSIH_M4x3 */ shader_hw_mnxn
,
5331 /* WINED3DSIH_M4x4 */ shader_hw_mnxn
,
5332 /* WINED3DSIH_MAD */ shader_hw_map2gl
,
5333 /* WINED3DSIH_MAX */ shader_hw_map2gl
,
5334 /* WINED3DSIH_MIN */ shader_hw_map2gl
,
5335 /* WINED3DSIH_MOV */ shader_hw_mov
,
5336 /* WINED3DSIH_MOVA */ shader_hw_mov
,
5337 /* WINED3DSIH_MOVC */ NULL
,
5338 /* WINED3DSIH_MUL */ shader_hw_map2gl
,
5339 /* WINED3DSIH_NE */ NULL
,
5340 /* WINED3DSIH_NOP */ shader_hw_nop
,
5341 /* WINED3DSIH_NOT */ NULL
,
5342 /* WINED3DSIH_NRM */ shader_hw_nrm
,
5343 /* WINED3DSIH_OR */ NULL
,
5344 /* WINED3DSIH_PHASE */ shader_hw_nop
,
5345 /* WINED3DSIH_POW */ shader_hw_pow
,
5346 /* WINED3DSIH_RCP */ shader_hw_scalar_op
,
5347 /* WINED3DSIH_REP */ shader_hw_rep
,
5348 /* WINED3DSIH_RESINFO */ NULL
,
5349 /* WINED3DSIH_RET */ shader_hw_ret
,
5350 /* WINED3DSIH_ROUND_NE */ NULL
,
5351 /* WINED3DSIH_ROUND_NI */ NULL
,
5352 /* WINED3DSIH_ROUND_PI */ NULL
,
5353 /* WINED3DSIH_ROUND_Z */ NULL
,
5354 /* WINED3DSIH_RSQ */ shader_hw_scalar_op
,
5355 /* WINED3DSIH_SAMPLE */ NULL
,
5356 /* WINED3DSIH_SAMPLE_B */ NULL
,
5357 /* WINED3DSIH_SAMPLE_C */ NULL
,
5358 /* WINED3DSIH_SAMPLE_C_LZ */ NULL
,
5359 /* WINED3DSIH_SAMPLE_GRAD */ NULL
,
5360 /* WINED3DSIH_SAMPLE_INFO */ NULL
,
5361 /* WINED3DSIH_SAMPLE_LOD */ NULL
,
5362 /* WINED3DSIH_SAMPLE_POS */ NULL
,
5363 /* WINED3DSIH_SETP */ NULL
,
5364 /* WINED3DSIH_SGE */ shader_hw_map2gl
,
5365 /* WINED3DSIH_SGN */ shader_hw_sgn
,
5366 /* WINED3DSIH_SINCOS */ shader_hw_sincos
,
5367 /* WINED3DSIH_SLT */ shader_hw_map2gl
,
5368 /* WINED3DSIH_SQRT */ NULL
,
5369 /* WINED3DSIH_STORE_RAW */ NULL
,
5370 /* WINED3DSIH_STORE_STRUCTURED */ NULL
,
5371 /* WINED3DSIH_STORE_UAV_TYPED */ NULL
,
5372 /* WINED3DSIH_SUB */ shader_hw_map2gl
,
5373 /* WINED3DSIH_SWAPC */ NULL
,
5374 /* WINED3DSIH_SWITCH */ NULL
,
5375 /* WINED3DSIH_SYNC */ NULL
,
5376 /* WINED3DSIH_TEX */ pshader_hw_tex
,
5377 /* WINED3DSIH_TEXBEM */ pshader_hw_texbem
,
5378 /* WINED3DSIH_TEXBEML */ pshader_hw_texbem
,
5379 /* WINED3DSIH_TEXCOORD */ pshader_hw_texcoord
,
5380 /* WINED3DSIH_TEXDEPTH */ pshader_hw_texdepth
,
5381 /* WINED3DSIH_TEXDP3 */ pshader_hw_texdp3
,
5382 /* WINED3DSIH_TEXDP3TEX */ pshader_hw_texdp3tex
,
5383 /* WINED3DSIH_TEXKILL */ pshader_hw_texkill
,
5384 /* WINED3DSIH_TEXLDD */ shader_hw_texldd
,
5385 /* WINED3DSIH_TEXLDL */ shader_hw_texldl
,
5386 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_hw_texm3x2depth
,
5387 /* WINED3DSIH_TEXM3x2PAD */ pshader_hw_texm3x2pad
,
5388 /* WINED3DSIH_TEXM3x2TEX */ pshader_hw_texm3x2tex
,
5389 /* WINED3DSIH_TEXM3x3 */ pshader_hw_texm3x3
,
5390 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
5391 /* WINED3DSIH_TEXM3x3PAD */ pshader_hw_texm3x3pad
,
5392 /* WINED3DSIH_TEXM3x3SPEC */ pshader_hw_texm3x3spec
,
5393 /* WINED3DSIH_TEXM3x3TEX */ pshader_hw_texm3x3tex
,
5394 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_hw_texm3x3vspec
,
5395 /* WINED3DSIH_TEXREG2AR */ pshader_hw_texreg2ar
,
5396 /* WINED3DSIH_TEXREG2GB */ pshader_hw_texreg2gb
,
5397 /* WINED3DSIH_TEXREG2RGB */ pshader_hw_texreg2rgb
,
5398 /* WINED3DSIH_UBFE */ NULL
,
5399 /* WINED3DSIH_UDIV */ NULL
,
5400 /* WINED3DSIH_UGE */ NULL
,
5401 /* WINED3DSIH_ULT */ NULL
,
5402 /* WINED3DSIH_UMAX */ NULL
,
5403 /* WINED3DSIH_UMIN */ NULL
,
5404 /* WINED3DSIH_USHR */ NULL
,
5405 /* WINED3DSIH_UTOF */ NULL
,
5406 /* WINED3DSIH_XOR */ NULL
,
5409 static BOOL
get_bool_const(const struct wined3d_shader_instruction
*ins
,
5410 const struct wined3d_shader
*shader
, DWORD idx
)
5412 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
5413 BOOL vshader
= shader_is_vshader_version(reg_maps
->shader_version
.type
);
5414 const struct wined3d_shader_lconst
*constant
;
5416 WORD flag
= (1u << idx
);
5417 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5419 if (reg_maps
->local_bool_consts
& flag
)
5421 /* What good is an if(bool) with a hardcoded local constant? I don't know, but handle it */
5422 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsB
, struct wined3d_shader_lconst
, entry
)
5424 if (constant
->idx
== idx
)
5426 return constant
->value
[0];
5429 ERR("Local constant not found\n");
5434 if(vshader
) bools
= priv
->cur_vs_args
->clip
.boolclip
.bools
;
5435 else bools
= priv
->cur_ps_args
->bools
;
5436 return bools
& flag
;
5440 static void get_loop_control_const(const struct wined3d_shader_instruction
*ins
,
5441 const struct wined3d_shader
*shader
, UINT idx
, struct wined3d_shader_loop_control
*loop_control
)
5443 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
5444 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5446 /* Integer constants can either be a local constant, or they can be stored in the shader
5447 * type specific compile args. */
5448 if (reg_maps
->local_int_consts
& (1u << idx
))
5450 const struct wined3d_shader_lconst
*constant
;
5452 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsI
, struct wined3d_shader_lconst
, entry
)
5454 if (constant
->idx
== idx
)
5456 loop_control
->count
= constant
->value
[0];
5457 loop_control
->start
= constant
->value
[1];
5458 /* Step is signed. */
5459 loop_control
->step
= (int)constant
->value
[2];
5463 /* If this happens the flag was set incorrectly */
5464 ERR("Local constant not found\n");
5465 loop_control
->count
= 0;
5466 loop_control
->start
= 0;
5467 loop_control
->step
= 0;
5471 switch (reg_maps
->shader_version
.type
)
5473 case WINED3D_SHADER_TYPE_VERTEX
:
5474 /* Count and aL start value are unsigned */
5475 loop_control
->count
= priv
->cur_vs_args
->loop_ctrl
[idx
][0];
5476 loop_control
->start
= priv
->cur_vs_args
->loop_ctrl
[idx
][1];
5477 /* Step is signed. */
5478 loop_control
->step
= ((char)priv
->cur_vs_args
->loop_ctrl
[idx
][2]);
5481 case WINED3D_SHADER_TYPE_PIXEL
:
5482 loop_control
->count
= priv
->cur_ps_args
->loop_ctrl
[idx
][0];
5483 loop_control
->start
= priv
->cur_ps_args
->loop_ctrl
[idx
][1];
5484 loop_control
->step
= ((char)priv
->cur_ps_args
->loop_ctrl
[idx
][2]);
5488 FIXME("Unhandled shader type %#x.\n", reg_maps
->shader_version
.type
);
5493 static void record_instruction(struct list
*list
, const struct wined3d_shader_instruction
*ins
)
5496 struct wined3d_shader_dst_param
*dst_param
;
5497 struct wined3d_shader_src_param
*src_param
= NULL
, *rel_addr
;
5498 struct recorded_instruction
*rec
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*rec
));
5501 ERR("Out of memory\n");
5506 dst_param
= HeapAlloc(GetProcessHeap(), 0, sizeof(*dst_param
));
5507 if(!dst_param
) goto free
;
5508 *dst_param
= *ins
->dst
;
5509 if (ins
->dst
->reg
.idx
[0].rel_addr
)
5511 rel_addr
= HeapAlloc(GetProcessHeap(), 0, sizeof(*rel_addr
));
5514 *rel_addr
= *ins
->dst
->reg
.idx
[0].rel_addr
;
5515 dst_param
->reg
.idx
[0].rel_addr
= rel_addr
;
5517 rec
->ins
.dst
= dst_param
;
5519 if (!(src_param
= wined3d_calloc(ins
->src_count
, sizeof(*src_param
))))
5521 for (i
= 0; i
< ins
->src_count
; ++i
)
5523 src_param
[i
] = ins
->src
[i
];
5524 if (ins
->src
[i
].reg
.idx
[0].rel_addr
)
5526 rel_addr
= HeapAlloc(GetProcessHeap(), 0, sizeof(*rel_addr
));
5529 *rel_addr
= *ins
->src
[i
].reg
.idx
[0].rel_addr
;
5530 src_param
[i
].reg
.idx
[0].rel_addr
= rel_addr
;
5533 rec
->ins
.src
= src_param
;
5534 list_add_tail(list
, &rec
->entry
);
5538 ERR("Out of memory\n");
5541 HeapFree(GetProcessHeap(), 0, (void *)dst_param
->reg
.idx
[0].rel_addr
);
5542 HeapFree(GetProcessHeap(), 0, dst_param
);
5546 for(i
= 0; i
< ins
->src_count
; i
++)
5548 HeapFree(GetProcessHeap(), 0, (void *)src_param
[i
].reg
.idx
[0].rel_addr
);
5550 HeapFree(GetProcessHeap(), 0, src_param
);
5552 HeapFree(GetProcessHeap(), 0, rec
);
5555 static void free_recorded_instruction(struct list
*list
)
5557 struct recorded_instruction
*rec_ins
, *entry2
;
5560 LIST_FOR_EACH_ENTRY_SAFE(rec_ins
, entry2
, list
, struct recorded_instruction
, entry
)
5562 list_remove(&rec_ins
->entry
);
5563 if (rec_ins
->ins
.dst
)
5565 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.dst
->reg
.idx
[0].rel_addr
);
5566 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.dst
);
5568 if (rec_ins
->ins
.src
)
5570 for (i
= 0; i
< rec_ins
->ins
.src_count
; ++i
)
5572 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.src
[i
].reg
.idx
[0].rel_addr
);
5574 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.src
);
5576 HeapFree(GetProcessHeap(), 0, rec_ins
);
5580 static void pop_control_frame(const struct wined3d_shader_instruction
*ins
)
5582 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5583 struct control_frame
*control_frame
;
5585 if (ins
->handler_idx
== WINED3DSIH_ENDLOOP
|| ins
->handler_idx
== WINED3DSIH_ENDREP
)
5587 struct list
*e
= list_head(&priv
->control_frames
);
5588 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5589 list_remove(&control_frame
->entry
);
5590 HeapFree(GetProcessHeap(), 0, control_frame
);
5593 else if (ins
->handler_idx
== WINED3DSIH_ENDIF
)
5595 /* Non-ifc ENDIFs were already handled previously. */
5596 struct list
*e
= list_head(&priv
->control_frames
);
5597 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5598 list_remove(&control_frame
->entry
);
5599 HeapFree(GetProcessHeap(), 0, control_frame
);
5603 static void shader_arb_handle_instruction(const struct wined3d_shader_instruction
*ins
) {
5604 SHADER_HANDLER hw_fct
;
5605 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5606 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
5607 struct control_frame
*control_frame
;
5608 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
5611 if(ins
->handler_idx
== WINED3DSIH_LOOP
|| ins
->handler_idx
== WINED3DSIH_REP
)
5613 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
5614 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5616 if(ins
->handler_idx
== WINED3DSIH_LOOP
) control_frame
->type
= LOOP
;
5617 if(ins
->handler_idx
== WINED3DSIH_REP
) control_frame
->type
= REP
;
5619 if(priv
->target_version
>= NV2
)
5621 control_frame
->no
.loop
= priv
->num_loops
++;
5626 /* Don't bother recording when we're in a not used if branch */
5632 if(!priv
->recording
)
5634 list_init(&priv
->record
);
5635 priv
->recording
= TRUE
;
5636 control_frame
->outer_loop
= TRUE
;
5637 get_loop_control_const(ins
, shader
, ins
->src
[0].reg
.idx
[0].offset
, &control_frame
->loop_control
);
5638 return; /* Instruction is handled */
5640 /* Record this loop in the outer loop's recording */
5643 else if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
|| ins
->handler_idx
== WINED3DSIH_ENDREP
)
5645 if(priv
->target_version
>= NV2
)
5647 /* Nothing to do. The control frame is popped after the HW instr handler */
5651 struct list
*e
= list_head(&priv
->control_frames
);
5652 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5653 list_remove(&control_frame
->entry
);
5655 if(control_frame
->outer_loop
)
5657 unsigned int iteration
;
5661 /* Turn off recording before playback */
5662 priv
->recording
= FALSE
;
5664 /* Move the recorded instructions to a separate list and get them out of the private data
5665 * structure. If there are nested loops, the shader_arb_handle_instruction below will
5666 * be recorded again, thus priv->record might be overwritten
5669 list_move_tail(©
, &priv
->record
);
5670 list_init(&priv
->record
);
5672 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5674 shader_addline(buffer
, "#unrolling loop: %u iterations, aL=%u, inc %d\n",
5675 control_frame
->loop_control
.count
, control_frame
->loop_control
.start
,
5676 control_frame
->loop_control
.step
);
5677 aL
= control_frame
->loop_control
.start
;
5681 shader_addline(buffer
, "#unrolling rep: %u iterations\n", control_frame
->loop_control
.count
);
5684 for (iteration
= 0; iteration
< control_frame
->loop_control
.count
; ++iteration
)
5686 struct recorded_instruction
*rec_ins
;
5687 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5690 shader_addline(buffer
, "#Iteration %u, aL=%d\n", iteration
, aL
);
5694 shader_addline(buffer
, "#Iteration %u\n", iteration
);
5697 LIST_FOR_EACH_ENTRY(rec_ins
, ©
, struct recorded_instruction
, entry
)
5699 shader_arb_handle_instruction(&rec_ins
->ins
);
5702 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5704 aL
+= control_frame
->loop_control
.step
;
5707 shader_addline(buffer
, "#end loop/rep\n");
5709 free_recorded_instruction(©
);
5710 HeapFree(GetProcessHeap(), 0, control_frame
);
5711 return; /* Instruction is handled */
5715 /* This is a nested loop. Proceed to the normal recording function */
5716 HeapFree(GetProcessHeap(), 0, control_frame
);
5723 record_instruction(&priv
->record
, ins
);
5728 if(ins
->handler_idx
== WINED3DSIH_IF
)
5730 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
5731 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5732 control_frame
->type
= IF
;
5734 bool_const
= get_bool_const(ins
, shader
, ins
->src
[0].reg
.idx
[0].offset
);
5735 if (ins
->src
[0].modifiers
== WINED3DSPSM_NOT
)
5736 bool_const
= !bool_const
;
5737 if (!priv
->muted
&& !bool_const
)
5739 shader_addline(buffer
, "#if(FALSE){\n");
5741 control_frame
->muting
= TRUE
;
5743 else shader_addline(buffer
, "#if(TRUE) {\n");
5745 return; /* Instruction is handled */
5747 else if(ins
->handler_idx
== WINED3DSIH_IFC
)
5749 /* IF(bool) and if_cond(a, b) use the same ELSE and ENDIF tokens */
5750 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
5751 control_frame
->type
= IFC
;
5752 control_frame
->no
.ifc
= priv
->num_ifcs
++;
5753 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5755 else if(ins
->handler_idx
== WINED3DSIH_ELSE
)
5757 struct list
*e
= list_head(&priv
->control_frames
);
5758 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5760 if(control_frame
->type
== IF
)
5762 shader_addline(buffer
, "#} else {\n");
5763 if(!priv
->muted
&& !control_frame
->muting
)
5766 control_frame
->muting
= TRUE
;
5768 else if(control_frame
->muting
) priv
->muted
= FALSE
;
5769 return; /* Instruction is handled. */
5771 /* In case of an ifc, generate a HW shader instruction */
5772 if (control_frame
->type
!= IFC
)
5773 ERR("Control frame does not match.\n");
5775 else if(ins
->handler_idx
== WINED3DSIH_ENDIF
)
5777 struct list
*e
= list_head(&priv
->control_frames
);
5778 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5780 if(control_frame
->type
== IF
)
5782 shader_addline(buffer
, "#} endif\n");
5783 if(control_frame
->muting
) priv
->muted
= FALSE
;
5784 list_remove(&control_frame
->entry
);
5785 HeapFree(GetProcessHeap(), 0, control_frame
);
5786 return; /* Instruction is handled */
5788 /* In case of an ifc, generate a HW shader instruction */
5789 if (control_frame
->type
!= IFC
)
5790 ERR("Control frame does not match.\n");
5795 pop_control_frame(ins
);
5799 /* Select handler */
5800 hw_fct
= shader_arb_instruction_handler_table
[ins
->handler_idx
];
5802 /* Unhandled opcode */
5805 FIXME("Backend can't handle opcode %s.\n", debug_d3dshaderinstructionhandler(ins
->handler_idx
));
5810 pop_control_frame(ins
);
5812 shader_arb_add_instruction_modifiers(ins
);
5815 static BOOL
shader_arb_has_ffp_proj_control(void *shader_priv
)
5817 struct shader_arb_priv
*priv
= shader_priv
;
5819 return priv
->ffp_proj_control
;
5822 const struct wined3d_shader_backend_ops arb_program_shader_backend
=
5824 shader_arb_handle_instruction
,
5827 shader_arb_select_depth_blt
,
5828 shader_arb_deselect_depth_blt
,
5829 shader_arb_update_float_vertex_constants
,
5830 shader_arb_update_float_pixel_constants
,
5831 shader_arb_load_constants
,
5835 shader_arb_allocate_context_data
,
5836 shader_arb_free_context_data
,
5837 shader_arb_init_context_state
,
5838 shader_arb_get_caps
,
5839 shader_arb_color_fixup_supported
,
5840 shader_arb_has_ffp_proj_control
,
5843 /* ARB_fragment_program fixed function pipeline replacement definitions */
5844 #define ARB_FFP_CONST_TFACTOR 0
5845 #define ARB_FFP_CONST_COLOR_KEY_LOW ((ARB_FFP_CONST_TFACTOR) + 1)
5846 #define ARB_FFP_CONST_COLOR_KEY_HIGH ((ARB_FFP_CONST_COLOR_KEY_LOW) + 1)
5847 #define ARB_FFP_CONST_SPECULAR_ENABLE ((ARB_FFP_CONST_COLOR_KEY_HIGH) + 1)
5848 #define ARB_FFP_CONST_CONSTANT(i) ((ARB_FFP_CONST_SPECULAR_ENABLE) + 1 + i)
5849 #define ARB_FFP_CONST_BUMPMAT(i) ((ARB_FFP_CONST_CONSTANT(7)) + 1 + i)
5850 #define ARB_FFP_CONST_LUMINANCE(i) ((ARB_FFP_CONST_BUMPMAT(7)) + 1 + i)
5852 struct arbfp_ffp_desc
5854 struct ffp_frag_desc parent
;
5858 /* Context activation is done by the caller. */
5859 static void arbfp_enable(const struct wined3d_gl_info
*gl_info
, BOOL enable
)
5863 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
5864 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
5868 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
5869 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
5873 static void *arbfp_alloc(const struct wined3d_shader_backend_ops
*shader_backend
, void *shader_priv
)
5875 struct shader_arb_priv
*priv
;
5877 /* Share private data between the shader backend and the pipeline
5878 * replacement, if both are the arb implementation. This is needed to
5879 * figure out whether ARBfp should be disabled if no pixel shader is bound
5881 if (shader_backend
== &arb_program_shader_backend
)
5883 else if (!(priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*priv
))))
5886 wine_rb_init(&priv
->fragment_shaders
, wined3d_ffp_frag_program_key_compare
);
5887 priv
->use_arbfp_fixed_func
= TRUE
;
5892 /* Context activation is done by the caller. */
5893 static void arbfp_free_ffpshader(struct wine_rb_entry
*entry
, void *context
)
5895 const struct wined3d_gl_info
*gl_info
= context
;
5896 struct arbfp_ffp_desc
*entry_arb
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_ffp_desc
, parent
.entry
);
5898 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb
->shader
));
5899 checkGLcall("glDeleteProgramsARB(1, &entry_arb->shader)");
5900 HeapFree(GetProcessHeap(), 0, entry_arb
);
5903 /* Context activation is done by the caller. */
5904 static void arbfp_free(struct wined3d_device
*device
)
5906 struct shader_arb_priv
*priv
= device
->fragment_priv
;
5908 wine_rb_destroy(&priv
->fragment_shaders
, arbfp_free_ffpshader
, &device
->adapter
->gl_info
);
5909 priv
->use_arbfp_fixed_func
= FALSE
;
5911 if (device
->shader_backend
!= &arb_program_shader_backend
)
5913 HeapFree(GetProcessHeap(), 0, device
->fragment_priv
);
5917 static void arbfp_get_caps(const struct wined3d_gl_info
*gl_info
, struct fragment_caps
*caps
)
5919 caps
->wined3d_caps
= WINED3D_FRAGMENT_CAP_PROJ_CONTROL
5920 | WINED3D_FRAGMENT_CAP_SRGB_WRITE
5921 | WINED3D_FRAGMENT_CAP_COLOR_KEY
;
5922 caps
->PrimitiveMiscCaps
= WINED3DPMISCCAPS_TSSARGTEMP
;
5923 caps
->TextureOpCaps
= WINED3DTEXOPCAPS_DISABLE
|
5924 WINED3DTEXOPCAPS_SELECTARG1
|
5925 WINED3DTEXOPCAPS_SELECTARG2
|
5926 WINED3DTEXOPCAPS_MODULATE4X
|
5927 WINED3DTEXOPCAPS_MODULATE2X
|
5928 WINED3DTEXOPCAPS_MODULATE
|
5929 WINED3DTEXOPCAPS_ADDSIGNED2X
|
5930 WINED3DTEXOPCAPS_ADDSIGNED
|
5931 WINED3DTEXOPCAPS_ADD
|
5932 WINED3DTEXOPCAPS_SUBTRACT
|
5933 WINED3DTEXOPCAPS_ADDSMOOTH
|
5934 WINED3DTEXOPCAPS_BLENDCURRENTALPHA
|
5935 WINED3DTEXOPCAPS_BLENDFACTORALPHA
|
5936 WINED3DTEXOPCAPS_BLENDTEXTUREALPHA
|
5937 WINED3DTEXOPCAPS_BLENDDIFFUSEALPHA
|
5938 WINED3DTEXOPCAPS_BLENDTEXTUREALPHAPM
|
5939 WINED3DTEXOPCAPS_MODULATEALPHA_ADDCOLOR
|
5940 WINED3DTEXOPCAPS_MODULATECOLOR_ADDALPHA
|
5941 WINED3DTEXOPCAPS_MODULATEINVCOLOR_ADDALPHA
|
5942 WINED3DTEXOPCAPS_MODULATEINVALPHA_ADDCOLOR
|
5943 WINED3DTEXOPCAPS_DOTPRODUCT3
|
5944 WINED3DTEXOPCAPS_MULTIPLYADD
|
5945 WINED3DTEXOPCAPS_LERP
|
5946 WINED3DTEXOPCAPS_BUMPENVMAP
|
5947 WINED3DTEXOPCAPS_BUMPENVMAPLUMINANCE
;
5949 /* TODO: Implement WINED3DTEXOPCAPS_PREMODULATE */
5951 caps
->MaxTextureBlendStages
= MAX_TEXTURES
;
5952 caps
->MaxSimultaneousTextures
= min(gl_info
->limits
.fragment_samplers
, MAX_TEXTURES
);
5955 static DWORD
arbfp_get_emul_mask(const struct wined3d_gl_info
*gl_info
)
5957 return GL_EXT_EMUL_ARB_MULTITEXTURE
| GL_EXT_EMUL_EXT_FOG_COORD
;
5960 static void state_texfactor_arbfp(struct wined3d_context
*context
,
5961 const struct wined3d_state
*state
, DWORD state_id
)
5963 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5964 struct wined3d_device
*device
= context
->device
;
5965 struct wined3d_color color
;
5967 if (device
->shader_backend
== &arb_program_shader_backend
)
5969 struct shader_arb_priv
*priv
;
5971 /* Don't load the parameter if we're using an arbfp pixel shader,
5972 * otherwise we'll overwrite application provided constants. */
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_tss_constant_arbfp(struct wined3d_context
*context
,
5987 const struct wined3d_state
*state
, DWORD state_id
)
5989 DWORD stage
= (state_id
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
5990 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5991 struct wined3d_device
*device
= context
->device
;
5992 struct wined3d_color color
;
5994 if (device
->shader_backend
== &arb_program_shader_backend
)
5996 struct shader_arb_priv
*priv
;
5998 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
5999 * application provided constants.
6004 priv
= device
->shader_priv
;
6005 priv
->pshader_const_dirty
[ARB_FFP_CONST_CONSTANT(stage
)] = 1;
6006 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_CONSTANT(stage
) + 1);
6009 wined3d_color_from_d3dcolor(&color
, state
->texture_states
[stage
][WINED3D_TSS_CONSTANT
]);
6010 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_CONSTANT(stage
), &color
.r
));
6011 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_CONSTANT(stage), &color.r)");
6014 static void state_arb_specularenable(struct wined3d_context
*context
,
6015 const struct wined3d_state
*state
, DWORD state_id
)
6017 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
6018 struct wined3d_device
*device
= context
->device
;
6021 if (device
->shader_backend
== &arb_program_shader_backend
)
6023 struct shader_arb_priv
*priv
;
6025 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
6026 * application provided constants.
6031 priv
= device
->shader_priv
;
6032 priv
->pshader_const_dirty
[ARB_FFP_CONST_SPECULAR_ENABLE
] = 1;
6033 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_SPECULAR_ENABLE
+ 1);
6036 if (state
->render_states
[WINED3D_RS_SPECULARENABLE
])
6038 /* The specular color has no alpha */
6039 col
[0] = 1.0f
; col
[1] = 1.0f
;
6040 col
[2] = 1.0f
; col
[3] = 0.0f
;
6042 col
[0] = 0.0f
; col
[1] = 0.0f
;
6043 col
[2] = 0.0f
; col
[3] = 0.0f
;
6045 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_SPECULAR_ENABLE
, col
));
6046 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_SPECULAR_ENABLE, col)");
6049 static void set_bumpmat_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6051 DWORD stage
= (state_id
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
6052 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
6053 struct wined3d_device
*device
= context
->device
;
6056 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_BUMP_ENV
;
6058 if (device
->shader_backend
== &arb_program_shader_backend
)
6060 struct shader_arb_priv
*priv
= device
->shader_priv
;
6062 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants. */
6066 priv
->pshader_const_dirty
[ARB_FFP_CONST_BUMPMAT(stage
)] = 1;
6067 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_BUMPMAT(stage
) + 1);
6070 mat
[0][0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT00
]);
6071 mat
[0][1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT01
]);
6072 mat
[1][0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT10
]);
6073 mat
[1][1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT11
]);
6075 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_BUMPMAT(stage
), &mat
[0][0]));
6076 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_BUMPMAT(stage), &mat[0][0])");
6079 static void tex_bumpenvlum_arbfp(struct wined3d_context
*context
,
6080 const struct wined3d_state
*state
, DWORD state_id
)
6082 DWORD stage
= (state_id
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
6083 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
6084 struct wined3d_device
*device
= context
->device
;
6087 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_BUMP_ENV
;
6089 if (device
->shader_backend
== &arb_program_shader_backend
)
6091 struct shader_arb_priv
*priv
= device
->shader_priv
;
6093 /* Exit now, don't set the luminance below, otherwise we may overwrite pixel shader constants. */
6097 priv
->pshader_const_dirty
[ARB_FFP_CONST_LUMINANCE(stage
)] = 1;
6098 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_LUMINANCE(stage
) + 1);
6101 param
[0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_LSCALE
]);
6102 param
[1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_LOFFSET
]);
6106 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_LUMINANCE(stage
), param
));
6107 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_LUMINANCE(stage), param)");
6110 static void alpha_test_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6112 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
6116 TRACE("context %p, state %p, state_id %#x.\n", context
, state
, state_id
);
6118 if (state
->render_states
[WINED3D_RS_ALPHATESTENABLE
])
6120 gl_info
->gl_ops
.gl
.p_glEnable(GL_ALPHA_TEST
);
6121 checkGLcall("glEnable GL_ALPHA_TEST");
6125 gl_info
->gl_ops
.gl
.p_glDisable(GL_ALPHA_TEST
);
6126 checkGLcall("glDisable GL_ALPHA_TEST");
6130 ref
= ((float)state
->render_states
[WINED3D_RS_ALPHAREF
]) / 255.0f
;
6131 glParm
= wined3d_gl_compare_func(state
->render_states
[WINED3D_RS_ALPHAFUNC
]);
6135 gl_info
->gl_ops
.gl
.p_glAlphaFunc(glParm
, ref
);
6136 checkGLcall("glAlphaFunc");
6140 static void color_key_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6142 const struct wined3d_texture
*texture
= state
->textures
[0];
6143 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
6144 struct wined3d_device
*device
= context
->device
;
6145 struct wined3d_color float_key
[2];
6150 if (device
->shader_backend
== &arb_program_shader_backend
)
6152 struct shader_arb_priv
*priv
;
6154 /* Don't load the parameter if we're using an arbfp pixel shader,
6155 * otherwise we'll overwrite application provided constants. */
6159 priv
= device
->shader_priv
;
6160 priv
->pshader_const_dirty
[ARB_FFP_CONST_COLOR_KEY_LOW
] = 1;
6161 priv
->pshader_const_dirty
[ARB_FFP_CONST_COLOR_KEY_HIGH
] = 1;
6162 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_COLOR_KEY_HIGH
+ 1);
6165 wined3d_format_get_float_color_key(texture
->resource
.format
, &texture
->async
.src_blt_color_key
, float_key
);
6167 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_COLOR_KEY_LOW
, &float_key
[0].r
));
6168 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_COLOR_KEY_LOW, &float_key[0].r)");
6169 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_COLOR_KEY_HIGH
, &float_key
[1].r
));
6170 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_COLOR_KEY_HIGH, &float_key[1].r)");
6173 static const char *get_argreg(struct wined3d_string_buffer
*buffer
, DWORD argnum
, unsigned int stage
, DWORD arg
)
6177 if(arg
== ARG_UNUSED
) return "unused"; /* This is the marker for unused registers */
6179 switch(arg
& WINED3DTA_SELECTMASK
) {
6180 case WINED3DTA_DIFFUSE
:
6181 ret
= "fragment.color.primary"; break;
6183 case WINED3DTA_CURRENT
:
6187 case WINED3DTA_TEXTURE
:
6189 case 0: ret
= "tex0"; break;
6190 case 1: ret
= "tex1"; break;
6191 case 2: ret
= "tex2"; break;
6192 case 3: ret
= "tex3"; break;
6193 case 4: ret
= "tex4"; break;
6194 case 5: ret
= "tex5"; break;
6195 case 6: ret
= "tex6"; break;
6196 case 7: ret
= "tex7"; break;
6197 default: ret
= "unknown texture";
6201 case WINED3DTA_TFACTOR
:
6202 ret
= "tfactor"; break;
6204 case WINED3DTA_SPECULAR
:
6205 ret
= "fragment.color.secondary"; break;
6207 case WINED3DTA_TEMP
:
6208 ret
= "tempreg"; break;
6210 case WINED3DTA_CONSTANT
:
6212 case 0: ret
= "const0"; break;
6213 case 1: ret
= "const1"; break;
6214 case 2: ret
= "const2"; break;
6215 case 3: ret
= "const3"; break;
6216 case 4: ret
= "const4"; break;
6217 case 5: ret
= "const5"; break;
6218 case 6: ret
= "const6"; break;
6219 case 7: ret
= "const7"; break;
6220 default: ret
= "unknown constant";
6228 if(arg
& WINED3DTA_COMPLEMENT
) {
6229 shader_addline(buffer
, "SUB arg%u, const.x, %s;\n", argnum
, ret
);
6230 if(argnum
== 0) ret
= "arg0";
6231 if(argnum
== 1) ret
= "arg1";
6232 if(argnum
== 2) ret
= "arg2";
6234 if(arg
& WINED3DTA_ALPHAREPLICATE
) {
6235 shader_addline(buffer
, "MOV arg%u, %s.w;\n", argnum
, ret
);
6236 if(argnum
== 0) ret
= "arg0";
6237 if(argnum
== 1) ret
= "arg1";
6238 if(argnum
== 2) ret
= "arg2";
6243 static void gen_ffp_instr(struct wined3d_string_buffer
*buffer
, unsigned int stage
, BOOL color
,
6244 BOOL alpha
, DWORD dst
, DWORD op
, DWORD dw_arg0
, DWORD dw_arg1
, DWORD dw_arg2
)
6246 const char *dstmask
, *dstreg
, *arg0
, *arg1
, *arg2
;
6247 unsigned int mul
= 1;
6249 if(color
&& alpha
) dstmask
= "";
6250 else if(color
) dstmask
= ".xyz";
6251 else dstmask
= ".w";
6253 if(dst
== tempreg
) dstreg
= "tempreg";
6254 else dstreg
= "ret";
6256 arg0
= get_argreg(buffer
, 0, stage
, dw_arg0
);
6257 arg1
= get_argreg(buffer
, 1, stage
, dw_arg1
);
6258 arg2
= get_argreg(buffer
, 2, stage
, dw_arg2
);
6262 case WINED3D_TOP_DISABLE
:
6265 case WINED3D_TOP_SELECT_ARG2
:
6268 case WINED3D_TOP_SELECT_ARG1
:
6269 shader_addline(buffer
, "MOV %s%s, %s;\n", dstreg
, dstmask
, arg1
);
6272 case WINED3D_TOP_MODULATE_4X
:
6275 case WINED3D_TOP_MODULATE_2X
:
6278 case WINED3D_TOP_MODULATE
:
6279 shader_addline(buffer
, "MUL %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6282 case WINED3D_TOP_ADD_SIGNED_2X
:
6285 case WINED3D_TOP_ADD_SIGNED
:
6286 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
6289 case WINED3D_TOP_ADD
:
6290 shader_addline(buffer
, "ADD_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6293 case WINED3D_TOP_SUBTRACT
:
6294 shader_addline(buffer
, "SUB_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6297 case WINED3D_TOP_ADD_SMOOTH
:
6298 shader_addline(buffer
, "SUB arg1, const.x, %s;\n", arg1
);
6299 shader_addline(buffer
, "MAD_SAT %s%s, arg1, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6302 case WINED3D_TOP_BLEND_CURRENT_ALPHA
:
6303 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_CURRENT
);
6304 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6306 case WINED3D_TOP_BLEND_FACTOR_ALPHA
:
6307 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TFACTOR
);
6308 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6310 case WINED3D_TOP_BLEND_TEXTURE_ALPHA
:
6311 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
6312 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6314 case WINED3D_TOP_BLEND_DIFFUSE_ALPHA
:
6315 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_DIFFUSE
);
6316 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6319 case WINED3D_TOP_BLEND_TEXTURE_ALPHA_PM
:
6320 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
6321 shader_addline(buffer
, "SUB arg0.w, const.x, %s.w;\n", arg0
);
6322 shader_addline(buffer
, "MAD_SAT %s%s, %s, arg0.w, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6325 /* D3DTOP_PREMODULATE ???? */
6327 case WINED3D_TOP_MODULATE_INVALPHA_ADD_COLOR
:
6328 shader_addline(buffer
, "SUB arg0.w, const.x, %s;\n", arg1
);
6329 shader_addline(buffer
, "MAD_SAT %s%s, arg0.w, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6331 case WINED3D_TOP_MODULATE_ALPHA_ADD_COLOR
:
6332 shader_addline(buffer
, "MAD_SAT %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
6334 case WINED3D_TOP_MODULATE_INVCOLOR_ADD_ALPHA
:
6335 shader_addline(buffer
, "SUB arg0, const.x, %s;\n", arg1
);
6336 shader_addline(buffer
, "MAD_SAT %s%s, arg0, %s, %s.w;\n", dstreg
, dstmask
, arg2
, arg1
);
6338 case WINED3D_TOP_MODULATE_COLOR_ADD_ALPHA
:
6339 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s.w;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
6342 case WINED3D_TOP_DOTPRODUCT3
:
6344 shader_addline(buffer
, "SUB arg1, %s, const.w;\n", arg1
);
6345 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
6346 shader_addline(buffer
, "DP3_SAT %s%s, arg1, arg2;\n", dstreg
, dstmask
);
6349 case WINED3D_TOP_MULTIPLY_ADD
:
6350 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg0
);
6353 case WINED3D_TOP_LERP
:
6354 /* The msdn is not quite right here */
6355 shader_addline(buffer
, "LRP %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6358 case WINED3D_TOP_BUMPENVMAP
:
6359 case WINED3D_TOP_BUMPENVMAP_LUMINANCE
:
6360 /* Those are handled in the first pass of the shader(generation pass 1 and 2) already */
6364 FIXME("Unhandled texture op %08x\n", op
);
6368 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.y;\n", dstreg
, dstmask
, dstreg
);
6370 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.z;\n", dstreg
, dstmask
, dstreg
);
6373 static const char *arbfp_texture_target(enum wined3d_gl_resource_type type
)
6377 case WINED3D_GL_RES_TYPE_TEX_1D
:
6379 case WINED3D_GL_RES_TYPE_TEX_2D
:
6381 case WINED3D_GL_RES_TYPE_TEX_3D
:
6383 case WINED3D_GL_RES_TYPE_TEX_CUBE
:
6385 case WINED3D_GL_RES_TYPE_TEX_RECT
:
6388 return "unexpected_resource_type";
6392 static GLuint
gen_arbfp_ffp_shader(const struct ffp_frag_settings
*settings
, const struct wined3d_gl_info
*gl_info
)
6394 BYTE tex_read
= 0, bump_used
= 0, luminance_used
= 0, constant_used
= 0;
6395 BOOL tempreg_used
= FALSE
, tfactor_used
= FALSE
;
6396 unsigned int stage
, lowest_disabled_stage
;
6397 struct wined3d_string_buffer buffer
;
6398 struct color_fixup_masks masks
;
6399 BOOL custom_linear_fog
= FALSE
;
6400 const char *textype
, *instr
;
6401 DWORD arg0
, arg1
, arg2
;
6402 char colorcor_dst
[8];
6406 if (!string_buffer_init(&buffer
))
6408 ERR("Failed to initialize shader buffer.\n");
6412 shader_addline(&buffer
, "!!ARBfp1.0\n");
6414 if (settings
->color_key_enabled
)
6416 shader_addline(&buffer
, "PARAM color_key_low = program.env[%u];\n", ARB_FFP_CONST_COLOR_KEY_LOW
);
6417 shader_addline(&buffer
, "PARAM color_key_high = program.env[%u];\n", ARB_FFP_CONST_COLOR_KEY_HIGH
);
6421 /* Find out which textures are read */
6422 for (stage
= 0; stage
< MAX_TEXTURES
; ++stage
)
6424 if (settings
->op
[stage
].cop
== WINED3D_TOP_DISABLE
)
6427 arg0
= settings
->op
[stage
].carg0
& WINED3DTA_SELECTMASK
;
6428 arg1
= settings
->op
[stage
].carg1
& WINED3DTA_SELECTMASK
;
6429 arg2
= settings
->op
[stage
].carg2
& WINED3DTA_SELECTMASK
;
6431 if (arg0
== WINED3DTA_TEXTURE
|| arg1
== WINED3DTA_TEXTURE
|| arg2
== WINED3DTA_TEXTURE
)
6432 tex_read
|= 1u << stage
;
6433 if (settings
->op
[stage
].dst
== tempreg
)
6434 tempreg_used
= TRUE
;
6435 if (arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
)
6436 tempreg_used
= TRUE
;
6437 if (arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
)
6438 tfactor_used
= TRUE
;
6439 if (arg0
== WINED3DTA_CONSTANT
|| arg1
== WINED3DTA_CONSTANT
|| arg2
== WINED3DTA_CONSTANT
)
6440 constant_used
|= 1u << stage
;
6442 switch (settings
->op
[stage
].cop
)
6444 case WINED3D_TOP_BUMPENVMAP_LUMINANCE
:
6445 luminance_used
|= 1u << stage
;
6447 case WINED3D_TOP_BUMPENVMAP
:
6448 bump_used
|= 1u << stage
;
6450 case WINED3D_TOP_BLEND_TEXTURE_ALPHA
:
6451 case WINED3D_TOP_BLEND_TEXTURE_ALPHA_PM
:
6452 tex_read
|= 1u << stage
;
6455 case WINED3D_TOP_BLEND_FACTOR_ALPHA
:
6456 tfactor_used
= TRUE
;
6463 if (settings
->op
[stage
].aop
== WINED3D_TOP_DISABLE
)
6466 arg0
= settings
->op
[stage
].aarg0
& WINED3DTA_SELECTMASK
;
6467 arg1
= settings
->op
[stage
].aarg1
& WINED3DTA_SELECTMASK
;
6468 arg2
= settings
->op
[stage
].aarg2
& WINED3DTA_SELECTMASK
;
6470 if (arg0
== WINED3DTA_TEXTURE
|| arg1
== WINED3DTA_TEXTURE
|| arg2
== WINED3DTA_TEXTURE
)
6471 tex_read
|= 1u << stage
;
6472 if (arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
)
6473 tempreg_used
= TRUE
;
6474 if (arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
)
6475 tfactor_used
= TRUE
;
6476 if (arg0
== WINED3DTA_CONSTANT
|| arg1
== WINED3DTA_CONSTANT
|| arg2
== WINED3DTA_CONSTANT
)
6477 constant_used
|= 1u << stage
;
6479 lowest_disabled_stage
= stage
;
6481 switch (settings
->fog
)
6483 case WINED3D_FFP_PS_FOG_OFF
: break;
6484 case WINED3D_FFP_PS_FOG_LINEAR
:
6485 if (gl_info
->quirks
& WINED3D_QUIRK_BROKEN_ARB_FOG
)
6487 custom_linear_fog
= TRUE
;
6490 shader_addline(&buffer
, "OPTION ARB_fog_linear;\n");
6493 case WINED3D_FFP_PS_FOG_EXP
: shader_addline(&buffer
, "OPTION ARB_fog_exp;\n"); break;
6494 case WINED3D_FFP_PS_FOG_EXP2
: shader_addline(&buffer
, "OPTION ARB_fog_exp2;\n"); break;
6495 default: FIXME("Unexpected fog setting %d\n", settings
->fog
);
6498 shader_addline(&buffer
, "PARAM const = {1, 2, 4, 0.5};\n");
6499 shader_addline(&buffer
, "TEMP TMP;\n");
6500 shader_addline(&buffer
, "TEMP ret;\n");
6501 if (tempreg_used
|| settings
->sRGB_write
) shader_addline(&buffer
, "TEMP tempreg;\n");
6502 shader_addline(&buffer
, "TEMP arg0;\n");
6503 shader_addline(&buffer
, "TEMP arg1;\n");
6504 shader_addline(&buffer
, "TEMP arg2;\n");
6505 for (stage
= 0; stage
< MAX_TEXTURES
; ++stage
)
6507 if (constant_used
& (1u << stage
))
6508 shader_addline(&buffer
, "PARAM const%u = program.env[%u];\n", stage
, ARB_FFP_CONST_CONSTANT(stage
));
6510 if (!(tex_read
& (1u << stage
)))
6513 shader_addline(&buffer
, "TEMP tex%u;\n", stage
);
6515 if (!(bump_used
& (1u << stage
)))
6517 shader_addline(&buffer
, "PARAM bumpmat%u = program.env[%u];\n", stage
, ARB_FFP_CONST_BUMPMAT(stage
));
6519 if (!(luminance_used
& (1u << stage
)))
6521 shader_addline(&buffer
, "PARAM luminance%u = program.env[%u];\n", stage
, ARB_FFP_CONST_LUMINANCE(stage
));
6524 shader_addline(&buffer
, "PARAM tfactor = program.env[%u];\n", ARB_FFP_CONST_TFACTOR
);
6525 shader_addline(&buffer
, "PARAM specular_enable = program.env[%u];\n", ARB_FFP_CONST_SPECULAR_ENABLE
);
6527 if (settings
->sRGB_write
)
6529 shader_addline(&buffer
, "PARAM srgb_consts0 = ");
6530 shader_arb_append_imm_vec4(&buffer
, wined3d_srgb_const0
);
6531 shader_addline(&buffer
, ";\n");
6532 shader_addline(&buffer
, "PARAM srgb_consts1 = ");
6533 shader_arb_append_imm_vec4(&buffer
, wined3d_srgb_const1
);
6534 shader_addline(&buffer
, ";\n");
6537 if (lowest_disabled_stage
< 7 && settings
->emul_clipplanes
)
6538 shader_addline(&buffer
, "KIL fragment.texcoord[7];\n");
6540 if (tempreg_used
|| settings
->sRGB_write
)
6541 shader_addline(&buffer
, "MOV tempreg, 0.0;\n");
6543 /* Generate texture sampling instructions */
6544 for (stage
= 0; stage
< MAX_TEXTURES
&& settings
->op
[stage
].cop
!= WINED3D_TOP_DISABLE
; ++stage
)
6546 if (!(tex_read
& (1u << stage
)))
6549 textype
= arbfp_texture_target(settings
->op
[stage
].tex_type
);
6551 if(settings
->op
[stage
].projected
== proj_none
) {
6553 } else if(settings
->op
[stage
].projected
== proj_count4
||
6554 settings
->op
[stage
].projected
== proj_count3
) {
6557 FIXME("Unexpected projection mode %d\n", settings
->op
[stage
].projected
);
6562 && (settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP
6563 || settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
))
6565 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, x, z, 0, 0;\n", stage
- 1);
6566 shader_addline(&buffer
, "DP3 ret.x, arg1, tex%u;\n", stage
- 1);
6567 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, y, w, 0, 0;\n", stage
- 1);
6568 shader_addline(&buffer
, "DP3 ret.y, arg1, tex%u;\n", stage
- 1);
6570 /* with projective textures, texbem only divides the static texture coord, not the displacement,
6571 * so multiply the displacement with the dividing parameter before passing it to TXP
6573 if (settings
->op
[stage
].projected
!= proj_none
) {
6574 if(settings
->op
[stage
].projected
== proj_count4
) {
6575 shader_addline(&buffer
, "MOV ret.w, fragment.texcoord[%u].w;\n", stage
);
6576 shader_addline(&buffer
, "MUL ret.xyz, ret, fragment.texcoord[%u].w, fragment.texcoord[%u];\n", stage
, stage
);
6578 shader_addline(&buffer
, "MOV ret.w, fragment.texcoord[%u].z;\n", stage
);
6579 shader_addline(&buffer
, "MAD ret.xyz, ret, fragment.texcoord[%u].z, fragment.texcoord[%u];\n", stage
, stage
);
6582 shader_addline(&buffer
, "ADD ret, ret, fragment.texcoord[%u];\n", stage
);
6585 shader_addline(&buffer
, "%s tex%u, ret, texture[%u], %s;\n",
6586 instr
, stage
, stage
, textype
);
6587 if (settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
)
6589 shader_addline(&buffer
, "MAD_SAT ret.x, tex%u.z, luminance%u.x, luminance%u.y;\n",
6590 stage
- 1, stage
- 1, stage
- 1);
6591 shader_addline(&buffer
, "MUL tex%u, tex%u, ret.x;\n", stage
, stage
);
6593 } else if(settings
->op
[stage
].projected
== proj_count3
) {
6594 shader_addline(&buffer
, "MOV ret, fragment.texcoord[%u];\n", stage
);
6595 shader_addline(&buffer
, "MOV ret.w, ret.z;\n");
6596 shader_addline(&buffer
, "%s tex%u, ret, texture[%u], %s;\n",
6597 instr
, stage
, stage
, textype
);
6599 shader_addline(&buffer
, "%s tex%u, fragment.texcoord[%u], texture[%u], %s;\n",
6600 instr
, stage
, stage
, stage
, textype
);
6603 sprintf(colorcor_dst
, "tex%u", stage
);
6604 masks
= calc_color_correction(settings
->op
[stage
].color_fixup
, WINED3DSP_WRITEMASK_ALL
);
6605 gen_color_correction(&buffer
, colorcor_dst
, colorcor_dst
, "const.x", "const.y",
6606 settings
->op
[stage
].color_fixup
, masks
);
6609 if (settings
->color_key_enabled
)
6611 shader_addline(&buffer
, "SLT TMP, tex0, color_key_low;\n"); /* below low key */
6612 shader_addline(&buffer
, "SGE ret, tex0, color_key_high;\n"); /* above high key */
6613 shader_addline(&buffer
, "ADD TMP, TMP, ret;\n"); /* or */
6614 shader_addline(&buffer
, "DP4 TMP.b, TMP, TMP;\n"); /* on any channel */
6615 shader_addline(&buffer
, "SGE TMP, -TMP.b, 0.0;\n"); /* logical not */
6616 shader_addline(&buffer
, "KIL -TMP;\n"); /* discard if true */
6619 shader_addline(&buffer
, "MOV ret, fragment.color.primary;\n");
6621 /* Generate the main shader */
6622 for (stage
= 0; stage
< MAX_TEXTURES
; ++stage
)
6624 if (settings
->op
[stage
].cop
== WINED3D_TOP_DISABLE
)
6627 if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG1
6628 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG1
)
6629 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
;
6630 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG1
6631 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG2
)
6632 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg2
;
6633 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG2
6634 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG1
)
6635 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg1
;
6636 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG2
6637 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG2
)
6638 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
6640 op_equal
= settings
->op
[stage
].aop
== settings
->op
[stage
].cop
6641 && settings
->op
[stage
].carg0
== settings
->op
[stage
].aarg0
6642 && settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
6643 && settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
6645 if (settings
->op
[stage
].aop
== WINED3D_TOP_DISABLE
)
6647 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].dst
,
6648 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6649 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6653 gen_ffp_instr(&buffer
, stage
, TRUE
, TRUE
, settings
->op
[stage
].dst
,
6654 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6655 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6657 else if (settings
->op
[stage
].cop
!= WINED3D_TOP_BUMPENVMAP
6658 && settings
->op
[stage
].cop
!= WINED3D_TOP_BUMPENVMAP_LUMINANCE
)
6660 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].dst
,
6661 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6662 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6663 gen_ffp_instr(&buffer
, stage
, FALSE
, TRUE
, settings
->op
[stage
].dst
,
6664 settings
->op
[stage
].aop
, settings
->op
[stage
].aarg0
,
6665 settings
->op
[stage
].aarg1
, settings
->op
[stage
].aarg2
);
6669 if (settings
->sRGB_write
|| custom_linear_fog
)
6671 shader_addline(&buffer
, "MAD ret, fragment.color.secondary, specular_enable, ret;\n");
6672 if (settings
->sRGB_write
)
6673 arbfp_add_sRGB_correction(&buffer
, "ret", "arg0", "arg1", "arg2", "tempreg", FALSE
);
6674 if (custom_linear_fog
)
6675 arbfp_add_linear_fog(&buffer
, "ret", "arg0");
6676 shader_addline(&buffer
, "MOV result.color, ret;\n");
6680 shader_addline(&buffer
, "MAD result.color, fragment.color.secondary, specular_enable, ret;\n");
6684 shader_addline(&buffer
, "END\n");
6686 /* Generate the shader */
6687 GL_EXTCALL(glGenProgramsARB(1, &ret
));
6688 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, ret
));
6689 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
6691 string_buffer_free(&buffer
);
6695 static void fragment_prog_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6697 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
6698 const struct wined3d_device
*device
= context
->device
;
6699 struct shader_arb_priv
*priv
= device
->fragment_priv
;
6700 BOOL use_pshader
= use_ps(state
);
6701 struct ffp_frag_settings settings
;
6702 const struct arbfp_ffp_desc
*desc
;
6705 TRACE("context %p, state %p, state_id %#x.\n", context
, state
, state_id
);
6707 if (isStateDirty(context
, STATE_RENDER(WINED3D_RS_FOGENABLE
)))
6709 if (!use_pshader
&& device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
)
6711 /* Reload fixed function constants since they collide with the
6712 * pixel shader constants. */
6713 for (i
= 0; i
< MAX_TEXTURES
; ++i
)
6715 set_bumpmat_arbfp(context
, state
, STATE_TEXTURESTAGE(i
, WINED3D_TSS_BUMPENV_MAT00
));
6716 state_tss_constant_arbfp(context
, state
, STATE_TEXTURESTAGE(i
, WINED3D_TSS_CONSTANT
));
6718 state_texfactor_arbfp(context
, state
, STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
));
6719 state_arb_specularenable(context
, state
, STATE_RENDER(WINED3D_RS_SPECULARENABLE
));
6720 color_key_arbfp(context
, state
, STATE_COLOR_KEY
);
6722 else if (use_pshader
)
6724 context
->shader_update_mask
|= 1u << WINED3D_SHADER_TYPE_PIXEL
;
6731 /* Find or create a shader implementing the fixed function pipeline
6732 * settings, then activate it. */
6733 gen_ffp_frag_op(context
, state
, &settings
, FALSE
);
6734 desc
= (const struct arbfp_ffp_desc
*)find_ffp_frag_shader(&priv
->fragment_shaders
, &settings
);
6737 struct arbfp_ffp_desc
*new_desc
= HeapAlloc(GetProcessHeap(), 0, sizeof(*new_desc
));
6740 ERR("Out of memory\n");
6744 new_desc
->parent
.settings
= settings
;
6745 new_desc
->shader
= gen_arbfp_ffp_shader(&settings
, gl_info
);
6746 add_ffp_frag_shader(&priv
->fragment_shaders
, &new_desc
->parent
);
6747 TRACE("Allocated fixed function replacement shader descriptor %p\n", new_desc
);
6751 /* Now activate the replacement program. GL_FRAGMENT_PROGRAM_ARB is already active (however, note the
6752 * comment above the shader_select call below). If e.g. GLSL is active, the shader_select call will
6755 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, desc
->shader
));
6756 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, desc->shader)");
6757 priv
->current_fprogram_id
= desc
->shader
;
6759 if (device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
)
6761 /* Reload fixed function constants since they collide with the
6762 * pixel shader constants. */
6763 for (i
= 0; i
< MAX_TEXTURES
; ++i
)
6765 set_bumpmat_arbfp(context
, state
, STATE_TEXTURESTAGE(i
, WINED3D_TSS_BUMPENV_MAT00
));
6766 state_tss_constant_arbfp(context
, state
, STATE_TEXTURESTAGE(i
, WINED3D_TSS_CONSTANT
));
6768 state_texfactor_arbfp(context
, state
, STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
));
6769 state_arb_specularenable(context
, state
, STATE_RENDER(WINED3D_RS_SPECULARENABLE
));
6770 color_key_arbfp(context
, state
, STATE_COLOR_KEY
);
6772 context
->last_was_pshader
= FALSE
;
6774 else if (!context
->last_was_pshader
)
6776 if (device
->shader_backend
== &arb_program_shader_backend
)
6777 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_F
;
6778 context
->last_was_pshader
= TRUE
;
6781 context
->shader_update_mask
|= 1u << WINED3D_SHADER_TYPE_PIXEL
;
6784 /* We can't link the fog states to the fragment state directly since the
6785 * vertex pipeline links them to FOGENABLE. A different linking in different
6786 * pipeline parts can't be expressed in the combined state table, so we need
6787 * to handle that with a forwarding function. The other invisible side effect
6788 * is that changing the fog start and fog end (which links to FOGENABLE in
6789 * vertex) results in the fragment_prog_arbfp function being called because
6790 * FOGENABLE is dirty, which calls this function here. */
6791 static void state_arbfp_fog(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6793 enum fogsource new_source
;
6794 DWORD fogstart
= state
->render_states
[WINED3D_RS_FOGSTART
];
6795 DWORD fogend
= state
->render_states
[WINED3D_RS_FOGEND
];
6797 TRACE("context %p, state %p, state_id %#x.\n", context
, state
, state_id
);
6799 if (!isStateDirty(context
, STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
)))
6800 fragment_prog_arbfp(context
, state
, state_id
);
6802 if (!state
->render_states
[WINED3D_RS_FOGENABLE
])
6805 if (state
->render_states
[WINED3D_RS_FOGTABLEMODE
] == WINED3D_FOG_NONE
)
6809 new_source
= FOGSOURCE_VS
;
6813 if (state
->render_states
[WINED3D_RS_FOGVERTEXMODE
] == WINED3D_FOG_NONE
|| context
->last_was_rhw
)
6814 new_source
= FOGSOURCE_COORD
;
6816 new_source
= FOGSOURCE_FFP
;
6821 new_source
= FOGSOURCE_FFP
;
6824 if (new_source
!= context
->fog_source
|| fogstart
== fogend
)
6826 context
->fog_source
= new_source
;
6827 state_fogstartend(context
, state
, STATE_RENDER(WINED3D_RS_FOGSTART
));
6831 static void textransform(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6833 if (!isStateDirty(context
, STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
)))
6834 fragment_prog_arbfp(context
, state
, state_id
);
6837 static const struct StateEntryTemplate arbfp_fragmentstate_template
[] =
6839 {STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
), { STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
), state_texfactor_arbfp
}, WINED3D_GL_EXT_NONE
},
6840 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6841 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6842 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6843 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6844 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6845 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6846 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6847 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6848 {STATE_TEXTURESTAGE(0, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6849 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6850 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6851 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6852 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6853 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6854 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6855 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6856 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6857 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6858 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6859 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6860 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6861 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6862 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6863 {STATE_TEXTURESTAGE(1, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6864 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6865 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6866 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6867 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6868 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6869 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6870 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6871 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6872 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6873 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6874 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6875 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6876 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6877 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6878 {STATE_TEXTURESTAGE(2, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6879 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6880 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6881 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6882 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6883 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6884 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6885 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6886 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6887 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6888 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6889 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6890 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6891 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6892 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6893 {STATE_TEXTURESTAGE(3, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6894 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6895 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6896 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6897 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6898 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6899 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6900 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6901 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6902 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6903 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6904 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6905 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6906 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6907 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6908 {STATE_TEXTURESTAGE(4, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6909 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6910 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6911 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6912 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6913 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6914 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6915 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6916 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6917 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6918 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6919 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6920 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6921 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6922 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6923 {STATE_TEXTURESTAGE(5, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6924 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6925 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6926 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6927 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6928 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6929 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6930 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6931 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6932 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6933 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6934 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6935 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6936 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6937 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6938 {STATE_TEXTURESTAGE(6, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6939 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6940 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6941 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6942 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6943 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6944 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6945 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6946 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6947 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6948 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6949 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6950 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6951 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6952 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6953 {STATE_TEXTURESTAGE(7, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6954 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6955 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6956 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6957 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6958 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6959 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6960 {STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
6961 {STATE_RENDER(WINED3D_RS_ALPHAFUNC
), { STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6962 {STATE_RENDER(WINED3D_RS_ALPHAREF
), { STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6963 {STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), { STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), alpha_test_arbfp
}, WINED3D_GL_EXT_NONE
},
6964 {STATE_RENDER(WINED3D_RS_COLORKEYENABLE
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6965 {STATE_COLOR_KEY
, { STATE_COLOR_KEY
, color_key_arbfp
}, WINED3D_GL_EXT_NONE
},
6966 {STATE_RENDER(WINED3D_RS_FOGENABLE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), state_arbfp_fog
}, WINED3D_GL_EXT_NONE
},
6967 {STATE_RENDER(WINED3D_RS_FOGTABLEMODE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6968 {STATE_RENDER(WINED3D_RS_FOGVERTEXMODE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6969 {STATE_RENDER(WINED3D_RS_FOGSTART
), { STATE_RENDER(WINED3D_RS_FOGSTART
), state_fogstartend
}, WINED3D_GL_EXT_NONE
},
6970 {STATE_RENDER(WINED3D_RS_FOGEND
), { STATE_RENDER(WINED3D_RS_FOGSTART
), NULL
}, WINED3D_GL_EXT_NONE
},
6971 {STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), { STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), state_srgbwrite
}, ARB_FRAMEBUFFER_SRGB
},
6972 {STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6973 {STATE_RENDER(WINED3D_RS_FOGCOLOR
), { STATE_RENDER(WINED3D_RS_FOGCOLOR
), state_fogcolor
}, WINED3D_GL_EXT_NONE
},
6974 {STATE_RENDER(WINED3D_RS_FOGDENSITY
), { STATE_RENDER(WINED3D_RS_FOGDENSITY
), state_fogdensity
}, WINED3D_GL_EXT_NONE
},
6975 {STATE_TEXTURESTAGE(0,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(0,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6976 {STATE_TEXTURESTAGE(1,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(1,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6977 {STATE_TEXTURESTAGE(2,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(2,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6978 {STATE_TEXTURESTAGE(3,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(3,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6979 {STATE_TEXTURESTAGE(4,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(4,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6980 {STATE_TEXTURESTAGE(5,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(5,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6981 {STATE_TEXTURESTAGE(6,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(6,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6982 {STATE_TEXTURESTAGE(7,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(7,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6983 {STATE_TEXTURESTAGE(0, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6984 {STATE_TEXTURESTAGE(1, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6985 {STATE_TEXTURESTAGE(2, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6986 {STATE_TEXTURESTAGE(3, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6987 {STATE_TEXTURESTAGE(4, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6988 {STATE_TEXTURESTAGE(5, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6989 {STATE_TEXTURESTAGE(6, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6990 {STATE_TEXTURESTAGE(7, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6991 {STATE_RENDER(WINED3D_RS_SPECULARENABLE
), { STATE_RENDER(WINED3D_RS_SPECULARENABLE
), state_arb_specularenable
}, WINED3D_GL_EXT_NONE
},
6992 {STATE_RENDER(WINED3D_RS_SHADEMODE
), { STATE_RENDER(WINED3D_RS_SHADEMODE
), state_shademode
}, WINED3D_GL_EXT_NONE
},
6993 {0 /* Terminate */, { 0, 0 }, WINED3D_GL_EXT_NONE
},
6996 static BOOL
arbfp_alloc_context_data(struct wined3d_context
*context
)
7001 static void arbfp_free_context_data(struct wined3d_context
*context
)
7005 const struct fragment_pipeline arbfp_fragment_pipeline
= {
7008 arbfp_get_emul_mask
,
7011 arbfp_alloc_context_data
,
7012 arbfp_free_context_data
,
7013 shader_arb_color_fixup_supported
,
7014 arbfp_fragmentstate_template
,
7017 struct arbfp_blit_type
7019 enum complex_fixup fixup
: 4;
7020 enum wined3d_gl_resource_type res_type
: 3;
7021 DWORD use_color_key
: 1;
7025 struct arbfp_blit_desc
7028 struct arbfp_blit_type type
;
7029 struct wine_rb_entry entry
;
7032 #define ARBFP_BLIT_PARAM_SIZE 0
7033 #define ARBFP_BLIT_PARAM_COLOR_KEY_LOW 1
7034 #define ARBFP_BLIT_PARAM_COLOR_KEY_HIGH 2
7036 struct arbfp_blit_priv
7038 struct wine_rb_tree shaders
;
7039 GLuint palette_texture
;
7042 static int arbfp_blit_type_compare(const void *key
, const struct wine_rb_entry
*entry
)
7044 const struct arbfp_blit_type
*ka
= key
;
7045 const struct arbfp_blit_type
*kb
= &WINE_RB_ENTRY_VALUE(entry
, const struct arbfp_blit_desc
, entry
)->type
;
7047 return memcmp(ka
, kb
, sizeof(*ka
));
7050 /* Context activation is done by the caller. */
7051 static void arbfp_free_blit_shader(struct wine_rb_entry
*entry
, void *context
)
7053 const struct wined3d_gl_info
*gl_info
= context
;
7054 struct arbfp_blit_desc
*entry_arb
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_blit_desc
, entry
);
7056 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb
->shader
));
7057 checkGLcall("glDeleteProgramsARB(1, &entry_arb->shader)");
7058 HeapFree(GetProcessHeap(), 0, entry_arb
);
7061 static HRESULT
arbfp_blit_alloc(struct wined3d_device
*device
)
7063 struct arbfp_blit_priv
*priv
;
7065 if (!(priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*priv
))))
7066 return E_OUTOFMEMORY
;
7068 wine_rb_init(&priv
->shaders
, arbfp_blit_type_compare
);
7070 device
->blit_priv
= priv
;
7075 /* Context activation is done by the caller. */
7076 static void arbfp_blit_free(struct wined3d_device
*device
)
7078 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
7079 struct arbfp_blit_priv
*priv
= device
->blit_priv
;
7081 wine_rb_destroy(&priv
->shaders
, arbfp_free_blit_shader
, &device
->adapter
->gl_info
);
7082 checkGLcall("Delete blit programs");
7084 if (priv
->palette_texture
)
7085 gl_info
->gl_ops
.gl
.p_glDeleteTextures(1, &priv
->palette_texture
);
7087 HeapFree(GetProcessHeap(), 0, device
->blit_priv
);
7088 device
->blit_priv
= NULL
;
7091 static BOOL
gen_planar_yuv_read(struct wined3d_string_buffer
*buffer
, const struct arbfp_blit_type
*type
,
7095 const char *tex
, *texinstr
= "TXP";
7097 if (type
->fixup
== COMPLEX_FIXUP_UYVY
)
7108 tex
= arbfp_texture_target(type
->res_type
);
7109 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_RECT
)
7112 /* First we have to read the chroma values. This means we need at least two pixels(no filtering),
7113 * or 4 pixels(with filtering). To get the unmodified chromas, we have to rid ourselves of the
7114 * filtering when we sample the texture.
7116 * These are the rules for reading the chroma:
7122 * So we have to get the sampling x position in non-normalized coordinates in integers
7124 if (type
->res_type
!= WINED3D_GL_RES_TYPE_TEX_RECT
)
7126 shader_addline(buffer
, "MUL texcrd.xy, fragment.texcoord[0], size.x;\n");
7127 shader_addline(buffer
, "MOV texcrd.w, size.x;\n");
7131 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7133 /* We must not allow filtering between pixel x and x+1, this would mix U and V
7134 * Vertical filtering is ok. However, bear in mind that the pixel center is at
7137 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
7138 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, coef.y;\n");
7140 /* Divide the x coordinate by 0.5 and get the fraction. This gives 0.25 and 0.75 for the
7141 * even and odd pixels respectively
7143 shader_addline(buffer
, "MUL texcrd2, texcrd, coef.y;\n");
7144 shader_addline(buffer
, "FRC texcrd2, texcrd2;\n");
7146 /* Sample Pixel 1 */
7147 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
7149 /* Put the value into either of the chroma values */
7150 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
7151 shader_addline(buffer
, "MUL chroma.x, luminance.%c, temp.x;\n", chroma
);
7152 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
7153 shader_addline(buffer
, "MUL chroma.y, luminance.%c, temp.x;\n", chroma
);
7155 /* Sample pixel 2. If we read an even pixel(SLT above returned 1), sample
7156 * the pixel right to the current one. Otherwise, sample the left pixel.
7157 * Bias and scale the SLT result to -1;1 and add it to the texcrd.x.
7159 shader_addline(buffer
, "MAD temp.x, temp.x, coef.z, -coef.x;\n");
7160 shader_addline(buffer
, "ADD texcrd.x, texcrd, temp.x;\n");
7161 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
7163 /* Put the value into the other chroma */
7164 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
7165 shader_addline(buffer
, "MAD chroma.y, luminance.%c, temp.x, chroma.y;\n", chroma
);
7166 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
7167 shader_addline(buffer
, "MAD chroma.x, luminance.%c, temp.x, chroma.x;\n", chroma
);
7169 /* TODO: If filtering is enabled, sample a 2nd pair of pixels left or right of
7170 * the current one and lerp the two U and V values
7173 /* This gives the correctly filtered luminance value */
7174 shader_addline(buffer
, "TEX luminance, fragment.texcoord[0], texture[0], %s;\n", tex
);
7179 static BOOL
gen_yv12_read(struct wined3d_string_buffer
*buffer
, const struct arbfp_blit_type
*type
,
7183 static const float yv12_coef
[]
7184 = {2.0f
/ 3.0f
, 1.0f
/ 6.0f
, (2.0f
/ 3.0f
) + (1.0f
/ 6.0f
), 1.0f
/ 3.0f
};
7186 tex
= arbfp_texture_target(type
->res_type
);
7188 /* YV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
7189 * V and a (W/2)x(H/2) U plane, each with 8 bit per pixel. So the effective
7190 * bitdepth is 12 bits per pixel. Since the U and V planes have only half the
7191 * pitch of the luminance plane, the packing into the gl texture is a bit
7192 * unfortunate. If the whole texture is interpreted as luminance data it looks
7193 * approximately like this:
7195 * +----------------------------------+----
7207 * +----------------+-----------------+----
7209 * | V even rows | V odd rows |
7211 * +----------------+------------------ -
7213 * | U even rows | U odd rows |
7215 * +----------------+-----------------+----
7219 * So it appears as if there are 4 chroma images, but in fact the odd rows
7220 * in the chroma images are in the same row as the even ones. So it is
7221 * kinda tricky to read
7223 * When reading from rectangle textures, keep in mind that the input y coordinates
7224 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height
7226 shader_addline(buffer
, "PARAM yv12_coef = ");
7227 shader_arb_append_imm_vec4(buffer
, yv12_coef
);
7228 shader_addline(buffer
, ";\n");
7230 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7231 /* the chroma planes have only half the width */
7232 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, coef.y;\n");
7234 /* The first value is between 2/3 and 5/6th of the texture's height, so scale+bias
7235 * the coordinate. Also read the right side of the image when reading odd lines
7237 * Don't forget to clamp the y values in into the range, otherwise we'll get filtering
7240 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7243 shader_addline(buffer
, "RCP chroma.w, size.y;\n");
7245 shader_addline(buffer
, "MUL texcrd2.y, texcrd.y, size.y;\n");
7247 shader_addline(buffer
, "FLR texcrd2.y, texcrd2.y;\n");
7248 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, yv12_coef.y, yv12_coef.x;\n");
7250 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
7251 shader_addline(buffer
, "ADD texcrd2.x, texcrd2.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
7252 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
7253 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
7254 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
7256 /* clamp, keep the half pixel origin in mind */
7257 shader_addline(buffer
, "MAD temp.y, coef.y, chroma.w, yv12_coef.x;\n");
7258 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7259 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.z;\n");
7260 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7264 /* Read from [size - size+size/4] */
7265 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
7266 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, coef.w, size.y;\n");
7268 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
7269 shader_addline(buffer
, "ADD texcrd2.x, texcrd.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
7270 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
7271 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
7272 shader_addline(buffer
, "MUL texcrd2.x, texcrd2.x, size.x;\n");
7273 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
7275 /* Make sure to read exactly from the pixel center */
7276 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
7277 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, coef.y;\n");
7280 shader_addline(buffer
, "MAD temp.y, size.y, coef.w, size.y;\n");
7281 shader_addline(buffer
, "ADD temp.y, temp.y, -coef.y;\n");
7282 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7283 shader_addline(buffer
, "ADD temp.y, size.y, coef.y;\n");
7284 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7286 /* Read the texture, put the result into the output register */
7287 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7288 shader_addline(buffer
, "MOV chroma.x, temp.w;\n");
7290 /* The other chroma value is 1/6th of the texture lower, from 5/6th to 6/6th
7291 * No need to clamp because we're just reusing the already clamped value from above
7293 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7294 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, yv12_coef.y;\n");
7296 shader_addline(buffer
, "MAD texcrd.y, size.y, coef.w, texcrd.y;\n");
7297 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7298 shader_addline(buffer
, "MOV chroma.y, temp.w;\n");
7300 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
7301 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
7302 * values due to filtering
7304 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7305 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7307 /* Multiply the y coordinate by 2/3 and clamp it */
7308 shader_addline(buffer
, "MUL texcrd.y, texcrd.y, yv12_coef.x;\n");
7309 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.x;\n");
7310 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7311 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7315 /* Reading from texture_rectangles is pretty straightforward, just use the unmodified
7316 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
7319 shader_addline(buffer
, "ADD temp.x, size.y, -coef.y;\n");
7320 shader_addline(buffer
, "MIN texcrd.y, texcrd.y, size.x;\n");
7321 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7328 static BOOL
gen_nv12_read(struct wined3d_string_buffer
*buffer
, const struct arbfp_blit_type
*type
,
7332 static const float nv12_coef
[]
7333 = {2.0f
/ 3.0f
, 1.0f
/ 3.0f
, 1.0f
, 1.0f
};
7335 tex
= arbfp_texture_target(type
->res_type
);
7337 /* NV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
7338 * sized plane where each component is an UV pair. So the effective
7339 * bitdepth is 12 bits per pixel If the whole texture is interpreted as luminance
7340 * data it looks approximately like this:
7342 * +----------------------------------+----
7354 * +----------------------------------+----
7355 * |UVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUV|
7356 * |UVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUV|
7362 * +----------------------------------+----
7364 * When reading from rectangle textures, keep in mind that the input y coordinates
7365 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height. */
7367 shader_addline(buffer
, "PARAM nv12_coef = ");
7368 shader_arb_append_imm_vec4(buffer
, nv12_coef
);
7369 shader_addline(buffer
, ";\n");
7371 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7372 /* We only have half the number of chroma pixels. */
7373 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, coef.y;\n");
7375 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7377 shader_addline(buffer
, "RCP chroma.w, size.x;\n");
7378 shader_addline(buffer
, "RCP chroma.z, size.y;\n");
7380 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, nv12_coef.y, nv12_coef.x;\n");
7382 /* We must not allow filtering horizontally, this would mix U and V.
7383 * Vertical filtering is ok. However, bear in mind that the pixel center is at
7384 * 0.5, so add 0.5. */
7386 /* Convert to non-normalized coordinates so we can find the
7387 * individual pixel. */
7388 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, size.x;\n");
7389 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
7390 /* Multiply by 2 since chroma components are stored in UV pixel pairs,
7391 * add 0.5 to hit the center of the pixel. */
7392 shader_addline(buffer
, "MAD texcrd.x, texcrd.x, coef.z, coef.y;\n");
7394 /* Convert back to normalized coordinates. */
7395 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, chroma.w;\n");
7397 /* Clamp, keep the half pixel origin in mind. */
7398 shader_addline(buffer
, "MAD temp.y, coef.y, chroma.z, nv12_coef.x;\n");
7399 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7400 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.z, nv12_coef.z;\n");
7401 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7405 /* Read from [size - size+size/2] */
7406 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, coef.y, size.y;\n");
7408 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
7409 /* Multiply by 2 since chroma components are stored in UV pixel pairs,
7410 * add 0.5 to hit the center of the pixel. */
7411 shader_addline(buffer
, "MAD texcrd.x, texcrd.x, coef.z, coef.y;\n");
7414 shader_addline(buffer
, "MAD temp.y, size.y, coef.y, size.y;\n");
7415 shader_addline(buffer
, "ADD temp.y, temp.y, -coef.y;\n");
7416 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7417 shader_addline(buffer
, "ADD temp.y, size.y, coef.y;\n");
7418 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7420 /* Read the texture, put the result into the output register. */
7421 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7422 shader_addline(buffer
, "MOV chroma.y, temp.w;\n");
7424 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7427 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, chroma.w;\n");
7432 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, coef.x;\n");
7434 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7435 shader_addline(buffer
, "MOV chroma.x, temp.w;\n");
7437 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
7438 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
7439 * values due to filtering. */
7440 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7441 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7443 /* Multiply the y coordinate by 2/3 and clamp it */
7444 shader_addline(buffer
, "MUL texcrd.y, texcrd.y, nv12_coef.x;\n");
7445 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, nv12_coef.x;\n");
7446 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7447 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7451 /* Reading from texture_rectangles is pretty straightforward, just use the unmodified
7452 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
7455 shader_addline(buffer
, "ADD temp.x, size.y, -coef.y;\n");
7456 shader_addline(buffer
, "MIN texcrd.y, texcrd.y, size.x;\n");
7457 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7464 /* Context activation is done by the caller. */
7465 static GLuint
gen_p8_shader(struct arbfp_blit_priv
*priv
,
7466 const struct wined3d_gl_info
*gl_info
, const struct arbfp_blit_type
*type
)
7469 struct wined3d_string_buffer buffer
;
7470 const char *tex_target
= arbfp_texture_target(type
->res_type
);
7472 /* This should not happen because we only use this conversion for
7473 * present blits which don't use color keying. */
7474 if (type
->use_color_key
)
7475 FIXME("Implement P8 color keying.\n");
7478 if (!string_buffer_init(&buffer
))
7480 ERR("Failed to initialize shader buffer.\n");
7484 GL_EXTCALL(glGenProgramsARB(1, &shader
));
7485 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7488 string_buffer_free(&buffer
);
7492 shader_addline(&buffer
, "!!ARBfp1.0\n");
7493 shader_addline(&buffer
, "TEMP index;\n");
7495 /* { 255/256, 0.5/255*255/256, 0, 0 } */
7496 shader_addline(&buffer
, "PARAM constants = { 0.996, 0.00195, 0, 0 };\n");
7498 /* The alpha-component contains the palette index */
7499 shader_addline(&buffer
, "TEX index, fragment.texcoord[0], texture[0], %s;\n", tex_target
);
7501 /* Scale the index by 255/256 and add a bias of '0.5' in order to sample in the middle */
7502 shader_addline(&buffer
, "MAD index.a, index.a, constants.x, constants.y;\n");
7504 /* Use the alpha-component as an index in the palette to get the final color */
7505 shader_addline(&buffer
, "TEX result.color, index.a, texture[1], 1D;\n");
7506 shader_addline(&buffer
, "END\n");
7508 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
7510 string_buffer_free(&buffer
);
7515 /* Context activation is done by the caller. */
7516 static void upload_palette(const struct wined3d_texture
*texture
, struct wined3d_context
*context
)
7518 const struct wined3d_palette
*palette
= texture
->swapchain
? texture
->swapchain
->palette
: NULL
;
7519 struct wined3d_device
*device
= texture
->resource
.device
;
7520 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
7521 struct arbfp_blit_priv
*priv
= device
->blit_priv
;
7523 if (!priv
->palette_texture
)
7524 gl_info
->gl_ops
.gl
.p_glGenTextures(1, &priv
->palette_texture
);
7526 GL_EXTCALL(glActiveTexture(GL_TEXTURE1
));
7527 gl_info
->gl_ops
.gl
.p_glBindTexture(GL_TEXTURE_1D
, priv
->palette_texture
);
7529 gl_info
->gl_ops
.gl
.p_glTexEnvi(GL_TEXTURE_ENV
, GL_TEXTURE_ENV_MODE
, GL_REPLACE
);
7531 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_MIN_FILTER
, GL_NEAREST
);
7532 /* Make sure we have discrete color levels. */
7533 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_MAG_FILTER
, GL_NEAREST
);
7534 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_WRAP_S
, GL_CLAMP_TO_EDGE
);
7535 /* TODO: avoid unneeded uploads in the future by adding some SFLAG_PALETTE_DIRTY mechanism */
7538 gl_info
->gl_ops
.gl
.p_glTexImage1D(GL_TEXTURE_1D
, 0, GL_RGB
, 256, 0, GL_BGRA
,
7539 GL_UNSIGNED_INT_8_8_8_8_REV
, palette
->colors
);
7543 static const DWORD black
;
7544 FIXME("P8 surface loaded without a palette.\n");
7545 gl_info
->gl_ops
.gl
.p_glTexImage1D(GL_TEXTURE_1D
, 0, GL_RGB
, 1, 0, GL_BGRA
,
7546 GL_UNSIGNED_INT_8_8_8_8_REV
, &black
);
7549 /* Switch back to unit 0 in which the 2D texture will be stored. */
7550 context_active_texture(context
, gl_info
, 0);
7553 /* Context activation is done by the caller. */
7554 static GLuint
gen_yuv_shader(struct arbfp_blit_priv
*priv
, const struct wined3d_gl_info
*gl_info
,
7555 const struct arbfp_blit_type
*type
)
7558 struct wined3d_string_buffer buffer
;
7559 char luminance_component
;
7561 if (type
->use_color_key
)
7562 FIXME("Implement YUV color keying.\n");
7565 if (!string_buffer_init(&buffer
))
7567 ERR("Failed to initialize shader buffer.\n");
7571 GL_EXTCALL(glGenProgramsARB(1, &shader
));
7572 checkGLcall("GL_EXTCALL(glGenProgramsARB(1, &shader))");
7573 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7574 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
7577 string_buffer_free(&buffer
);
7581 /* The YUY2 and UYVY formats contain two pixels packed into a 32 bit macropixel,
7582 * giving effectively 16 bit per pixel. The color consists of a luminance(Y) and
7583 * two chroma(U and V) values. Each macropixel has two luminance values, one for
7584 * each single pixel it contains, and one U and one V value shared between both
7587 * The data is loaded into an A8L8 texture. With YUY2, the luminance component
7588 * contains the luminance and alpha the chroma. With UYVY it is vice versa. Thus
7589 * take the format into account when generating the read swizzles
7591 * Reading the Y value is straightforward - just sample the texture. The hardware
7592 * takes care of filtering in the horizontal and vertical direction.
7594 * Reading the U and V values is harder. We have to avoid filtering horizontally,
7595 * because that would mix the U and V values of one pixel or two adjacent pixels.
7596 * Thus floor the texture coordinate and add 0.5 to get an unfiltered read,
7597 * regardless of the filtering setting. Vertical filtering works automatically
7598 * though - the U and V values of two rows are mixed nicely.
7600 * Apart of avoiding filtering issues, the code has to know which value it just
7601 * read, and where it can find the other one. To determine this, it checks if
7602 * it sampled an even or odd pixel, and shifts the 2nd read accordingly.
7604 * Handling horizontal filtering of U and V values requires reading a 2nd pair
7605 * of pixels, extracting U and V and mixing them. This is not implemented yet.
7607 * An alternative implementation idea is to load the texture as A8R8G8B8 texture,
7608 * with width / 2. This way one read gives all 3 values, finding U and V is easy
7609 * in an unfiltered situation. Finding the luminance on the other hand requires
7610 * finding out if it is an odd or even pixel. The real drawback of this approach
7611 * is filtering. This would have to be emulated completely in the shader, reading
7612 * up two 2 packed pixels in up to 2 rows and interpolating both horizontally and
7613 * vertically. Beyond that it would require adjustments to the texture handling
7614 * code to deal with the width scaling
7616 shader_addline(&buffer
, "!!ARBfp1.0\n");
7617 shader_addline(&buffer
, "TEMP luminance;\n");
7618 shader_addline(&buffer
, "TEMP temp;\n");
7619 shader_addline(&buffer
, "TEMP chroma;\n");
7620 shader_addline(&buffer
, "TEMP texcrd;\n");
7621 shader_addline(&buffer
, "TEMP texcrd2;\n");
7622 shader_addline(&buffer
, "PARAM coef = {1.0, 0.5, 2.0, 0.25};\n");
7623 shader_addline(&buffer
, "PARAM yuv_coef = {1.403, 0.344, 0.714, 1.770};\n");
7624 shader_addline(&buffer
, "PARAM size = program.local[%u];\n", ARBFP_BLIT_PARAM_SIZE
);
7626 switch (type
->fixup
)
7628 case COMPLEX_FIXUP_UYVY
:
7629 case COMPLEX_FIXUP_YUY2
:
7630 if (!gen_planar_yuv_read(&buffer
, type
, &luminance_component
))
7632 string_buffer_free(&buffer
);
7637 case COMPLEX_FIXUP_YV12
:
7638 if (!gen_yv12_read(&buffer
, type
, &luminance_component
))
7640 string_buffer_free(&buffer
);
7645 case COMPLEX_FIXUP_NV12
:
7646 if (!gen_nv12_read(&buffer
, type
, &luminance_component
))
7648 string_buffer_free(&buffer
);
7654 FIXME("Unsupported YUV fixup %#x\n", type
->fixup
);
7655 string_buffer_free(&buffer
);
7659 /* Calculate the final result. Formula is taken from
7660 * http://www.fourcc.org/fccyvrgb.php. Note that the chroma
7661 * ranges from -0.5 to 0.5
7663 shader_addline(&buffer
, "SUB chroma.xy, chroma, coef.y;\n");
7665 shader_addline(&buffer
, "MAD result.color.x, chroma.x, yuv_coef.x, luminance.%c;\n", luminance_component
);
7666 shader_addline(&buffer
, "MAD temp.x, -chroma.y, yuv_coef.y, luminance.%c;\n", luminance_component
);
7667 shader_addline(&buffer
, "MAD result.color.y, -chroma.x, yuv_coef.z, temp.x;\n");
7668 shader_addline(&buffer
, "MAD result.color.z, chroma.y, yuv_coef.w, luminance.%c;\n", luminance_component
);
7669 shader_addline(&buffer
, "END\n");
7671 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
7673 string_buffer_free(&buffer
);
7678 /* Context activation is done by the caller. */
7679 static GLuint
arbfp_gen_plain_shader(struct arbfp_blit_priv
*priv
,
7680 const struct wined3d_gl_info
*gl_info
, const struct arbfp_blit_type
*type
)
7683 struct wined3d_string_buffer buffer
;
7684 const char *tex_target
= arbfp_texture_target(type
->res_type
);
7687 if (!string_buffer_init(&buffer
))
7689 ERR("Failed to initialize shader buffer.\n");
7693 GL_EXTCALL(glGenProgramsARB(1, &shader
));
7696 string_buffer_free(&buffer
);
7699 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7701 shader_addline(&buffer
, "!!ARBfp1.0\n");
7703 if (type
->use_color_key
)
7705 shader_addline(&buffer
, "TEMP color;\n");
7706 shader_addline(&buffer
, "TEMP less, greater;\n");
7707 shader_addline(&buffer
, "PARAM color_key_low = program.local[%u];\n", ARBFP_BLIT_PARAM_COLOR_KEY_LOW
);
7708 shader_addline(&buffer
, "PARAM color_key_high = program.local[%u];\n", ARBFP_BLIT_PARAM_COLOR_KEY_HIGH
);
7709 shader_addline(&buffer
, "TEX color, fragment.texcoord[0], texture[0], %s;\n", tex_target
);
7710 shader_addline(&buffer
, "SLT less, color, color_key_low;\n"); /* below low key */
7711 shader_addline(&buffer
, "SGE greater, color, color_key_high;\n"); /* above high key */
7712 shader_addline(&buffer
, "ADD less, less, greater;\n"); /* or */
7713 shader_addline(&buffer
, "DP4 less.b, less, less;\n"); /* on any channel */
7714 shader_addline(&buffer
, "SGE less, -less.b, 0.0;\n"); /* logical not */
7715 shader_addline(&buffer
, "KIL -less;\n"); /* discard if true */
7716 shader_addline(&buffer
, "MOV result.color, color;\n");
7720 shader_addline(&buffer
, "TEX result.color, fragment.texcoord[0], texture[0], %s;\n", tex_target
);
7723 shader_addline(&buffer
, "END\n");
7725 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
7727 string_buffer_free(&buffer
);
7732 /* Context activation is done by the caller. */
7733 static HRESULT
arbfp_blit_set(void *blit_priv
, struct wined3d_context
*context
, const struct wined3d_surface
*surface
,
7734 const struct wined3d_color_key
*color_key
)
7736 const struct wined3d_texture
*texture
= surface
->container
;
7737 struct arbfp_blit_priv
*priv
= blit_priv
;
7738 enum complex_fixup fixup
;
7739 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
7740 struct wine_rb_entry
*entry
;
7741 struct arbfp_blit_type type
;
7742 struct arbfp_blit_desc
*desc
;
7743 struct wined3d_color float_color_key
[2];
7744 struct wined3d_vec4 size
;
7747 size
.x
= wined3d_texture_get_level_pow2_width(texture
, surface
->texture_level
);
7748 size
.y
= wined3d_texture_get_level_pow2_height(texture
, surface
->texture_level
);
7752 if (is_complex_fixup(texture
->resource
.format
->color_fixup
))
7753 fixup
= get_complex_fixup(texture
->resource
.format
->color_fixup
);
7755 fixup
= COMPLEX_FIXUP_NONE
;
7757 switch (texture
->target
)
7760 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_1D
;
7764 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_2D
;
7768 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_3D
;
7771 case GL_TEXTURE_CUBE_MAP_ARB
:
7772 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_CUBE
;
7775 case GL_TEXTURE_RECTANGLE_ARB
:
7776 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_RECT
;
7780 ERR("Unexpected GL texture type %#x.\n", texture
->target
);
7781 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_2D
;
7784 type
.use_color_key
= !!color_key
;
7787 entry
= wine_rb_get(&priv
->shaders
, &type
);
7790 desc
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_blit_desc
, entry
);
7791 shader
= desc
->shader
;
7797 case COMPLEX_FIXUP_NONE
:
7798 if (!is_identity_fixup(texture
->resource
.format
->color_fixup
))
7799 FIXME("Implement support for sign or swizzle fixups.\n");
7800 shader
= arbfp_gen_plain_shader(priv
, gl_info
, &type
);
7803 case COMPLEX_FIXUP_P8
:
7804 shader
= gen_p8_shader(priv
, gl_info
, &type
);
7807 case COMPLEX_FIXUP_YUY2
:
7808 case COMPLEX_FIXUP_UYVY
:
7809 case COMPLEX_FIXUP_YV12
:
7810 case COMPLEX_FIXUP_NV12
:
7811 shader
= gen_yuv_shader(priv
, gl_info
, &type
);
7817 FIXME("Unsupported complex fixup %#x, not setting a shader\n", fixup
);
7821 desc
= HeapAlloc(GetProcessHeap(), 0, sizeof(*desc
));
7826 desc
->shader
= shader
;
7827 if (wine_rb_put(&priv
->shaders
, &desc
->type
, &desc
->entry
) == -1)
7830 ERR("Out of memory\n");
7831 GL_EXTCALL(glDeleteProgramsARB(1, &shader
));
7832 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader))");
7833 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, 0));
7834 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, 0)");
7835 HeapFree(GetProcessHeap(), 0, desc
);
7836 return E_OUTOFMEMORY
;
7840 if (fixup
== COMPLEX_FIXUP_P8
)
7841 upload_palette(texture
, context
);
7843 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
7844 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
7845 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7846 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
7847 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARBFP_BLIT_PARAM_SIZE
, &size
.x
));
7848 checkGLcall("glProgramLocalParameter4fvARB");
7849 if (type
.use_color_key
)
7851 wined3d_format_get_float_color_key(texture
->resource
.format
, color_key
, float_color_key
);
7852 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
7853 ARBFP_BLIT_PARAM_COLOR_KEY_LOW
, &float_color_key
[0].r
));
7854 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
7855 ARBFP_BLIT_PARAM_COLOR_KEY_HIGH
, &float_color_key
[1].r
));
7856 checkGLcall("glProgramLocalParameter4fvARB");
7862 /* Context activation is done by the caller. */
7863 static void arbfp_blit_unset(const struct wined3d_gl_info
*gl_info
)
7865 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
7866 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
7869 static BOOL
arbfp_blit_supported(const struct wined3d_gl_info
*gl_info
,
7870 const struct wined3d_d3d_info
*d3d_info
, enum wined3d_blit_op blit_op
,
7871 const RECT
*src_rect
, DWORD src_usage
, enum wined3d_pool src_pool
, const struct wined3d_format
*src_format
,
7872 const RECT
*dst_rect
, DWORD dst_usage
, enum wined3d_pool dst_pool
, const struct wined3d_format
*dst_format
)
7874 enum complex_fixup src_fixup
;
7876 if (!gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
7881 case WINED3D_BLIT_OP_COLOR_BLIT_CKEY
:
7882 if (!d3d_info
->shader_color_key
)
7884 /* The conversion modifies the alpha channel so the color key might no longer match. */
7885 TRACE("Color keying not supported with converted textures.\n");
7888 case WINED3D_BLIT_OP_COLOR_BLIT_ALPHATEST
:
7889 case WINED3D_BLIT_OP_COLOR_BLIT
:
7893 TRACE("Unsupported blit_op=%d\n", blit_op
);
7897 if (src_pool
== WINED3D_POOL_SYSTEM_MEM
|| dst_pool
== WINED3D_POOL_SYSTEM_MEM
)
7900 src_fixup
= get_complex_fixup(src_format
->color_fixup
);
7901 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
7903 TRACE("Checking support for fixup:\n");
7904 dump_color_fixup_desc(src_format
->color_fixup
);
7907 if (!is_identity_fixup(dst_format
->color_fixup
))
7909 TRACE("Destination fixups are not supported\n");
7913 if (is_identity_fixup(src_format
->color_fixup
))
7919 /* We only support YUV conversions. */
7920 if (!is_complex_fixup(src_format
->color_fixup
))
7922 TRACE("[FAILED]\n");
7928 case COMPLEX_FIXUP_YUY2
:
7929 case COMPLEX_FIXUP_UYVY
:
7930 case COMPLEX_FIXUP_YV12
:
7931 case COMPLEX_FIXUP_NV12
:
7932 case COMPLEX_FIXUP_P8
:
7937 FIXME("Unsupported YUV fixup %#x\n", src_fixup
);
7938 TRACE("[FAILED]\n");
7943 static void arbfp_blit_surface(struct wined3d_device
*device
, enum wined3d_blit_op op
, DWORD filter
,
7944 struct wined3d_surface
*src_surface
, const RECT
*src_rect_in
,
7945 struct wined3d_surface
*dst_surface
, const RECT
*dst_rect_in
,
7946 const struct wined3d_color_key
*color_key
)
7948 unsigned int dst_sub_resource_idx
= surface_get_sub_resource_idx(dst_surface
);
7949 struct wined3d_texture
*src_texture
= src_surface
->container
;
7950 struct wined3d_texture
*dst_texture
= dst_surface
->container
;
7951 struct wined3d_context
*context
;
7952 RECT src_rect
= *src_rect_in
;
7953 RECT dst_rect
= *dst_rect_in
;
7954 struct wined3d_color_key alpha_test_key
;
7956 /* Activate the destination context, set it up for blitting */
7957 context
= context_acquire(device
, dst_surface
);
7959 /* Now load the surface */
7960 if (wined3d_settings
.offscreen_rendering_mode
!= ORM_FBO
7961 && (surface_get_sub_resource(src_surface
)->locations
7962 & (WINED3D_LOCATION_TEXTURE_RGB
| WINED3D_LOCATION_DRAWABLE
))
7963 == WINED3D_LOCATION_DRAWABLE
7964 && !wined3d_resource_is_offscreen(&src_texture
->resource
))
7966 /* Without FBO blits transferring from the drawable to the texture is
7967 * expensive, because we have to flip the data in sysmem. Since we can
7968 * flip in the blitter, we don't actually need that flip anyway. So we
7969 * use the surface's texture as scratch texture, and flip the source
7970 * rectangle instead. */
7971 surface_load_fb_texture(src_surface
, FALSE
, context
);
7973 src_rect
.top
= wined3d_texture_get_level_height(src_texture
, src_surface
->texture_level
) - src_rect
.top
;
7974 src_rect
.bottom
= wined3d_texture_get_level_height(src_texture
, src_surface
->texture_level
) - src_rect
.bottom
;
7977 wined3d_texture_load(src_texture
, context
, FALSE
);
7979 context_apply_blit_state(context
, device
);
7981 if (!wined3d_resource_is_offscreen(&dst_texture
->resource
))
7982 surface_translate_drawable_coords(dst_surface
, context
->win_handle
, &dst_rect
);
7984 if (op
== WINED3D_BLIT_OP_COLOR_BLIT_ALPHATEST
)
7986 const struct wined3d_format
*fmt
= src_texture
->resource
.format
;
7987 alpha_test_key
.color_space_low_value
= 0;
7988 alpha_test_key
.color_space_high_value
= ~(((1u << fmt
->alpha_size
) - 1) << fmt
->alpha_offset
);
7989 color_key
= &alpha_test_key
;
7992 arbfp_blit_set(device
->blit_priv
, context
, src_surface
, color_key
);
7994 /* Draw a textured quad */
7995 draw_textured_quad(src_surface
, context
, &src_rect
, &dst_rect
, filter
);
7997 /* Leave the opengl state valid for blitting */
7998 arbfp_blit_unset(context
->gl_info
);
8000 if (wined3d_settings
.strict_draw_ordering
8001 || (dst_texture
->swapchain
&& (dst_texture
->swapchain
->front_buffer
== dst_texture
)))
8002 context
->gl_info
->gl_ops
.gl
.p_glFlush(); /* Flush to ensure ordering across contexts. */
8004 context_release(context
);
8006 wined3d_texture_validate_location(dst_texture
, dst_sub_resource_idx
, dst_texture
->resource
.draw_binding
);
8007 wined3d_texture_invalidate_location(dst_texture
, dst_sub_resource_idx
, ~dst_texture
->resource
.draw_binding
);
8010 static HRESULT
arbfp_blit_color_fill(struct wined3d_device
*device
, struct wined3d_rendertarget_view
*view
,
8011 const RECT
*rect
, const struct wined3d_color
*color
)
8013 FIXME("Color filling not implemented by arbfp_blit\n");
8014 return WINED3DERR_INVALIDCALL
;
8017 static HRESULT
arbfp_blit_depth_fill(struct wined3d_device
*device
, struct wined3d_rendertarget_view
*view
,
8018 const RECT
*rect
, DWORD clear_flags
, float depth
, DWORD stencil
)
8020 FIXME("Depth/stencil filling not implemented by arbfp_blit.\n");
8021 return WINED3DERR_INVALIDCALL
;
8024 const struct blit_shader arbfp_blit
= {
8029 arbfp_blit_supported
,
8030 arbfp_blit_color_fill
,
8031 arbfp_blit_depth_fill
,