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
30 #include "wined3d_private.h"
32 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader
);
33 WINE_DECLARE_DEBUG_CHANNEL(d3d_constants
);
34 WINE_DECLARE_DEBUG_CHANNEL(d3d
);
36 static BOOL
shader_is_pshader_version(enum wined3d_shader_type type
)
38 return type
== WINED3D_SHADER_TYPE_PIXEL
;
41 static BOOL
shader_is_vshader_version(enum wined3d_shader_type type
)
43 return type
== WINED3D_SHADER_TYPE_VERTEX
;
46 /* Extract a line. Note that this modifies the source string. */
47 static char *get_line(char **ptr
)
52 if (!(q
= strstr(p
, "\n")))
64 static void shader_arb_dump_program_source(const char *source
)
67 char *ptr
, *line
, *tmp
;
69 source_size
= strlen(source
) + 1;
70 tmp
= HeapAlloc(GetProcessHeap(), 0, source_size
);
73 ERR("Failed to allocate %u bytes for shader source.\n", source_size
);
76 memcpy(tmp
, source
, source_size
);
79 while ((line
= get_line(&ptr
))) FIXME(" %s\n", line
);
82 HeapFree(GetProcessHeap(), 0, tmp
);
96 static const char *arb_get_helper_value(enum wined3d_shader_type shader
, enum arb_helper_value value
)
98 if (shader
== WINED3D_SHADER_TYPE_GEOMETRY
)
100 ERR("Geometry shaders are unsupported\n");
104 if (shader
== WINED3D_SHADER_TYPE_PIXEL
)
108 case ARB_ZERO
: return "ps_helper_const.x";
109 case ARB_ONE
: return "ps_helper_const.y";
110 case ARB_TWO
: return "coefmul.x";
111 case ARB_0001
: return "ps_helper_const.xxxy";
112 case ARB_EPS
: return "ps_helper_const.z";
120 case ARB_ZERO
: return "helper_const.x";
121 case ARB_ONE
: return "helper_const.y";
122 case ARB_TWO
: return "helper_const.z";
123 case ARB_EPS
: return "helper_const.w";
124 case ARB_0001
: return "helper_const.xxxy";
125 case ARB_VS_REL_OFFSET
: return "rel_addr_const.y";
128 FIXME("Unmanaged %s shader helper constant requested: %u\n",
129 shader
== WINED3D_SHADER_TYPE_PIXEL
? "pixel" : "vertex", value
);
132 case ARB_ZERO
: return "0.0";
133 case ARB_ONE
: return "1.0";
134 case ARB_TWO
: return "2.0";
135 case ARB_0001
: return "{0.0, 0.0, 0.0, 1.0}";
136 case ARB_EPS
: return "1e-8";
137 default: return "bad";
141 static inline BOOL
ffp_clip_emul(const struct wined3d_context
*context
)
143 return context
->lowest_disabled_stage
< 7;
146 /* ARB_program_shader private data */
165 struct wined3d_shader_loop_control loop_control
;
169 struct arb_ps_np2fixup_info
171 struct ps_np2fixup_info super
;
172 /* For ARB we need an offset value:
173 * With both GLSL and ARB mode the NP2 fixup information (the texture dimensions) are stored in a
174 * consecutive way (GLSL uses a uniform array). Since ARB doesn't know the notion of a "standalone"
175 * array we need an offset to the index inside the program local parameter array. */
179 struct arb_ps_compile_args
181 struct ps_compile_args super
;
183 WORD clip
; /* only a boolean, use a WORD for alignment */
184 unsigned char loop_ctrl
[MAX_CONST_I
][3];
187 struct stb_const_desc
189 unsigned char texunit
;
193 struct arb_ps_compiled_shader
195 struct arb_ps_compile_args args
;
196 struct arb_ps_np2fixup_info np2fixup_info
;
197 struct stb_const_desc bumpenvmatconst
[MAX_TEXTURES
];
198 struct stb_const_desc luminanceconst
[MAX_TEXTURES
];
199 UINT int_consts
[MAX_CONST_I
];
202 unsigned char numbumpenvmatconsts
;
206 struct arb_vs_compile_args
208 struct vs_compile_args super
;
214 unsigned char clip_texcoord
;
215 unsigned char clipplane_mask
;
217 DWORD boolclip_compare
;
222 unsigned char samplers
[4];
223 DWORD samplers_compare
;
225 unsigned char loop_ctrl
[MAX_CONST_I
][3];
228 struct arb_vs_compiled_shader
230 struct arb_vs_compile_args args
;
232 UINT int_consts
[MAX_CONST_I
];
234 char need_color_unclamp
;
238 struct recorded_instruction
240 struct wined3d_shader_instruction ins
;
244 struct shader_arb_ctx_priv
249 /* plain GL_ARB_vertex_program or GL_ARB_fragment_program */
251 /* GL_NV_vertex_progam2_option or GL_NV_fragment_program_option */
253 /* GL_NV_vertex_program3 or GL_NV_fragment_program2 */
257 const struct arb_vs_compile_args
*cur_vs_args
;
258 const struct arb_ps_compile_args
*cur_ps_args
;
259 const struct arb_ps_compiled_shader
*compiled_fprog
;
260 const struct arb_vs_compiled_shader
*compiled_vprog
;
261 struct arb_ps_np2fixup_info
*cur_np2fixup_info
;
262 struct list control_frames
;
266 unsigned int num_loops
, loop_depth
, num_ifcs
;
268 BOOL ps_post_process
;
270 unsigned int vs_clipplanes
;
274 /* For 3.0 vertex shaders */
275 const char *vs_output
[MAX_REG_OUTPUT
];
276 /* For 2.x and earlier vertex shaders */
277 const char *texcrd_output
[8], *color_output
[2], *fog_output
;
279 /* 3.0 pshader input for compatibility with fixed function */
280 const char *ps_input
[MAX_REG_INPUT
];
285 struct wined3d_shader_signature_element
*sig
;
287 struct wine_rb_entry entry
;
290 struct arb_pshader_private
{
291 struct arb_ps_compiled_shader
*gl_shaders
;
292 UINT num_gl_shaders
, shader_array_size
;
293 DWORD input_signature_idx
;
294 DWORD clipplane_emulation
;
298 struct arb_vshader_private
{
299 struct arb_vs_compiled_shader
*gl_shaders
;
300 UINT num_gl_shaders
, shader_array_size
;
304 struct shader_arb_priv
306 GLuint current_vprogram_id
;
307 GLuint current_fprogram_id
;
308 const struct arb_ps_compiled_shader
*compiled_fprog
;
309 const struct arb_vs_compiled_shader
*compiled_vprog
;
310 GLuint depth_blt_vprogram_id
;
311 GLuint depth_blt_fprogram_id_full
[tex_type_count
];
312 GLuint depth_blt_fprogram_id_masked
[tex_type_count
];
313 BOOL use_arbfp_fixed_func
;
314 struct wine_rb_tree fragment_shaders
;
315 BOOL last_ps_const_clamped
;
316 BOOL last_vs_color_unclamp
;
318 struct wine_rb_tree signature_tree
;
321 unsigned int highest_dirty_ps_const
, highest_dirty_vs_const
;
322 char *vshader_const_dirty
, *pshader_const_dirty
;
323 const struct wined3d_context
*last_context
;
325 const struct wined3d_vertex_pipe_ops
*vertex_pipe
;
326 const struct fragment_pipeline
*fragment_pipe
;
327 BOOL ffp_proj_control
;
330 /* Context activation for state handlers is done by the caller. */
332 static BOOL
need_rel_addr_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 (shader_data
->rel_offset
) return TRUE
;
336 if (!reg_maps
->usesmova
) return FALSE
;
337 return !gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
];
340 /* Returns TRUE if result.clip from GL_NV_vertex_program2 should be used and FALSE otherwise */
341 static inline BOOL
use_nv_clip(const struct wined3d_gl_info
*gl_info
)
343 return gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]
344 && !(gl_info
->quirks
& WINED3D_QUIRK_NV_CLIP_BROKEN
);
347 static BOOL
need_helper_const(const struct arb_vshader_private
*shader_data
,
348 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
)
350 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
)) return TRUE
;
351 if (!gl_info
->supported
[NV_VERTEX_PROGRAM
]) return TRUE
; /* Need to init colors. */
352 if (gl_info
->quirks
& WINED3D_QUIRK_ARB_VS_OFFSET_LIMIT
) return TRUE
; /* Load the immval offset. */
353 if (gl_info
->quirks
& WINED3D_QUIRK_SET_TEXCOORD_W
) return TRUE
; /* Have to init texcoords. */
354 if (!use_nv_clip(gl_info
)) return TRUE
; /* Init the clip texcoord */
355 if (reg_maps
->usesnrm
) return TRUE
; /* 0.0 */
356 if (reg_maps
->usespow
) return TRUE
; /* EPS, 0.0 and 1.0 */
357 if (reg_maps
->fog
) return TRUE
; /* Clamping fog coord, 0.0 and 1.0 */
361 static unsigned int reserved_vs_const(const struct arb_vshader_private
*shader_data
,
362 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
)
364 unsigned int ret
= 1;
365 /* We use one PARAM for the pos fixup, and in some cases one to load
366 * some immediate values into the shader. */
367 if (need_helper_const(shader_data
, reg_maps
, gl_info
)) ++ret
;
368 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
)) ++ret
;
372 /* Loads floating point constants into the currently set ARB_vertex/fragment_program.
373 * When constant_list == NULL, it will load all the constants.
375 * @target_type should be either GL_VERTEX_PROGRAM_ARB (for vertex shaders)
376 * or GL_FRAGMENT_PROGRAM_ARB (for pixel shaders)
378 /* Context activation is done by the caller. */
379 static unsigned int shader_arb_load_constantsF(const struct wined3d_shader
*shader
,
380 const struct wined3d_gl_info
*gl_info
, GLuint target_type
, unsigned int max_constants
,
381 const float *constants
, char *dirty_consts
)
383 struct wined3d_shader_lconst
*lconst
;
387 if (TRACE_ON(d3d_constants
))
389 for(i
= 0; i
< max_constants
; i
++) {
390 if(!dirty_consts
[i
]) continue;
391 TRACE_(d3d_constants
)("Loading constants %i: %f, %f, %f, %f\n", i
,
392 constants
[i
* 4 + 0], constants
[i
* 4 + 1],
393 constants
[i
* 4 + 2], constants
[i
* 4 + 3]);
399 /* In 1.X pixel shaders constants are implicitly clamped in the range [-1;1] */
400 if (target_type
== GL_FRAGMENT_PROGRAM_ARB
&& shader
->reg_maps
.shader_version
.major
== 1)
403 /* ps 1.x supports only 8 constants, clamp only those. When switching between 1.x and higher
404 * shaders, the first 8 constants are marked dirty for reload
406 for(; i
< min(8, max_constants
); i
++) {
407 if(!dirty_consts
[i
]) continue;
411 if (constants
[j
+ 0] > 1.0f
) lcl_const
[0] = 1.0f
;
412 else if (constants
[j
+ 0] < -1.0f
) lcl_const
[0] = -1.0f
;
413 else lcl_const
[0] = constants
[j
+ 0];
415 if (constants
[j
+ 1] > 1.0f
) lcl_const
[1] = 1.0f
;
416 else if (constants
[j
+ 1] < -1.0f
) lcl_const
[1] = -1.0f
;
417 else lcl_const
[1] = constants
[j
+ 1];
419 if (constants
[j
+ 2] > 1.0f
) lcl_const
[2] = 1.0f
;
420 else if (constants
[j
+ 2] < -1.0f
) lcl_const
[2] = -1.0f
;
421 else lcl_const
[2] = constants
[j
+ 2];
423 if (constants
[j
+ 3] > 1.0f
) lcl_const
[3] = 1.0f
;
424 else if (constants
[j
+ 3] < -1.0f
) lcl_const
[3] = -1.0f
;
425 else lcl_const
[3] = constants
[j
+ 3];
427 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, i
, lcl_const
));
430 /* If further constants are dirty, reload them without clamping.
432 * The alternative is not to touch them, but then we cannot reset the dirty constant count
433 * to zero. That's bad for apps that only use PS 1.x shaders, because in that case the code
434 * above would always re-check the first 8 constants since max_constant remains at the init
439 if (gl_info
->supported
[EXT_GPU_PROGRAM_PARAMETERS
])
441 /* TODO: Benchmark if we're better of with finding the dirty constants ourselves,
442 * or just reloading *all* constants at once
444 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type, i, max_constants, constants + (i * 4)));
446 for(; i
< max_constants
; i
++) {
447 if(!dirty_consts
[i
]) continue;
449 /* Find the next block of dirty constants */
452 for(i
++; (i
< max_constants
) && dirty_consts
[i
]; i
++) {
456 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type
, j
, i
- j
, constants
+ (j
* 4)));
459 for(; i
< max_constants
; i
++) {
460 if(dirty_consts
[i
]) {
462 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, i
, constants
+ (i
* 4)));
466 checkGLcall("glProgramEnvParameter4fvARB()");
468 /* Load immediate constants */
469 if (shader
->load_local_constsF
)
471 if (TRACE_ON(d3d_shader
))
473 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
475 GLfloat
* values
= (GLfloat
*)lconst
->value
;
476 TRACE_(d3d_constants
)("Loading local constants %i: %f, %f, %f, %f\n", lconst
->idx
,
477 values
[0], values
[1], values
[2], values
[3]);
480 /* Immediate constants are clamped for 1.X shaders at loading times */
482 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
484 dirty_consts
[lconst
->idx
] = 1; /* Dirtify so the non-immediate constant overwrites it next time */
485 ret
= max(ret
, lconst
->idx
+ 1);
486 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, lconst
->idx
, (GLfloat
*)lconst
->value
));
488 checkGLcall("glProgramEnvParameter4fvARB()");
489 return ret
; /* The loaded immediate constants need reloading for the next shader */
491 return 0; /* No constants are dirty now */
495 /* Loads the texture dimensions for NP2 fixup into the currently set
496 * ARB_[vertex/fragment]_programs. */
497 static void shader_arb_load_np2fixup_constants(const struct arb_ps_np2fixup_info
*fixup
,
498 const struct wined3d_gl_info
*gl_info
, const struct wined3d_state
*state
)
500 GLfloat np2fixup_constants
[4 * MAX_FRAGMENT_SAMPLERS
];
501 WORD active
= fixup
->super
.active
;
507 for (i
= 0; active
; active
>>= 1, ++i
)
509 const struct wined3d_texture
*tex
= state
->textures
[i
];
510 unsigned char idx
= fixup
->super
.idx
[i
];
511 GLfloat
*tex_dim
= &np2fixup_constants
[(idx
>> 1) * 4];
518 ERR("Nonexistent texture is flagged for NP2 texcoord fixup.\n");
524 tex_dim
[2] = tex
->pow2_matrix
[0];
525 tex_dim
[3] = tex
->pow2_matrix
[5];
529 tex_dim
[0] = tex
->pow2_matrix
[0];
530 tex_dim
[1] = tex
->pow2_matrix
[5];
534 for (i
= 0; i
< fixup
->super
.num_consts
; ++i
)
536 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
537 fixup
->offset
+ i
, &np2fixup_constants
[i
* 4]));
541 /* Context activation is done by the caller. */
542 static void shader_arb_ps_local_constants(const struct arb_ps_compiled_shader
*gl_shader
,
543 const struct wined3d_context
*context
, const struct wined3d_state
*state
, UINT rt_height
)
545 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
548 for(i
= 0; i
< gl_shader
->numbumpenvmatconsts
; i
++)
550 int texunit
= gl_shader
->bumpenvmatconst
[i
].texunit
;
552 /* The state manager takes care that this function is always called if the bump env matrix changes */
553 const float *data
= (const float *)&state
->texture_states
[texunit
][WINED3D_TSS_BUMPENV_MAT00
];
554 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
555 gl_shader
->bumpenvmatconst
[i
].const_num
, data
));
557 if (gl_shader
->luminanceconst
[i
].const_num
!= WINED3D_CONST_NUM_UNUSED
)
559 /* WINED3D_TSS_BUMPENVLSCALE and WINED3D_TSS_BUMPENVLOFFSET are next to each other.
560 * point gl to the scale, and load 4 floats. x = scale, y = offset, z and w are junk, we
561 * don't care about them. The pointers are valid for sure because the stateblock is bigger.
562 * (they're WINED3D_TSS_TEXTURETRANSFORMFLAGS and WINED3D_TSS_ADDRESSW, so most likely 0 or NaN
564 const float *scale
= (const float *)&state
->texture_states
[texunit
][WINED3D_TSS_BUMPENV_LSCALE
];
565 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
566 gl_shader
->luminanceconst
[i
].const_num
, scale
));
569 checkGLcall("Load bumpmap consts");
571 if(gl_shader
->ycorrection
!= WINED3D_CONST_NUM_UNUSED
)
573 /* ycorrection.x: Backbuffer height(onscreen) or 0(offscreen).
574 * ycorrection.y: -1.0(onscreen), 1.0(offscreen)
579 val
[0] = context
->render_offscreen
? 0.0f
: (float) rt_height
;
580 val
[1] = context
->render_offscreen
? 1.0f
: -1.0f
;
583 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, gl_shader
->ycorrection
, val
));
584 checkGLcall("y correction loading");
587 if (!gl_shader
->num_int_consts
) return;
589 for(i
= 0; i
< MAX_CONST_I
; i
++)
591 if(gl_shader
->int_consts
[i
] != WINED3D_CONST_NUM_UNUSED
)
594 val
[0] = (float)state
->ps_consts_i
[4 * i
];
595 val
[1] = (float)state
->ps_consts_i
[4 * i
+ 1];
596 val
[2] = (float)state
->ps_consts_i
[4 * i
+ 2];
599 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, gl_shader
->int_consts
[i
], val
));
602 checkGLcall("Load ps int consts");
605 /* Context activation is done by the caller. */
606 static void shader_arb_vs_local_constants(const struct arb_vs_compiled_shader
*gl_shader
,
607 const struct wined3d_context
*context
, const struct wined3d_state
*state
)
609 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
610 float position_fixup
[4];
613 /* Upload the position fixup */
614 shader_get_position_fixup(context
, state
, position_fixup
);
615 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, gl_shader
->pos_fixup
, position_fixup
));
617 if (!gl_shader
->num_int_consts
) return;
619 for(i
= 0; i
< MAX_CONST_I
; i
++)
621 if(gl_shader
->int_consts
[i
] != WINED3D_CONST_NUM_UNUSED
)
624 val
[0] = (float)state
->vs_consts_i
[4 * i
];
625 val
[1] = (float)state
->vs_consts_i
[4 * i
+ 1];
626 val
[2] = (float)state
->vs_consts_i
[4 * i
+ 2];
629 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, gl_shader
->int_consts
[i
], val
));
632 checkGLcall("Load vs int consts");
635 static void shader_arb_select(void *shader_priv
, struct wined3d_context
*context
,
636 const struct wined3d_state
*state
);
639 * Loads the app-supplied constants into the currently set ARB_[vertex/fragment]_programs.
641 * We only support float constants in ARB at the moment, so don't
642 * worry about the Integers or Booleans
644 /* Context activation is done by the caller (state handler). */
645 static void shader_arb_load_constants_internal(struct shader_arb_priv
*priv
,
646 struct wined3d_context
*context
, const struct wined3d_state
*state
,
647 BOOL usePixelShader
, BOOL useVertexShader
, BOOL from_shader_select
)
649 const struct wined3d_d3d_info
*d3d_info
= context
->d3d_info
;
650 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
652 if (!from_shader_select
)
654 const struct wined3d_shader
*vshader
= state
->shader
[WINED3D_SHADER_TYPE_VERTEX
];
655 const struct wined3d_shader
*pshader
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
658 && (vshader
->reg_maps
.boolean_constants
659 || (!gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]
660 && (vshader
->reg_maps
.integer_constants
& ~vshader
->reg_maps
.local_int_consts
))))
662 TRACE("bool/integer vertex shader constants potentially modified, forcing shader reselection.\n");
663 shader_arb_select(priv
, context
, state
);
666 && (pshader
->reg_maps
.boolean_constants
667 || (!gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
]
668 && (pshader
->reg_maps
.integer_constants
& ~pshader
->reg_maps
.local_int_consts
))))
670 TRACE("bool/integer pixel shader constants potentially modified, forcing shader reselection.\n");
671 shader_arb_select(priv
, context
, state
);
675 if (context
!= priv
->last_context
)
677 memset(priv
->vshader_const_dirty
, 1,
678 sizeof(*priv
->vshader_const_dirty
) * d3d_info
->limits
.vs_uniform_count
);
679 priv
->highest_dirty_vs_const
= d3d_info
->limits
.vs_uniform_count
;
681 memset(priv
->pshader_const_dirty
, 1,
682 sizeof(*priv
->pshader_const_dirty
) * d3d_info
->limits
.ps_uniform_count
);
683 priv
->highest_dirty_ps_const
= d3d_info
->limits
.ps_uniform_count
;
685 priv
->last_context
= context
;
690 const struct wined3d_shader
*vshader
= state
->shader
[WINED3D_SHADER_TYPE_VERTEX
];
691 const struct arb_vs_compiled_shader
*gl_shader
= priv
->compiled_vprog
;
693 /* Load DirectX 9 float constants for vertex shader */
694 priv
->highest_dirty_vs_const
= shader_arb_load_constantsF(vshader
, gl_info
, GL_VERTEX_PROGRAM_ARB
,
695 priv
->highest_dirty_vs_const
, state
->vs_consts_f
, priv
->vshader_const_dirty
);
696 shader_arb_vs_local_constants(gl_shader
, context
, state
);
701 const struct wined3d_shader
*pshader
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
702 const struct arb_ps_compiled_shader
*gl_shader
= priv
->compiled_fprog
;
703 UINT rt_height
= state
->fb
->render_targets
[0]->resource
.height
;
705 /* Load DirectX 9 float constants for pixel shader */
706 priv
->highest_dirty_ps_const
= shader_arb_load_constantsF(pshader
, gl_info
, GL_FRAGMENT_PROGRAM_ARB
,
707 priv
->highest_dirty_ps_const
, state
->ps_consts_f
, priv
->pshader_const_dirty
);
708 shader_arb_ps_local_constants(gl_shader
, context
, state
, rt_height
);
710 if (context
->constant_update_mask
& WINED3D_SHADER_CONST_PS_NP2_FIXUP
)
711 shader_arb_load_np2fixup_constants(&gl_shader
->np2fixup_info
, gl_info
, state
);
715 static void shader_arb_load_constants(void *shader_priv
, struct wined3d_context
*context
,
716 const struct wined3d_state
*state
)
718 BOOL vs
= use_vs(state
);
719 BOOL ps
= use_ps(state
);
721 shader_arb_load_constants_internal(shader_priv
, context
, state
, ps
, vs
, FALSE
);
724 static void shader_arb_update_float_vertex_constants(struct wined3d_device
*device
, UINT start
, UINT count
)
726 struct wined3d_context
*context
= context_get_current();
727 struct shader_arb_priv
*priv
= device
->shader_priv
;
730 for (i
= 0; i
< device
->context_count
; ++i
)
732 device
->contexts
[i
]->constant_update_mask
|= WINED3D_SHADER_CONST_VS_F
;
735 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
736 * context. On a context switch the old context will be fully dirtified */
737 if (!context
|| context
->swapchain
->device
!= device
) return;
739 memset(priv
->vshader_const_dirty
+ start
, 1, sizeof(*priv
->vshader_const_dirty
) * count
);
740 priv
->highest_dirty_vs_const
= max(priv
->highest_dirty_vs_const
, start
+ count
);
743 static void shader_arb_update_float_pixel_constants(struct wined3d_device
*device
, UINT start
, UINT count
)
745 struct wined3d_context
*context
= context_get_current();
746 struct shader_arb_priv
*priv
= device
->shader_priv
;
749 for (i
= 0; i
< device
->context_count
; ++i
)
751 device
->contexts
[i
]->constant_update_mask
|= WINED3D_SHADER_CONST_PS_F
;
754 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
755 * context. On a context switch the old context will be fully dirtified */
756 if (!context
|| context
->swapchain
->device
!= device
) return;
758 memset(priv
->pshader_const_dirty
+ start
, 1, sizeof(*priv
->pshader_const_dirty
) * count
);
759 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, start
+ count
);
762 static void shader_arb_append_imm_vec4(struct wined3d_shader_buffer
*buffer
, const float *values
)
766 wined3d_ftoa(values
[0], str
[0]);
767 wined3d_ftoa(values
[1], str
[1]);
768 wined3d_ftoa(values
[2], str
[2]);
769 wined3d_ftoa(values
[3], str
[3]);
770 shader_addline(buffer
, "{%s, %s, %s, %s}", str
[0], str
[1], str
[2], str
[3]);
773 /* Generate the variable & register declarations for the ARB_vertex_program output target */
774 static void shader_generate_arb_declarations(const struct wined3d_shader
*shader
,
775 const struct wined3d_shader_reg_maps
*reg_maps
, struct wined3d_shader_buffer
*buffer
,
776 const struct wined3d_gl_info
*gl_info
, DWORD
*num_clipplanes
,
777 const struct shader_arb_ctx_priv
*ctx
)
780 char pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
781 const struct wined3d_shader_lconst
*lconst
;
782 unsigned max_constantsF
;
785 /* In pixel shaders, all private constants are program local, we don't need anything
786 * from program.env. Thus we can advertise the full set of constants in pixel shaders.
787 * If we need a private constant the GL implementation will squeeze it in somewhere
789 * With vertex shaders we need the posFixup and on some GL implementations 4 helper
790 * immediate values. The posFixup is loaded using program.env for now, so always
791 * subtract one from the number of constants. If the shader uses indirect addressing,
792 * account for the helper const too because we have to declare all available d3d constants
793 * and don't know which are actually used.
797 max_constantsF
= gl_info
->limits
.arb_ps_native_constants
;
798 /* 24 is the minimum MAX_PROGRAM_ENV_PARAMETERS_ARB value. */
799 if (max_constantsF
< 24)
800 max_constantsF
= gl_info
->limits
.arb_ps_float_constants
;
804 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
805 max_constantsF
= gl_info
->limits
.arb_vs_native_constants
;
806 /* 96 is the minimum MAX_PROGRAM_ENV_PARAMETERS_ARB value.
807 * Also prevents max_constantsF from becoming less than 0 and
809 if (max_constantsF
< 96)
810 max_constantsF
= gl_info
->limits
.arb_vs_float_constants
;
812 if (reg_maps
->usesrelconstF
)
814 DWORD highest_constf
= 0, clip_limit
;
816 max_constantsF
-= reserved_vs_const(shader_data
, reg_maps
, gl_info
);
817 max_constantsF
-= count_bits(reg_maps
->integer_constants
);
818 max_constantsF
-= gl_info
->reserved_arb_constants
;
820 for (i
= 0; i
< shader
->limits
.constant_float
; ++i
)
823 DWORD shift
= i
& 0x1f;
824 if(reg_maps
->constf
[idx
] & (1 << shift
)) highest_constf
= i
;
827 if(use_nv_clip(gl_info
) && ctx
->target_version
>= NV2
)
829 if(ctx
->cur_vs_args
->super
.clip_enabled
)
830 clip_limit
= gl_info
->limits
.clipplanes
;
836 unsigned int mask
= ctx
->cur_vs_args
->clip
.boolclip
.clipplane_mask
;
837 clip_limit
= min(count_bits(mask
), 4);
839 *num_clipplanes
= min(clip_limit
, max_constantsF
- highest_constf
- 1);
840 max_constantsF
-= *num_clipplanes
;
841 if(*num_clipplanes
< clip_limit
)
843 WARN("Only %u clipplanes out of %u enabled\n", *num_clipplanes
, gl_info
->limits
.clipplanes
);
848 if (ctx
->target_version
>= NV2
) *num_clipplanes
= gl_info
->limits
.clipplanes
;
849 else *num_clipplanes
= min(gl_info
->limits
.clipplanes
, 4);
853 for (i
= 0, map
= reg_maps
->temporary
; map
; map
>>= 1, ++i
)
855 if (map
& 1) shader_addline(buffer
, "TEMP R%u;\n", i
);
858 for (i
= 0, map
= reg_maps
->address
; map
; map
>>= 1, ++i
)
860 if (map
& 1) shader_addline(buffer
, "ADDRESS A%u;\n", i
);
863 if (pshader
&& reg_maps
->shader_version
.major
== 1 && reg_maps
->shader_version
.minor
<= 3)
865 for (i
= 0, map
= reg_maps
->texcoord
; map
; map
>>= 1, ++i
)
867 if (map
& 1) shader_addline(buffer
, "TEMP T%u;\n", i
);
871 if (!shader
->load_local_constsF
)
873 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
876 value
= (const float *)lconst
->value
;
877 shader_addline(buffer
, "PARAM C%u = ", lconst
->idx
);
878 shader_arb_append_imm_vec4(buffer
, value
);
879 shader_addline(buffer
, ";\n");
883 /* After subtracting privately used constants from the hardware limit(they are loaded as
884 * local constants), make sure the shader doesn't violate the env constant limit
888 max_constantsF
= min(max_constantsF
, gl_info
->limits
.arb_ps_float_constants
);
892 max_constantsF
= min(max_constantsF
, gl_info
->limits
.arb_vs_float_constants
);
895 /* Avoid declaring more constants than needed */
896 max_constantsF
= min(max_constantsF
, shader
->limits
.constant_float
);
898 /* we use the array-based constants array if the local constants are marked for loading,
899 * because then we use indirect addressing, or when the local constant list is empty,
900 * because then we don't know if we're using indirect addressing or not. If we're hardcoding
901 * local constants do not declare the loaded constants as an array because ARB compilers usually
902 * do not optimize unused constants away
904 if (reg_maps
->usesrelconstF
)
906 /* Need to PARAM the environment parameters (constants) so we can use relative addressing */
907 shader_addline(buffer
, "PARAM C[%d] = { program.env[0..%d] };\n",
908 max_constantsF
, max_constantsF
- 1);
910 for(i
= 0; i
< max_constantsF
; i
++) {
913 mask
= 1 << (i
& 0x1f);
914 if (!shader_constant_is_local(shader
, i
) && (reg_maps
->constf
[idx
] & mask
))
916 shader_addline(buffer
, "PARAM C%d = program.env[%d];\n",i
, i
);
922 static const char * const shift_tab
[] = {
923 "dummy", /* 0 (none) */
924 "coefmul.x", /* 1 (x2) */
925 "coefmul.y", /* 2 (x4) */
926 "coefmul.z", /* 3 (x8) */
927 "coefmul.w", /* 4 (x16) */
928 "dummy", /* 5 (x32) */
929 "dummy", /* 6 (x64) */
930 "dummy", /* 7 (x128) */
931 "dummy", /* 8 (d256) */
932 "dummy", /* 9 (d128) */
933 "dummy", /* 10 (d64) */
934 "dummy", /* 11 (d32) */
935 "coefdiv.w", /* 12 (d16) */
936 "coefdiv.z", /* 13 (d8) */
937 "coefdiv.y", /* 14 (d4) */
938 "coefdiv.x" /* 15 (d2) */
941 static void shader_arb_get_write_mask(const struct wined3d_shader_instruction
*ins
,
942 const struct wined3d_shader_dst_param
*dst
, char *write_mask
)
944 char *ptr
= write_mask
;
946 if (dst
->write_mask
!= WINED3DSP_WRITEMASK_ALL
)
949 if (dst
->write_mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
950 if (dst
->write_mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
951 if (dst
->write_mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
952 if (dst
->write_mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
958 static void shader_arb_get_swizzle(const struct wined3d_shader_src_param
*param
, BOOL fixup
, char *swizzle_str
)
960 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
961 * but addressed as "rgba". To fix this we need to swap the register's x
962 * and z components. */
963 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
964 char *ptr
= swizzle_str
;
966 /* swizzle bits fields: wwzzyyxx */
967 DWORD swizzle
= param
->swizzle
;
968 DWORD swizzle_x
= swizzle
& 0x03;
969 DWORD swizzle_y
= (swizzle
>> 2) & 0x03;
970 DWORD swizzle_z
= (swizzle
>> 4) & 0x03;
971 DWORD swizzle_w
= (swizzle
>> 6) & 0x03;
973 /* If the swizzle is the default swizzle (ie, "xyzw"), we don't need to
974 * generate a swizzle string. Unless we need to our own swizzling. */
975 if (swizzle
!= WINED3DSP_NOSWIZZLE
|| fixup
)
978 if (swizzle_x
== swizzle_y
&& swizzle_x
== swizzle_z
&& swizzle_x
== swizzle_w
) {
979 *ptr
++ = swizzle_chars
[swizzle_x
];
981 *ptr
++ = swizzle_chars
[swizzle_x
];
982 *ptr
++ = swizzle_chars
[swizzle_y
];
983 *ptr
++ = swizzle_chars
[swizzle_z
];
984 *ptr
++ = swizzle_chars
[swizzle_w
];
991 static void shader_arb_request_a0(const struct wined3d_shader_instruction
*ins
, const char *src
)
993 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
994 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
996 if (!strcmp(priv
->addr_reg
, src
)) return;
998 strcpy(priv
->addr_reg
, src
);
999 shader_addline(buffer
, "ARL A0.x, %s;\n", src
);
1002 static void shader_arb_get_src_param(const struct wined3d_shader_instruction
*ins
,
1003 const struct wined3d_shader_src_param
*src
, unsigned int tmpreg
, char *outregstr
);
1005 static void shader_arb_get_register_name(const struct wined3d_shader_instruction
*ins
,
1006 const struct wined3d_shader_register
*reg
, char *register_name
, BOOL
*is_color
)
1008 /* oPos, oFog and oPts in D3D */
1009 static const char * const rastout_reg_names
[] = {"TMP_OUT", "TMP_FOGCOORD", "result.pointsize"};
1010 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
1011 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
1012 BOOL pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
1013 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1019 case WINED3DSPR_TEMP
:
1020 sprintf(register_name
, "R%u", reg
->idx
[0].offset
);
1023 case WINED3DSPR_INPUT
:
1026 if (reg_maps
->shader_version
.major
< 3)
1028 if (!reg
->idx
[0].offset
)
1029 strcpy(register_name
, "fragment.color.primary");
1031 strcpy(register_name
, "fragment.color.secondary");
1035 if (reg
->idx
[0].rel_addr
)
1038 shader_arb_get_src_param(ins
, reg
->idx
[0].rel_addr
, 0, rel_reg
);
1040 if (!strcmp(rel_reg
, "**aL_emul**"))
1042 DWORD idx
= ctx
->aL
+ reg
->idx
[0].offset
;
1043 if(idx
< MAX_REG_INPUT
)
1045 strcpy(register_name
, ctx
->ps_input
[idx
]);
1049 ERR("Pixel shader input register out of bounds: %u\n", idx
);
1050 sprintf(register_name
, "out_of_bounds_%u", idx
);
1053 else if (reg_maps
->input_registers
& 0x0300)
1055 /* There are two ways basically:
1057 * 1) Use the unrolling code that is used for loop emulation and unroll the loop.
1058 * That means trouble if the loop also contains a breakc or if the control values
1059 * aren't local constants.
1060 * 2) Generate an if block that checks if aL.y < 8, == 8 or == 9 and selects the
1061 * source dynamically. The trouble is that we cannot simply read aL.y because it
1062 * is an ADDRESS register. We could however push it, load .zw with a value and use
1063 * ADAC to load the condition code register and pop it again afterwards
1065 FIXME("Relative input register addressing with more than 8 registers\n");
1067 /* This is better than nothing for now */
1068 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1070 else if(ctx
->cur_ps_args
->super
.vp_mode
!= vertexshader
)
1072 /* This is problematic because we'd have to consult the ctx->ps_input strings
1073 * for where to find the varying. Some may be "0.0", others can be texcoords or
1074 * colors. This needs either a pipeline replacement to make the vertex shader feed
1075 * proper varyings, or loop unrolling
1077 * For now use the texcoords and hope for the best
1079 FIXME("Non-vertex shader varying input with indirect addressing\n");
1080 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1084 /* D3D supports indirect addressing only with aL in loop registers. The loop instruction
1085 * pulls GL_NV_fragment_program2 in
1087 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1092 if (reg
->idx
[0].offset
< MAX_REG_INPUT
)
1094 strcpy(register_name
, ctx
->ps_input
[reg
->idx
[0].offset
]);
1098 ERR("Pixel shader input register out of bounds: %u\n", reg
->idx
[0].offset
);
1099 sprintf(register_name
, "out_of_bounds_%u", reg
->idx
[0].offset
);
1106 if (ctx
->cur_vs_args
->super
.swizzle_map
& (1 << reg
->idx
[0].offset
))
1108 sprintf(register_name
, "vertex.attrib[%u]", reg
->idx
[0].offset
);
1112 case WINED3DSPR_CONST
:
1113 if (!pshader
&& reg
->idx
[0].rel_addr
)
1115 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1116 UINT rel_offset
= ctx
->target_version
== ARB
? shader_data
->rel_offset
: 0;
1119 if (reg_maps
->shader_version
.major
< 2)
1121 sprintf(rel_reg
, "A0.x");
1125 shader_arb_get_src_param(ins
, reg
->idx
[0].rel_addr
, 0, rel_reg
);
1126 if (ctx
->target_version
== ARB
)
1128 if (!strcmp(rel_reg
, "**aL_emul**"))
1132 shader_arb_request_a0(ins
, rel_reg
);
1133 sprintf(rel_reg
, "A0.x");
1138 sprintf(register_name
, "C[%u]", ctx
->aL
+ reg
->idx
[0].offset
);
1139 else if (reg
->idx
[0].offset
>= rel_offset
)
1140 sprintf(register_name
, "C[%s + %u]", rel_reg
, reg
->idx
[0].offset
- rel_offset
);
1142 sprintf(register_name
, "C[%s - %u]", rel_reg
, rel_offset
- reg
->idx
[0].offset
);
1146 if (reg_maps
->usesrelconstF
)
1147 sprintf(register_name
, "C[%u]", reg
->idx
[0].offset
);
1149 sprintf(register_name
, "C%u", reg
->idx
[0].offset
);
1153 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
1156 if (reg_maps
->shader_version
.major
== 1
1157 && reg_maps
->shader_version
.minor
<= 3)
1158 /* In ps <= 1.3, Tx is a temporary register as destination
1159 * to all instructions, and as source to most instructions.
1160 * For some instructions it is the texcoord input. Those
1161 * instructions know about the special use. */
1162 sprintf(register_name
, "T%u", reg
->idx
[0].offset
);
1164 /* In ps 1.4 and 2.x Tx is always a (read-only) varying. */
1165 sprintf(register_name
, "fragment.texcoord[%u]", reg
->idx
[0].offset
);
1169 if (reg_maps
->shader_version
.major
== 1 || ctx
->target_version
>= NV2
)
1170 sprintf(register_name
, "A%u", reg
->idx
[0].offset
);
1172 sprintf(register_name
, "A%u_SHADOW", reg
->idx
[0].offset
);
1176 case WINED3DSPR_COLOROUT
:
1177 if (ctx
->ps_post_process
&& !reg
->idx
[0].offset
)
1179 strcpy(register_name
, "TMP_COLOR");
1183 if (ctx
->cur_ps_args
->super
.srgb_correction
)
1184 FIXME("sRGB correction on higher render targets.\n");
1185 if (reg_maps
->rt_mask
> 1)
1186 sprintf(register_name
, "result.color[%u]", reg
->idx
[0].offset
);
1188 strcpy(register_name
, "result.color");
1192 case WINED3DSPR_RASTOUT
:
1193 if (reg
->idx
[0].offset
== 1)
1194 sprintf(register_name
, "%s", ctx
->fog_output
);
1196 sprintf(register_name
, "%s", rastout_reg_names
[reg
->idx
[0].offset
]);
1199 case WINED3DSPR_DEPTHOUT
:
1200 strcpy(register_name
, "result.depth");
1203 case WINED3DSPR_ATTROUT
:
1204 /* case WINED3DSPR_OUTPUT: */
1206 sprintf(register_name
, "oD[%u]", reg
->idx
[0].offset
);
1208 strcpy(register_name
, ctx
->color_output
[reg
->idx
[0].offset
]);
1211 case WINED3DSPR_TEXCRDOUT
:
1213 sprintf(register_name
, "oT[%u]", reg
->idx
[0].offset
);
1214 else if (reg_maps
->shader_version
.major
< 3)
1215 strcpy(register_name
, ctx
->texcrd_output
[reg
->idx
[0].offset
]);
1217 strcpy(register_name
, ctx
->vs_output
[reg
->idx
[0].offset
]);
1220 case WINED3DSPR_LOOP
:
1221 if(ctx
->target_version
>= NV2
)
1223 /* Pshader has an implicitly declared loop index counter A0.x that cannot be renamed */
1224 if(pshader
) sprintf(register_name
, "A0.x");
1225 else sprintf(register_name
, "aL.y");
1229 /* Unfortunately this code cannot return the value of ctx->aL here. An immediate value
1230 * would be valid, but if aL is used for indexing(its only use), there's likely an offset,
1231 * thus the result would be something like C[15 + 30], which is not valid in the ARB program
1232 * grammar. So return a marker for the emulated aL and intercept it in constant and varying
1235 sprintf(register_name
, "**aL_emul**");
1240 case WINED3DSPR_CONSTINT
:
1241 sprintf(register_name
, "I%u", reg
->idx
[0].offset
);
1244 case WINED3DSPR_MISCTYPE
:
1245 if (!reg
->idx
[0].offset
)
1246 sprintf(register_name
, "vpos");
1247 else if (reg
->idx
[0].offset
== 1)
1248 sprintf(register_name
, "fragment.facing.x");
1250 FIXME("Unknown MISCTYPE register index %u.\n", reg
->idx
[0].offset
);
1254 FIXME("Unhandled register type %#x[%u].\n", reg
->type
, reg
->idx
[0].offset
);
1255 sprintf(register_name
, "unrecognized_register[%u]", reg
->idx
[0].offset
);
1260 static void shader_arb_get_dst_param(const struct wined3d_shader_instruction
*ins
,
1261 const struct wined3d_shader_dst_param
*wined3d_dst
, char *str
)
1263 char register_name
[255];
1267 shader_arb_get_register_name(ins
, &wined3d_dst
->reg
, register_name
, &is_color
);
1268 strcpy(str
, register_name
);
1270 shader_arb_get_write_mask(ins
, wined3d_dst
, write_mask
);
1271 strcat(str
, write_mask
);
1274 static const char *shader_arb_get_fixup_swizzle(enum fixup_channel_source channel_source
)
1276 switch(channel_source
)
1278 case CHANNEL_SOURCE_ZERO
: return "0";
1279 case CHANNEL_SOURCE_ONE
: return "1";
1280 case CHANNEL_SOURCE_X
: return "x";
1281 case CHANNEL_SOURCE_Y
: return "y";
1282 case CHANNEL_SOURCE_Z
: return "z";
1283 case CHANNEL_SOURCE_W
: return "w";
1285 FIXME("Unhandled channel source %#x\n", channel_source
);
1290 static void gen_color_correction(struct wined3d_shader_buffer
*buffer
, const char *reg
,
1291 DWORD dst_mask
, const char *one
, const char *two
, struct color_fixup_desc fixup
)
1295 if (is_complex_fixup(fixup
))
1297 enum complex_fixup complex_fixup
= get_complex_fixup(fixup
);
1298 FIXME("Complex fixup (%#x) not supported\n", complex_fixup
);
1303 if (fixup
.x_source
!= CHANNEL_SOURCE_X
) mask
|= WINED3DSP_WRITEMASK_0
;
1304 if (fixup
.y_source
!= CHANNEL_SOURCE_Y
) mask
|= WINED3DSP_WRITEMASK_1
;
1305 if (fixup
.z_source
!= CHANNEL_SOURCE_Z
) mask
|= WINED3DSP_WRITEMASK_2
;
1306 if (fixup
.w_source
!= CHANNEL_SOURCE_W
) mask
|= WINED3DSP_WRITEMASK_3
;
1311 shader_addline(buffer
, "SWZ %s, %s, %s, %s, %s, %s;\n", reg
, reg
,
1312 shader_arb_get_fixup_swizzle(fixup
.x_source
), shader_arb_get_fixup_swizzle(fixup
.y_source
),
1313 shader_arb_get_fixup_swizzle(fixup
.z_source
), shader_arb_get_fixup_swizzle(fixup
.w_source
));
1317 if (fixup
.x_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_0
;
1318 if (fixup
.y_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_1
;
1319 if (fixup
.z_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_2
;
1320 if (fixup
.w_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_3
;
1326 char *ptr
= reg_mask
;
1328 if (mask
!= WINED3DSP_WRITEMASK_ALL
)
1331 if (mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
1332 if (mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
1333 if (mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
1334 if (mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
1338 shader_addline(buffer
, "MAD %s%s, %s, %s, -%s;\n", reg
, reg_mask
, reg
, two
, one
);
1342 static const char *shader_arb_get_modifier(const struct wined3d_shader_instruction
*ins
)
1345 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1346 if (!ins
->dst_count
) return "";
1348 mod
= ins
->dst
[0].modifiers
;
1350 /* Silently ignore PARTIALPRECISION if its not supported */
1351 if(priv
->target_version
== ARB
) mod
&= ~WINED3DSPDM_PARTIALPRECISION
;
1353 if(mod
& WINED3DSPDM_MSAMPCENTROID
)
1355 FIXME("Unhandled modifier WINED3DSPDM_MSAMPCENTROID\n");
1356 mod
&= ~WINED3DSPDM_MSAMPCENTROID
;
1361 case WINED3DSPDM_SATURATE
| WINED3DSPDM_PARTIALPRECISION
:
1364 case WINED3DSPDM_SATURATE
:
1367 case WINED3DSPDM_PARTIALPRECISION
:
1374 FIXME("Unknown modifiers 0x%08x\n", mod
);
1379 #define TEX_PROJ 0x1
1380 #define TEX_BIAS 0x2
1382 #define TEX_DERIV 0x10
1384 static void shader_hw_sample(const struct wined3d_shader_instruction
*ins
, DWORD sampler_idx
,
1385 const char *dst_str
, const char *coord_reg
, WORD flags
, const char *dsx
, const char *dsy
)
1387 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1388 DWORD sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[sampler_idx
];
1389 const char *tex_type
;
1390 BOOL np2_fixup
= FALSE
;
1391 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1393 BOOL pshader
= shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
1394 const struct wined3d_shader
*shader
;
1395 const struct wined3d_device
*device
;
1396 const struct wined3d_gl_info
*gl_info
;
1398 /* D3D vertex shader sampler IDs are vertex samplers(0-3), not global d3d samplers */
1399 if(!pshader
) sampler_idx
+= MAX_FRAGMENT_SAMPLERS
;
1401 switch(sampler_type
) {
1407 shader
= ins
->ctx
->shader
;
1408 device
= shader
->device
;
1409 gl_info
= &device
->adapter
->gl_info
;
1411 if (pshader
&& priv
->cur_ps_args
->super
.np2_fixup
& (1 << sampler_idx
)
1412 && gl_info
->supported
[ARB_TEXTURE_RECTANGLE
])
1416 if (shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
))
1418 if (priv
->cur_np2fixup_info
->super
.active
& (1 << sampler_idx
))
1420 if (flags
) FIXME("Only ordinary sampling from NP2 textures is supported.\n");
1421 else np2_fixup
= TRUE
;
1426 case WINED3DSTT_VOLUME
:
1430 case WINED3DSTT_CUBE
:
1435 ERR("Unexpected texture type %d\n", sampler_type
);
1439 /* TEX, TXL, TXD and TXP do not support the "H" modifier,
1440 * so don't use shader_arb_get_modifier
1442 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
) mod
= "_SAT";
1445 /* Fragment samplers always have indentity mapping */
1446 if(sampler_idx
>= MAX_FRAGMENT_SAMPLERS
)
1448 sampler_idx
= priv
->cur_vs_args
->vertex
.samplers
[sampler_idx
- MAX_FRAGMENT_SAMPLERS
];
1451 if (flags
& TEX_DERIV
)
1453 if(flags
& TEX_PROJ
) FIXME("Projected texture sampling with custom derivatives\n");
1454 if(flags
& TEX_BIAS
) FIXME("Biased texture sampling with custom derivatives\n");
1455 shader_addline(buffer
, "TXD%s %s, %s, %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
,
1456 dsx
, dsy
,sampler_idx
, tex_type
);
1458 else if(flags
& TEX_LOD
)
1460 if(flags
& TEX_PROJ
) FIXME("Projected texture sampling with explicit lod\n");
1461 if(flags
& TEX_BIAS
) FIXME("Biased texture sampling with explicit lod\n");
1462 shader_addline(buffer
, "TXL%s %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
,
1463 sampler_idx
, tex_type
);
1465 else if (flags
& TEX_BIAS
)
1467 /* Shouldn't be possible, but let's check for it */
1468 if(flags
& TEX_PROJ
) FIXME("Biased and Projected texture sampling\n");
1469 /* TXB takes the 4th component of the source vector automatically, as d3d. Nothing more to do */
1470 shader_addline(buffer
, "TXB%s %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
, sampler_idx
, tex_type
);
1472 else if (flags
& TEX_PROJ
)
1474 shader_addline(buffer
, "TXP%s %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
, sampler_idx
, tex_type
);
1480 const unsigned char idx
= priv
->cur_np2fixup_info
->super
.idx
[sampler_idx
];
1481 shader_addline(buffer
, "MUL TA, np2fixup[%u].%s, %s;\n", idx
>> 1,
1482 (idx
% 2) ? "zwxy" : "xyzw", coord_reg
);
1484 shader_addline(buffer
, "TEX%s %s, TA, texture[%u], %s;\n", mod
, dst_str
, sampler_idx
, tex_type
);
1487 shader_addline(buffer
, "TEX%s %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
, sampler_idx
, tex_type
);
1492 gen_color_correction(buffer
, dst_str
, ins
->dst
[0].write_mask
,
1493 arb_get_helper_value(WINED3D_SHADER_TYPE_PIXEL
, ARB_ONE
),
1494 arb_get_helper_value(WINED3D_SHADER_TYPE_PIXEL
, ARB_TWO
),
1495 priv
->cur_ps_args
->super
.color_fixup
[sampler_idx
]);
1499 static void shader_arb_get_src_param(const struct wined3d_shader_instruction
*ins
,
1500 const struct wined3d_shader_src_param
*src
, unsigned int tmpreg
, char *outregstr
)
1502 /* Generate a line that does the input modifier computation and return the input register to use */
1503 BOOL is_color
= FALSE
, insert_line
;
1506 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1507 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1508 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
1509 const char *two
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_TWO
);
1511 /* Assume a new line will be added */
1514 /* Get register name */
1515 shader_arb_get_register_name(ins
, &src
->reg
, regstr
, &is_color
);
1516 shader_arb_get_swizzle(src
, is_color
, swzstr
);
1518 switch (src
->modifiers
)
1520 case WINED3DSPSM_NONE
:
1521 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
1522 insert_line
= FALSE
;
1524 case WINED3DSPSM_NEG
:
1525 sprintf(outregstr
, "-%s%s", regstr
, swzstr
);
1526 insert_line
= FALSE
;
1528 case WINED3DSPSM_BIAS
:
1529 shader_addline(buffer
, "ADD T%c, %s, -coefdiv.x;\n", 'A' + tmpreg
, regstr
);
1531 case WINED3DSPSM_BIASNEG
:
1532 shader_addline(buffer
, "ADD T%c, -%s, coefdiv.x;\n", 'A' + tmpreg
, regstr
);
1534 case WINED3DSPSM_SIGN
:
1535 shader_addline(buffer
, "MAD T%c, %s, %s, -%s;\n", 'A' + tmpreg
, regstr
, two
, one
);
1537 case WINED3DSPSM_SIGNNEG
:
1538 shader_addline(buffer
, "MAD T%c, %s, -%s, %s;\n", 'A' + tmpreg
, regstr
, two
, one
);
1540 case WINED3DSPSM_COMP
:
1541 shader_addline(buffer
, "SUB T%c, %s, %s;\n", 'A' + tmpreg
, one
, regstr
);
1543 case WINED3DSPSM_X2
:
1544 shader_addline(buffer
, "ADD T%c, %s, %s;\n", 'A' + tmpreg
, regstr
, regstr
);
1546 case WINED3DSPSM_X2NEG
:
1547 shader_addline(buffer
, "ADD T%c, -%s, -%s;\n", 'A' + tmpreg
, regstr
, regstr
);
1549 case WINED3DSPSM_DZ
:
1550 shader_addline(buffer
, "RCP T%c, %s.z;\n", 'A' + tmpreg
, regstr
);
1551 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
1553 case WINED3DSPSM_DW
:
1554 shader_addline(buffer
, "RCP T%c, %s.w;\n", 'A' + tmpreg
, regstr
);
1555 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
1557 case WINED3DSPSM_ABS
:
1558 if(ctx
->target_version
>= NV2
) {
1559 sprintf(outregstr
, "|%s%s|", regstr
, swzstr
);
1560 insert_line
= FALSE
;
1562 shader_addline(buffer
, "ABS T%c, %s;\n", 'A' + tmpreg
, regstr
);
1565 case WINED3DSPSM_ABSNEG
:
1566 if(ctx
->target_version
>= NV2
) {
1567 sprintf(outregstr
, "-|%s%s|", regstr
, swzstr
);
1569 shader_addline(buffer
, "ABS T%c, %s;\n", 'A' + tmpreg
, regstr
);
1570 sprintf(outregstr
, "-T%c%s", 'A' + tmpreg
, swzstr
);
1572 insert_line
= FALSE
;
1575 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
1576 insert_line
= FALSE
;
1579 /* Return modified or original register, with swizzle */
1581 sprintf(outregstr
, "T%c%s", 'A' + tmpreg
, swzstr
);
1584 static void pshader_hw_bem(const struct wined3d_shader_instruction
*ins
)
1586 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1587 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1588 DWORD sampler_code
= dst
->reg
.idx
[0].offset
;
1590 char src_name
[2][50];
1592 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1594 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed
1596 * Keep in mind that src_name[1] can be "TB" and src_name[0] can be "TA" because modifiers like _x2 are valid
1597 * with bem. So delay loading the first parameter until after the perturbation calculation which needs two
1600 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1601 shader_addline(buffer
, "SWZ TA, bumpenvmat%d, x, z, 0, 0;\n", sampler_code
);
1602 shader_addline(buffer
, "DP3 TC.r, TA, %s;\n", src_name
[1]);
1603 shader_addline(buffer
, "SWZ TA, bumpenvmat%d, y, w, 0, 0;\n", sampler_code
);
1604 shader_addline(buffer
, "DP3 TC.g, TA, %s;\n", src_name
[1]);
1606 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1607 shader_addline(buffer
, "ADD %s, %s, TC;\n", dst_name
, src_name
[0]);
1610 static DWORD
negate_modifiers(DWORD mod
, char *extra_char
)
1615 case WINED3DSPSM_NONE
: return WINED3DSPSM_NEG
;
1616 case WINED3DSPSM_NEG
: return WINED3DSPSM_NONE
;
1617 case WINED3DSPSM_BIAS
: return WINED3DSPSM_BIASNEG
;
1618 case WINED3DSPSM_BIASNEG
: return WINED3DSPSM_BIAS
;
1619 case WINED3DSPSM_SIGN
: return WINED3DSPSM_SIGNNEG
;
1620 case WINED3DSPSM_SIGNNEG
: return WINED3DSPSM_SIGN
;
1621 case WINED3DSPSM_COMP
: *extra_char
= '-'; return WINED3DSPSM_COMP
;
1622 case WINED3DSPSM_X2
: return WINED3DSPSM_X2NEG
;
1623 case WINED3DSPSM_X2NEG
: return WINED3DSPSM_X2
;
1624 case WINED3DSPSM_DZ
: *extra_char
= '-'; return WINED3DSPSM_DZ
;
1625 case WINED3DSPSM_DW
: *extra_char
= '-'; return WINED3DSPSM_DW
;
1626 case WINED3DSPSM_ABS
: return WINED3DSPSM_ABSNEG
;
1627 case WINED3DSPSM_ABSNEG
: return WINED3DSPSM_ABS
;
1629 FIXME("Unknown modifier %u\n", mod
);
1633 static void pshader_hw_cnd(const struct wined3d_shader_instruction
*ins
)
1635 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1636 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1638 char src_name
[3][50];
1639 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1640 ins
->ctx
->reg_maps
->shader_version
.minor
);
1642 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1643 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1645 if (shader_version
<= WINED3D_SHADER_VERSION(1, 3) && ins
->coissue
1646 && ins
->dst
->write_mask
!= WINED3DSP_WRITEMASK_3
)
1648 shader_addline(buffer
, "MOV%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
[1]);
1652 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
1655 /* src0 may have a negate srcmod set, so we can't blindly add "-" to the name */
1656 src0_copy
.modifiers
= negate_modifiers(src0_copy
.modifiers
, &extra_neg
);
1658 shader_arb_get_src_param(ins
, &src0_copy
, 0, src_name
[0]);
1659 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1660 shader_addline(buffer
, "ADD TA, %c%s, coefdiv.x;\n", extra_neg
, src_name
[0]);
1661 shader_addline(buffer
, "CMP%s %s, TA, %s, %s;\n", shader_arb_get_modifier(ins
),
1662 dst_name
, src_name
[1], src_name
[2]);
1666 static void pshader_hw_cmp(const struct wined3d_shader_instruction
*ins
)
1668 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1669 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1671 char src_name
[3][50];
1673 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1675 /* Generate input register names (with modifiers) */
1676 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1677 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1678 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1680 shader_addline(buffer
, "CMP%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1681 dst_name
, src_name
[0], src_name
[2], src_name
[1]);
1684 /** Process the WINED3DSIO_DP2ADD instruction in ARB.
1685 * dst = dot2(src0, src1) + src2 */
1686 static void pshader_hw_dp2add(const struct wined3d_shader_instruction
*ins
)
1688 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1689 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1691 char src_name
[3][50];
1692 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1694 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1695 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1696 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1698 if(ctx
->target_version
>= NV3
)
1700 /* GL_NV_fragment_program2 has a 1:1 matching instruction */
1701 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1702 shader_addline(buffer
, "DP2A%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1703 dst_name
, src_name
[0], src_name
[1], src_name
[2]);
1705 else if(ctx
->target_version
>= NV2
)
1707 /* dst.x = src2.?, src0.x, src1.x + src0.y * src1.y
1708 * dst.y = src2.?, src0.x, src1.z + src0.y * src1.w
1709 * dst.z = src2.?, src0.x, src1.x + src0.y * src1.y
1710 * dst.z = src2.?, src0.x, src1.z + src0.y * src1.w
1712 * Make sure that src1.zw = src1.xy, then we get a classic dp2add
1714 * .xyxy and other swizzles that we could get with this are not valid in
1715 * plain ARBfp, but luckily the NV extension grammar lifts this limitation.
1717 struct wined3d_shader_src_param tmp_param
= ins
->src
[1];
1718 DWORD swizzle
= tmp_param
.swizzle
& 0xf; /* Selects .xy */
1719 tmp_param
.swizzle
= swizzle
| (swizzle
<< 4); /* Creates .xyxy */
1721 shader_arb_get_src_param(ins
, &tmp_param
, 1, src_name
[1]);
1723 shader_addline(buffer
, "X2D%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1724 dst_name
, src_name
[2], src_name
[0], src_name
[1]);
1728 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1729 /* 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]
1730 * src_name[0] can be TA, but TA is a private temp for modifiers, so it is save to overwrite
1732 shader_addline(buffer
, "MOV TA, %s;\n", src_name
[0]);
1733 shader_addline(buffer
, "MOV TA.z, 0.0;\n");
1734 shader_addline(buffer
, "DP3 TA, TA, %s;\n", src_name
[1]);
1735 shader_addline(buffer
, "ADD%s %s, TA, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
[2]);
1739 /* Map the opcode 1-to-1 to the GL code */
1740 static void shader_hw_map2gl(const struct wined3d_shader_instruction
*ins
)
1742 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1743 const char *instruction
;
1744 char arguments
[256], dst_str
[50];
1746 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1748 switch (ins
->handler_idx
)
1750 case WINED3DSIH_ABS
: instruction
= "ABS"; break;
1751 case WINED3DSIH_ADD
: instruction
= "ADD"; break;
1752 case WINED3DSIH_CRS
: instruction
= "XPD"; break;
1753 case WINED3DSIH_DP3
: instruction
= "DP3"; break;
1754 case WINED3DSIH_DP4
: instruction
= "DP4"; break;
1755 case WINED3DSIH_DST
: instruction
= "DST"; break;
1756 case WINED3DSIH_FRC
: instruction
= "FRC"; break;
1757 case WINED3DSIH_LIT
: instruction
= "LIT"; break;
1758 case WINED3DSIH_LRP
: instruction
= "LRP"; break;
1759 case WINED3DSIH_MAD
: instruction
= "MAD"; break;
1760 case WINED3DSIH_MAX
: instruction
= "MAX"; break;
1761 case WINED3DSIH_MIN
: instruction
= "MIN"; break;
1762 case WINED3DSIH_MOV
: instruction
= "MOV"; break;
1763 case WINED3DSIH_MUL
: instruction
= "MUL"; break;
1764 case WINED3DSIH_SGE
: instruction
= "SGE"; break;
1765 case WINED3DSIH_SLT
: instruction
= "SLT"; break;
1766 case WINED3DSIH_SUB
: instruction
= "SUB"; break;
1767 case WINED3DSIH_MOVA
:instruction
= "ARR"; break;
1768 case WINED3DSIH_DSX
: instruction
= "DDX"; break;
1769 default: instruction
= "";
1770 FIXME("Unhandled opcode %#x\n", ins
->handler_idx
);
1774 /* Note that shader_arb_add_dst_param() adds spaces. */
1775 arguments
[0] = '\0';
1776 shader_arb_get_dst_param(ins
, dst
, dst_str
);
1777 for (i
= 0; i
< ins
->src_count
; ++i
)
1780 strcat(arguments
, ", ");
1781 shader_arb_get_src_param(ins
, &ins
->src
[i
], i
, operand
);
1782 strcat(arguments
, operand
);
1784 shader_addline(buffer
, "%s%s %s%s;\n", instruction
, shader_arb_get_modifier(ins
), dst_str
, arguments
);
1787 static void shader_hw_nop(const struct wined3d_shader_instruction
*ins
) {}
1789 static DWORD
shader_arb_select_component(DWORD swizzle
, DWORD component
)
1791 return ((swizzle
>> 2 * component
) & 0x3) * 0x55;
1794 static void shader_hw_mov(const struct wined3d_shader_instruction
*ins
)
1796 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
1797 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
1798 BOOL pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
1799 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1800 const char *zero
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ZERO
);
1801 const char *one
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ONE
);
1802 const char *two
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_TWO
);
1804 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1805 char src0_param
[256];
1807 if (ins
->handler_idx
== WINED3DSIH_MOVA
)
1809 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1811 const char *offset
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_VS_REL_OFFSET
);
1813 if(ctx
->target_version
>= NV2
) {
1814 shader_hw_map2gl(ins
);
1817 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_param
);
1818 shader_arb_get_write_mask(ins
, &ins
->dst
[0], write_mask
);
1820 /* This implements the mova formula used in GLSL. The first two instructions
1821 * prepare the sign() part. Note that it is fine to have my_sign(0.0) = 1.0
1825 * A0.x = arl(floor(abs(0.0) + 0.5) * 1.0) = floor(0.5) = 0.0 since arl does a floor
1827 * The ARL is performed when A0 is used - the requested component is read from A0_SHADOW into
1828 * A0.x. We can use the overwritten component of A0_shadow as temporary storage for the sign.
1830 shader_addline(buffer
, "SGE A0_SHADOW%s, %s, %s;\n", write_mask
, src0_param
, zero
);
1831 shader_addline(buffer
, "MAD A0_SHADOW%s, A0_SHADOW, %s, -%s;\n", write_mask
, two
, one
);
1833 shader_addline(buffer
, "ABS TA%s, %s;\n", write_mask
, src0_param
);
1834 shader_addline(buffer
, "ADD TA%s, TA, rel_addr_const.x;\n", write_mask
);
1835 shader_addline(buffer
, "FLR TA%s, TA;\n", write_mask
);
1836 if (shader_data
->rel_offset
)
1838 shader_addline(buffer
, "ADD TA%s, TA, %s;\n", write_mask
, offset
);
1840 shader_addline(buffer
, "MUL A0_SHADOW%s, TA, A0_SHADOW;\n", write_mask
);
1842 ((struct shader_arb_ctx_priv
*)ins
->ctx
->backend_data
)->addr_reg
[0] = '\0';
1844 else if (reg_maps
->shader_version
.major
== 1
1845 && !shader_is_pshader_version(reg_maps
->shader_version
.type
)
1846 && ins
->dst
[0].reg
.type
== WINED3DSPR_ADDR
)
1848 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1849 src0_param
[0] = '\0';
1851 if (shader_data
->rel_offset
&& ctx
->target_version
== ARB
)
1853 const char *offset
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_VS_REL_OFFSET
);
1854 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_param
);
1855 shader_addline(buffer
, "ADD TA.x, %s, %s;\n", src0_param
, offset
);
1856 shader_addline(buffer
, "ARL A0.x, TA.x;\n");
1860 /* Apple's ARB_vertex_program implementation does not accept an ARL source argument
1861 * with more than one component. Thus replicate the first source argument over all
1862 * 4 components. For example, .xyzw -> .x (or better: .xxxx), .zwxy -> .z, etc) */
1863 struct wined3d_shader_src_param tmp_src
= ins
->src
[0];
1864 tmp_src
.swizzle
= shader_arb_select_component(tmp_src
.swizzle
, 0);
1865 shader_arb_get_src_param(ins
, &tmp_src
, 0, src0_param
);
1866 shader_addline(buffer
, "ARL A0.x, %s;\n", src0_param
);
1869 else if (ins
->dst
[0].reg
.type
== WINED3DSPR_COLOROUT
&& !ins
->dst
[0].reg
.idx
[0].offset
&& pshader
)
1871 if (ctx
->ps_post_process
&& shader
->u
.ps
.color0_mov
)
1873 shader_addline(buffer
, "#mov handled in srgb write or fog code\n");
1876 shader_hw_map2gl(ins
);
1880 shader_hw_map2gl(ins
);
1884 static void pshader_hw_texkill(const struct wined3d_shader_instruction
*ins
)
1886 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1887 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1890 /* No swizzles are allowed in d3d's texkill. PS 1.x ignores the 4th component as documented,
1891 * but >= 2.0 honors it (undocumented, but tested by the d3d9 testsuite)
1893 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1895 if (ins
->ctx
->reg_maps
->shader_version
.major
>= 2)
1897 const char *kilsrc
= "TA";
1900 shader_arb_get_register_name(ins
, &dst
->reg
, reg_dest
, &is_color
);
1901 if(dst
->write_mask
== WINED3DSP_WRITEMASK_ALL
)
1907 /* Sigh. KIL doesn't support swizzles/writemasks. KIL passes a writemask, but ".xy" for example
1908 * is not valid as a swizzle in ARB (needs ".xyyy"). Use SWZ to load the register properly, and set
1909 * masked out components to 0(won't kill)
1911 char x
= '0', y
= '0', z
= '0', w
= '0';
1912 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
) x
= 'x';
1913 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
) y
= 'y';
1914 if(dst
->write_mask
& WINED3DSP_WRITEMASK_2
) z
= 'z';
1915 if(dst
->write_mask
& WINED3DSP_WRITEMASK_3
) w
= 'w';
1916 shader_addline(buffer
, "SWZ TA, %s, %c, %c, %c, %c;\n", reg_dest
, x
, y
, z
, w
);
1918 shader_addline(buffer
, "KIL %s;\n", kilsrc
);
1922 /* ARB fp doesn't like swizzles on the parameter of the KIL instruction. To mask the 4th component,
1923 * copy the register into our general purpose TMP variable, overwrite .w and pass TMP to KIL
1925 * ps_1_3 shaders use the texcoord incarnation of the Tx register. ps_1_4 shaders can use the same,
1926 * or pass in any temporary register(in shader phase 2)
1928 if (ins
->ctx
->reg_maps
->shader_version
.minor
<= 3)
1929 sprintf(reg_dest
, "fragment.texcoord[%u]", dst
->reg
.idx
[0].offset
);
1931 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1932 shader_addline(buffer
, "SWZ TA, %s, x, y, z, 1;\n", reg_dest
);
1933 shader_addline(buffer
, "KIL TA;\n");
1937 static void pshader_hw_tex(const struct wined3d_shader_instruction
*ins
)
1939 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1940 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1941 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1942 ins
->ctx
->reg_maps
->shader_version
.minor
);
1943 struct wined3d_shader_src_param src
;
1947 DWORD reg_sampler_code
;
1949 BOOL swizzle_coord
= FALSE
;
1951 /* All versions have a destination register */
1952 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1954 /* 1.0-1.4: Use destination register number as texture code.
1955 2.0+: Use provided sampler number as texure code. */
1956 if (shader_version
< WINED3D_SHADER_VERSION(2,0))
1957 reg_sampler_code
= dst
->reg
.idx
[0].offset
;
1959 reg_sampler_code
= ins
->src
[1].reg
.idx
[0].offset
;
1961 /* 1.0-1.3: Use the texcoord varying.
1962 1.4+: Use provided coordinate source register. */
1963 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
1964 sprintf(reg_coord
, "fragment.texcoord[%u]", reg_sampler_code
);
1966 /* TEX is the only instruction that can handle DW and DZ natively */
1968 if(src
.modifiers
== WINED3DSPSM_DW
) src
.modifiers
= WINED3DSPSM_NONE
;
1969 if(src
.modifiers
== WINED3DSPSM_DZ
) src
.modifiers
= WINED3DSPSM_NONE
;
1970 shader_arb_get_src_param(ins
, &src
, 0, reg_coord
);
1974 * 1.1, 1.2, 1.3: Use WINED3D_TSS_TEXTURETRANSFORMFLAGS
1975 * 1.4: Use WINED3DSPSM_DZ or WINED3DSPSM_DW on src[0]
1976 * 2.0+: Use WINED3DSI_TEXLD_PROJECT on the opcode
1978 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
1981 if (reg_sampler_code
< MAX_TEXTURES
)
1982 flags
= priv
->cur_ps_args
->super
.tex_transform
>> reg_sampler_code
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
;
1983 if (flags
& WINED3D_PSARGS_PROJECTED
)
1985 myflags
|= TEX_PROJ
;
1986 if ((flags
& ~WINED3D_PSARGS_PROJECTED
) == WINED3D_TTFF_COUNT3
)
1987 swizzle_coord
= TRUE
;
1990 else if (shader_version
< WINED3D_SHADER_VERSION(2,0))
1992 enum wined3d_shader_src_modifier src_mod
= ins
->src
[0].modifiers
;
1993 if (src_mod
== WINED3DSPSM_DZ
)
1995 swizzle_coord
= TRUE
;
1996 myflags
|= TEX_PROJ
;
1997 } else if(src_mod
== WINED3DSPSM_DW
) {
1998 myflags
|= TEX_PROJ
;
2001 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) myflags
|= TEX_PROJ
;
2002 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) myflags
|= TEX_BIAS
;
2007 /* TXP cannot handle DZ natively, so move the z coordinate to .w.
2008 * reg_coord is a read-only varying register, so we need a temp reg */
2009 shader_addline(ins
->ctx
->buffer
, "SWZ TA, %s, x, y, z, z;\n", reg_coord
);
2010 strcpy(reg_coord
, "TA");
2013 shader_hw_sample(ins
, reg_sampler_code
, reg_dest
, reg_coord
, myflags
, NULL
, NULL
);
2016 static void pshader_hw_texcoord(const struct wined3d_shader_instruction
*ins
)
2018 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2019 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2020 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
2021 ins
->ctx
->reg_maps
->shader_version
.minor
);
2024 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
2026 DWORD reg
= dst
->reg
.idx
[0].offset
;
2028 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2029 shader_addline(buffer
, "MOV_SAT %s, fragment.texcoord[%u];\n", dst_str
, reg
);
2033 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_src
);
2034 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2035 shader_addline(buffer
, "MOV %s, %s;\n", dst_str
, reg_src
);
2039 static void pshader_hw_texreg2ar(const struct wined3d_shader_instruction
*ins
)
2041 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2044 DWORD reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2048 /* Note that texreg2ar treats Tx as a temporary register, not as a varying */
2049 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2050 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2051 /* Move .x first in case src_str is "TA" */
2052 shader_addline(buffer
, "MOV TA.y, %s.x;\n", src_str
);
2053 shader_addline(buffer
, "MOV TA.x, %s.w;\n", src_str
);
2054 if (reg1
< MAX_TEXTURES
)
2056 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2057 flags
= priv
->cur_ps_args
->super
.tex_transform
>> reg1
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
;
2059 shader_hw_sample(ins
, reg1
, dst_str
, "TA", flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2062 static void pshader_hw_texreg2gb(const struct wined3d_shader_instruction
*ins
)
2064 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2066 DWORD reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2070 /* Note that texreg2gb treats Tx as a temporary register, not as a varying */
2071 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2072 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2073 shader_addline(buffer
, "MOV TA.x, %s.y;\n", src_str
);
2074 shader_addline(buffer
, "MOV TA.y, %s.z;\n", src_str
);
2075 shader_hw_sample(ins
, reg1
, dst_str
, "TA", 0, NULL
, NULL
);
2078 static void pshader_hw_texreg2rgb(const struct wined3d_shader_instruction
*ins
)
2080 DWORD reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2084 /* Note that texreg2rg treats Tx as a temporary register, not as a varying */
2085 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2086 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2087 shader_hw_sample(ins
, reg1
, dst_str
, src_str
, 0, NULL
, NULL
);
2090 static void pshader_hw_texbem(const struct wined3d_shader_instruction
*ins
)
2092 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2093 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2094 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2095 char reg_coord
[40], dst_reg
[50], src_reg
[50];
2096 DWORD reg_dest_code
;
2098 /* All versions have a destination register. The Tx where the texture coordinates come
2099 * from is the varying incarnation of the texture register
2101 reg_dest_code
= dst
->reg
.idx
[0].offset
;
2102 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_reg
);
2103 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_reg
);
2104 sprintf(reg_coord
, "fragment.texcoord[%u]", reg_dest_code
);
2106 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed
2107 * The Tx in which the perturbation map is stored is the tempreg incarnation of the texture register
2109 * GL_NV_fragment_program_option could handle this in one instruction via X2D:
2110 * X2D TA.xy, fragment.texcoord, T%u, bumpenvmat%u.xzyw
2112 * However, the NV extensions are never enabled for <= 2.0 shaders because of the performance penalty that
2113 * comes with it, and texbem is an 1.x only instruction. No 1.x instruction forces us to enable the NV
2116 shader_addline(buffer
, "SWZ TB, bumpenvmat%d, x, z, 0, 0;\n", reg_dest_code
);
2117 shader_addline(buffer
, "DP3 TA.x, TB, %s;\n", src_reg
);
2118 shader_addline(buffer
, "SWZ TB, bumpenvmat%d, y, w, 0, 0;\n", reg_dest_code
);
2119 shader_addline(buffer
, "DP3 TA.y, TB, %s;\n", src_reg
);
2121 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2122 * so we can't let the GL handle this.
2124 if ((priv
->cur_ps_args
->super
.tex_transform
>> reg_dest_code
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
)
2125 & WINED3D_PSARGS_PROJECTED
)
2127 shader_addline(buffer
, "RCP TB.w, %s.w;\n", reg_coord
);
2128 shader_addline(buffer
, "MUL TB.xy, %s, TB.w;\n", reg_coord
);
2129 shader_addline(buffer
, "ADD TA.xy, TA, TB;\n");
2131 shader_addline(buffer
, "ADD TA.xy, TA, %s;\n", reg_coord
);
2134 shader_hw_sample(ins
, reg_dest_code
, dst_reg
, "TA", 0, NULL
, NULL
);
2136 if (ins
->handler_idx
== WINED3DSIH_TEXBEML
)
2138 /* No src swizzles are allowed, so this is ok */
2139 shader_addline(buffer
, "MAD TA, %s.z, luminance%d.x, luminance%d.y;\n",
2140 src_reg
, reg_dest_code
, reg_dest_code
);
2141 shader_addline(buffer
, "MUL %s, %s, TA;\n", dst_reg
, dst_reg
);
2145 static void pshader_hw_texm3x2pad(const struct wined3d_shader_instruction
*ins
)
2147 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2148 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2149 char src0_name
[50], dst_name
[50];
2151 struct wined3d_shader_register tmp_reg
= ins
->dst
[0].reg
;
2153 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2154 /* The next instruction will be a texm3x2tex or texm3x2depth that writes to the uninitialized
2155 * T<reg+1> register. Use this register to store the calculated vector
2157 tmp_reg
.idx
[0].offset
= reg
+ 1;
2158 shader_arb_get_register_name(ins
, &tmp_reg
, dst_name
, &is_color
);
2159 shader_addline(buffer
, "DP3 %s.x, fragment.texcoord[%u], %s;\n", dst_name
, reg
, src0_name
);
2162 static void pshader_hw_texm3x2tex(const struct wined3d_shader_instruction
*ins
)
2164 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2166 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2167 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2173 /* We know that we're writing to the uninitialized T<reg> register, so use it for temporary storage */
2174 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2176 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2177 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2178 shader_addline(buffer
, "DP3 %s.y, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2179 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2180 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2183 static void pshader_hw_texm3x3pad(const struct wined3d_shader_instruction
*ins
)
2185 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2186 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2187 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2188 char src0_name
[50], dst_name
[50];
2189 struct wined3d_shader_register tmp_reg
= ins
->dst
[0].reg
;
2192 /* There are always 2 texm3x3pad instructions followed by one texm3x3[tex,vspec, ...] instruction, with
2193 * incrementing ins->dst[0].register_idx numbers. So the pad instruction already knows the final destination
2194 * register, and this register is uninitialized(otherwise the assembler complains that it is 'redeclared')
2196 tmp_reg
.idx
[0].offset
= reg
+ 2 - tex_mx
->current_row
;
2197 shader_arb_get_register_name(ins
, &tmp_reg
, dst_name
, &is_color
);
2199 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2200 shader_addline(buffer
, "DP3 %s.%c, fragment.texcoord[%u], %s;\n",
2201 dst_name
, 'x' + tex_mx
->current_row
, reg
, src0_name
);
2202 tex_mx
->texcoord_w
[tex_mx
->current_row
++] = reg
;
2205 static void pshader_hw_texm3x3tex(const struct wined3d_shader_instruction
*ins
)
2207 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2208 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2210 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2211 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2213 char src0_name
[50], dst_name
[50];
2216 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2217 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2218 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_name
, reg
, src0_name
);
2220 /* Sample the texture using the calculated coordinates */
2221 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2222 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2223 shader_hw_sample(ins
, reg
, dst_str
, dst_name
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2224 tex_mx
->current_row
= 0;
2227 static void pshader_hw_texm3x3vspec(const struct wined3d_shader_instruction
*ins
)
2229 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2230 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2232 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2233 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2239 /* Get the dst reg without writemask strings. We know this register is uninitialized, so we can use all
2240 * components for temporary data storage
2242 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2243 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2244 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2246 /* Construct the eye-ray vector from w coordinates */
2247 shader_addline(buffer
, "MOV TB.x, fragment.texcoord[%u].w;\n", tex_mx
->texcoord_w
[0]);
2248 shader_addline(buffer
, "MOV TB.y, fragment.texcoord[%u].w;\n", tex_mx
->texcoord_w
[1]);
2249 shader_addline(buffer
, "MOV TB.z, fragment.texcoord[%u].w;\n", reg
);
2251 /* Calculate reflection vector
2253 shader_addline(buffer
, "DP3 %s.w, %s, TB;\n", dst_reg
, dst_reg
);
2254 /* The .w is ignored when sampling, so I can use TB.w to calculate dot(N, N) */
2255 shader_addline(buffer
, "DP3 TB.w, %s, %s;\n", dst_reg
, dst_reg
);
2256 shader_addline(buffer
, "RCP TB.w, TB.w;\n");
2257 shader_addline(buffer
, "MUL %s.w, %s.w, TB.w;\n", dst_reg
, dst_reg
);
2258 shader_addline(buffer
, "MUL %s, %s.w, %s;\n", dst_reg
, dst_reg
, dst_reg
);
2259 shader_addline(buffer
, "MAD %s, coefmul.x, %s, -TB;\n", dst_reg
, dst_reg
);
2261 /* Sample the texture using the calculated coordinates */
2262 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2263 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2264 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2265 tex_mx
->current_row
= 0;
2268 static void pshader_hw_texm3x3spec(const struct wined3d_shader_instruction
*ins
)
2270 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2271 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2273 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2274 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2281 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2282 shader_arb_get_src_param(ins
, &ins
->src
[0], 1, src1_name
);
2283 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2284 /* Note: dst_reg.xy is input here, generated by two texm3x3pad instructions */
2285 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2287 /* Calculate reflection vector.
2290 * dst_reg.xyz = 2 * --------- * N - E
2293 * Which normalizes the normal vector
2295 shader_addline(buffer
, "DP3 %s.w, %s, %s;\n", dst_reg
, dst_reg
, src1_name
);
2296 shader_addline(buffer
, "DP3 TC.w, %s, %s;\n", dst_reg
, dst_reg
);
2297 shader_addline(buffer
, "RCP TC.w, TC.w;\n");
2298 shader_addline(buffer
, "MUL %s.w, %s.w, TC.w;\n", dst_reg
, dst_reg
);
2299 shader_addline(buffer
, "MUL %s, %s.w, %s;\n", dst_reg
, dst_reg
, dst_reg
);
2300 shader_addline(buffer
, "MAD %s, coefmul.x, %s, -%s;\n", dst_reg
, dst_reg
, src1_name
);
2302 /* Sample the texture using the calculated coordinates */
2303 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2304 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2305 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2306 tex_mx
->current_row
= 0;
2309 static void pshader_hw_texdepth(const struct wined3d_shader_instruction
*ins
)
2311 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2312 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2314 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2315 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2317 /* texdepth has an implicit destination, the fragment depth value. It's only parameter,
2318 * which is essentially an input, is the destination register because it is the first
2319 * parameter. According to the msdn, this must be register r5, but let's keep it more flexible
2320 * here(writemasks/swizzles are not valid on texdepth)
2322 shader_arb_get_dst_param(ins
, dst
, dst_name
);
2324 /* According to the msdn, the source register(must be r5) is unusable after
2325 * the texdepth instruction, so we're free to modify it
2327 shader_addline(buffer
, "MIN %s.y, %s.y, %s;\n", dst_name
, dst_name
, one
);
2329 /* How to deal with the special case dst_name.g == 0? if r != 0, then
2330 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
2331 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
2333 shader_addline(buffer
, "RCP %s.y, %s.y;\n", dst_name
, dst_name
);
2334 shader_addline(buffer
, "MUL TA.x, %s.x, %s.y;\n", dst_name
, dst_name
);
2335 shader_addline(buffer
, "MIN TA.x, TA.x, %s;\n", one
);
2336 shader_addline(buffer
, "MAX result.depth, TA.x, %s;\n", zero
);
2339 /** Process the WINED3DSIO_TEXDP3TEX instruction in ARB:
2340 * Take a 3-component dot product of the TexCoord[dstreg] and src,
2341 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
2342 static void pshader_hw_texdp3tex(const struct wined3d_shader_instruction
*ins
)
2344 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2345 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
[0].offset
;
2349 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2350 shader_addline(buffer
, "MOV TB, 0.0;\n");
2351 shader_addline(buffer
, "DP3 TB.x, fragment.texcoord[%u], %s;\n", sampler_idx
, src0
);
2353 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2354 shader_hw_sample(ins
, sampler_idx
, dst_str
, "TB", 0 /* Only one coord, can't be projected */, NULL
, NULL
);
2357 /** Process the WINED3DSIO_TEXDP3 instruction in ARB:
2358 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
2359 static void pshader_hw_texdp3(const struct wined3d_shader_instruction
*ins
)
2361 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2364 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2366 /* Handle output register */
2367 shader_arb_get_dst_param(ins
, dst
, dst_str
);
2368 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2369 shader_addline(buffer
, "DP3 %s, fragment.texcoord[%u], %s;\n", dst_str
, dst
->reg
.idx
[0].offset
, src0
);
2372 /** Process the WINED3DSIO_TEXM3X3 instruction in ARB
2373 * Perform the 3rd row of a 3x3 matrix multiply */
2374 static void pshader_hw_texm3x3(const struct wined3d_shader_instruction
*ins
)
2376 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2377 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2378 char dst_str
[50], dst_name
[50];
2382 shader_arb_get_dst_param(ins
, dst
, dst_str
);
2383 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2384 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2385 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_name
, dst
->reg
.idx
[0].offset
, src0
);
2386 shader_addline(buffer
, "MOV %s, %s;\n", dst_str
, dst_name
);
2389 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in ARB:
2390 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
2391 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
2392 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
2394 static void pshader_hw_texm3x2depth(const struct wined3d_shader_instruction
*ins
)
2396 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2397 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2398 char src0
[50], dst_name
[50];
2400 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2401 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2403 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2404 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2405 shader_addline(buffer
, "DP3 %s.y, fragment.texcoord[%u], %s;\n", dst_name
, dst
->reg
.idx
[0].offset
, src0
);
2407 /* How to deal with the special case dst_name.g == 0? if r != 0, then
2408 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
2409 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
2411 shader_addline(buffer
, "RCP %s.y, %s.y;\n", dst_name
, dst_name
);
2412 shader_addline(buffer
, "MUL %s.x, %s.x, %s.y;\n", dst_name
, dst_name
, dst_name
);
2413 shader_addline(buffer
, "MIN %s.x, %s.x, %s;\n", dst_name
, dst_name
, one
);
2414 shader_addline(buffer
, "MAX result.depth, %s.x, %s;\n", dst_name
, zero
);
2417 /** Handles transforming all WINED3DSIO_M?x? opcodes for
2418 Vertex/Pixel shaders to ARB_vertex_program codes */
2419 static void shader_hw_mnxn(const struct wined3d_shader_instruction
*ins
)
2422 int nComponents
= 0;
2423 struct wined3d_shader_dst_param tmp_dst
= {{0}};
2424 struct wined3d_shader_src_param tmp_src
[2] = {{{0}}};
2425 struct wined3d_shader_instruction tmp_ins
;
2427 memset(&tmp_ins
, 0, sizeof(tmp_ins
));
2429 /* Set constants for the temporary argument */
2430 tmp_ins
.ctx
= ins
->ctx
;
2431 tmp_ins
.dst_count
= 1;
2432 tmp_ins
.dst
= &tmp_dst
;
2433 tmp_ins
.src_count
= 2;
2434 tmp_ins
.src
= tmp_src
;
2436 switch(ins
->handler_idx
)
2438 case WINED3DSIH_M4x4
:
2440 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2442 case WINED3DSIH_M4x3
:
2444 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2446 case WINED3DSIH_M3x4
:
2448 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2450 case WINED3DSIH_M3x3
:
2452 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2454 case WINED3DSIH_M3x2
:
2456 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2459 FIXME("Unhandled opcode %#x\n", ins
->handler_idx
);
2463 tmp_dst
= ins
->dst
[0];
2464 tmp_src
[0] = ins
->src
[0];
2465 tmp_src
[1] = ins
->src
[1];
2466 for (i
= 0; i
< nComponents
; ++i
)
2468 tmp_dst
.write_mask
= WINED3DSP_WRITEMASK_0
<< i
;
2469 shader_hw_map2gl(&tmp_ins
);
2470 ++tmp_src
[1].reg
.idx
[0].offset
;
2474 static DWORD
abs_modifier(DWORD mod
, BOOL
*need_abs
)
2480 case WINED3DSPSM_NONE
: return WINED3DSPSM_ABS
;
2481 case WINED3DSPSM_NEG
: return WINED3DSPSM_ABS
;
2482 case WINED3DSPSM_BIAS
: *need_abs
= TRUE
; return WINED3DSPSM_BIAS
;
2483 case WINED3DSPSM_BIASNEG
: *need_abs
= TRUE
; return WINED3DSPSM_BIASNEG
;
2484 case WINED3DSPSM_SIGN
: *need_abs
= TRUE
; return WINED3DSPSM_SIGN
;
2485 case WINED3DSPSM_SIGNNEG
: *need_abs
= TRUE
; return WINED3DSPSM_SIGNNEG
;
2486 case WINED3DSPSM_COMP
: *need_abs
= TRUE
; return WINED3DSPSM_COMP
;
2487 case WINED3DSPSM_X2
: *need_abs
= TRUE
; return WINED3DSPSM_X2
;
2488 case WINED3DSPSM_X2NEG
: *need_abs
= TRUE
; return WINED3DSPSM_X2NEG
;
2489 case WINED3DSPSM_DZ
: *need_abs
= TRUE
; return WINED3DSPSM_DZ
;
2490 case WINED3DSPSM_DW
: *need_abs
= TRUE
; return WINED3DSPSM_DW
;
2491 case WINED3DSPSM_ABS
: return WINED3DSPSM_ABS
;
2492 case WINED3DSPSM_ABSNEG
: return WINED3DSPSM_ABS
;
2494 FIXME("Unknown modifier %u\n", mod
);
2498 static void shader_hw_scalar_op(const struct wined3d_shader_instruction
*ins
)
2500 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2501 const char *instruction
;
2502 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
2503 BOOL need_abs
= FALSE
;
2508 switch(ins
->handler_idx
)
2510 case WINED3DSIH_RSQ
: instruction
= "RSQ"; break;
2511 case WINED3DSIH_RCP
: instruction
= "RCP"; break;
2512 case WINED3DSIH_EXPP
:
2513 if (ins
->ctx
->reg_maps
->shader_version
.major
< 2)
2515 instruction
= "EXP";
2519 case WINED3DSIH_EXP
:
2520 instruction
= "EX2";
2522 case WINED3DSIH_LOG
:
2523 case WINED3DSIH_LOGP
:
2524 /* The precision requirements suggest that LOGP matches ARBvp's LOG
2525 * instruction, but notice that the output of those instructions is
2527 src0_copy
.modifiers
= abs_modifier(src0_copy
.modifiers
, &need_abs
);
2528 instruction
= "LG2";
2530 default: instruction
= "";
2531 FIXME("Unhandled opcode %#x\n", ins
->handler_idx
);
2535 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
2537 src0_copy
.swizzle
= shader_arb_select_component(src0_copy
.swizzle
, 3);
2539 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
); /* Destination */
2540 shader_arb_get_src_param(ins
, &src0_copy
, 0, src
);
2544 shader_addline(buffer
, "ABS TA.w, %s;\n", src
);
2545 shader_addline(buffer
, "%s%s %s, TA.w;\n", instruction
, shader_arb_get_modifier(ins
), dst
);
2549 shader_addline(buffer
, "%s%s %s, %s;\n", instruction
, shader_arb_get_modifier(ins
), dst
, src
);
2554 static void shader_hw_nrm(const struct wined3d_shader_instruction
*ins
)
2556 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2559 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2560 BOOL pshader
= shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2561 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2563 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2564 shader_arb_get_src_param(ins
, &ins
->src
[0], 1 /* Use TB */, src_name
);
2566 /* In D3D, NRM of a vector with length zero returns zero. Catch this situation, as
2567 * otherwise NRM or RSQ would return NaN */
2568 if(pshader
&& priv
->target_version
>= NV3
)
2570 /* GL_NV_fragment_program2's NRM needs protection against length zero vectors too
2572 * TODO: Find out if DP3+NRM+MOV is really faster than DP3+RSQ+MUL
2574 shader_addline(buffer
, "DP3C TA, %s, %s;\n", src_name
, src_name
);
2575 shader_addline(buffer
, "NRM%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
);
2576 shader_addline(buffer
, "MOV %s (EQ), %s;\n", dst_name
, zero
);
2578 else if(priv
->target_version
>= NV2
)
2580 shader_addline(buffer
, "DP3C TA.x, %s, %s;\n", src_name
, src_name
);
2581 shader_addline(buffer
, "RSQ TA.x (NE), TA.x;\n");
2582 shader_addline(buffer
, "MUL%s %s, %s, TA.x;\n", shader_arb_get_modifier(ins
), dst_name
,
2587 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2589 shader_addline(buffer
, "DP3 TA.x, %s, %s;\n", src_name
, src_name
);
2590 /* Pass any non-zero value to RSQ if the input vector has a length of zero. The
2591 * RSQ result doesn't matter, as long as multiplying it by 0 returns 0.
2593 shader_addline(buffer
, "SGE TA.y, -TA.x, %s;\n", zero
);
2594 shader_addline(buffer
, "MAD TA.x, %s, TA.y, TA.x;\n", one
);
2596 shader_addline(buffer
, "RSQ TA.x, TA.x;\n");
2597 /* dst.w = src[0].w * 1 / (src.x^2 + src.y^2 + src.z^2)^(1/2) according to msdn*/
2598 shader_addline(buffer
, "MUL%s %s, %s, TA.x;\n", shader_arb_get_modifier(ins
), dst_name
,
2603 static void shader_hw_lrp(const struct wined3d_shader_instruction
*ins
)
2605 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2607 char src_name
[3][50];
2609 /* ARB_fragment_program has a convenient LRP instruction */
2610 if(shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)) {
2611 shader_hw_map2gl(ins
);
2615 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2616 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
2617 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
2618 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
2620 shader_addline(buffer
, "SUB TA, %s, %s;\n", src_name
[1], src_name
[2]);
2621 shader_addline(buffer
, "MAD%s %s, %s, TA, %s;\n", shader_arb_get_modifier(ins
),
2622 dst_name
, src_name
[0], src_name
[2]);
2625 static void shader_hw_sincos(const struct wined3d_shader_instruction
*ins
)
2627 /* This instruction exists in ARB, but the d3d instruction takes two extra parameters which
2628 * must contain fixed constants. So we need a separate function to filter those constants and
2631 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2632 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2633 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2635 char src_name0
[50], src_name1
[50], src_name2
[50];
2638 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
2639 if(shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)) {
2640 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2641 /* No modifiers are supported on SCS */
2642 shader_addline(buffer
, "SCS %s, %s;\n", dst_name
, src_name0
);
2644 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
)
2646 shader_arb_get_register_name(ins
, &dst
->reg
, src_name0
, &is_color
);
2647 shader_addline(buffer
, "MOV_SAT %s, %s;\n", dst_name
, src_name0
);
2649 } else if(priv
->target_version
>= NV2
) {
2650 shader_arb_get_register_name(ins
, &dst
->reg
, dst_name
, &is_color
);
2652 /* Sincos writemask must be .x, .y or .xy */
2653 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
)
2654 shader_addline(buffer
, "COS%s %s.x, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name0
);
2655 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
)
2656 shader_addline(buffer
, "SIN%s %s.y, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name0
);
2658 /* Approximate sine and cosine with a taylor series, as per math textbook. The application passes 8
2659 * helper constants(D3DSINCOSCONST1 and D3DSINCOSCONST2) in src1 and src2.
2661 * sin(x) = x - x^3/3! + x^5/5! - x^7/7! + ...
2662 * cos(x) = 1 - x^2/2! + x^4/4! - x^6/6! + ...
2664 * The constants we get are:
2666 * +1 +1, -1 -1 +1 +1 -1 -1
2667 * ---- , ---- , ---- , ----- , ----- , ----- , ------
2668 * 1!*2 2!*4 3!*8 4!*16 5!*32 6!*64 7!*128
2670 * If used with x^2, x^3, x^4 etc they calculate sin(x/2) and cos(x/2):
2674 * (x/2)^4 = x^4 / 16
2675 * (x/2)^5 = x^5 / 32
2678 * To get the final result:
2679 * sin(x) = 2 * sin(x/2) * cos(x/2)
2680 * cos(x) = cos(x/2)^2 - sin(x/2)^2
2681 * (from sin(x+y) and cos(x+y) rules)
2683 * As per MSDN, dst.z is undefined after the operation, and so is
2684 * dst.x and dst.y if they're masked out by the writemask. Ie
2685 * sincos dst.y, src1, c0, c1
2686 * returns the sine in dst.y. dst.x and dst.z are undefined, dst.w is not touched. The assembler
2687 * vsa.exe also stops with an error if the dest register is the same register as the source
2688 * register. This means we can use dest.xyz as temporary storage. The assembler vsa.exe output also
2689 * indicates that sincos consumes 8 instruction slots in vs_2_0(and, strangely, in vs_3_0).
2691 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
2692 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name2
);
2693 shader_arb_get_register_name(ins
, &dst
->reg
, dst_name
, &is_color
);
2695 shader_addline(buffer
, "MUL %s.x, %s, %s;\n", dst_name
, src_name0
, src_name0
); /* x ^ 2 */
2696 shader_addline(buffer
, "MUL TA.y, %s.x, %s;\n", dst_name
, src_name0
); /* x ^ 3 */
2697 shader_addline(buffer
, "MUL %s.y, TA.y, %s;\n", dst_name
, src_name0
); /* x ^ 4 */
2698 shader_addline(buffer
, "MUL TA.z, %s.y, %s;\n", dst_name
, src_name0
); /* x ^ 5 */
2699 shader_addline(buffer
, "MUL %s.z, TA.z, %s;\n", dst_name
, src_name0
); /* x ^ 6 */
2700 shader_addline(buffer
, "MUL TA.w, %s.z, %s;\n", dst_name
, src_name0
); /* x ^ 7 */
2704 * Unfortunately we don't get the constants in a DP4-capable form. Is there a way to
2705 * properly merge that with MULs in the code above?
2706 * The swizzles .yz and xw however fit into the .yzxw swizzle added to ps_2_0. Maybe
2707 * we can merge the sine and cosine MAD rows to calculate them together.
2709 shader_addline(buffer
, "MUL TA.x, %s, %s.w;\n", src_name0
, src_name2
); /* x^1, +1/(1!*2) */
2710 shader_addline(buffer
, "MAD TA.x, TA.y, %s.x, TA.x;\n", src_name2
); /* -1/(3!*8) */
2711 shader_addline(buffer
, "MAD TA.x, TA.z, %s.w, TA.x;\n", src_name1
); /* +1/(5!*32) */
2712 shader_addline(buffer
, "MAD TA.x, TA.w, %s.x, TA.x;\n", src_name1
); /* -1/(7!*128) */
2715 shader_addline(buffer
, "MAD TA.y, %s.x, %s.y, %s.z;\n", dst_name
, src_name2
, src_name2
); /* -1/(2!*4), +1.0 */
2716 shader_addline(buffer
, "MAD TA.y, %s.y, %s.z, TA.y;\n", dst_name
, src_name1
); /* +1/(4!*16) */
2717 shader_addline(buffer
, "MAD TA.y, %s.z, %s.y, TA.y;\n", dst_name
, src_name1
); /* -1/(6!*64) */
2719 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
) {
2721 shader_addline(buffer
, "MUL TA.z, TA.y, TA.y;\n");
2722 shader_addline(buffer
, "MAD %s.x, -TA.x, TA.x, TA.z;\n", dst_name
);
2724 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
) {
2726 shader_addline(buffer
, "MUL %s.y, TA.x, TA.y;\n", dst_name
);
2727 shader_addline(buffer
, "ADD %s.y, %s.y, %s.y;\n", dst_name
, dst_name
, dst_name
);
2732 static void shader_hw_sgn(const struct wined3d_shader_instruction
*ins
)
2734 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2737 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
2739 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2740 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
);
2742 /* SGN is only valid in vertex shaders */
2743 if(ctx
->target_version
>= NV2
) {
2744 shader_addline(buffer
, "SSG%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
);
2748 /* If SRC > 0.0, -SRC < SRC = TRUE, otherwise false.
2749 * if SRC < 0.0, SRC < -SRC = TRUE. If neither is true, src = 0.0
2751 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
) {
2752 shader_addline(buffer
, "SLT %s, -%s, %s;\n", dst_name
, src_name
, src_name
);
2754 /* src contains TA? Write to the dest first. This won't overwrite our destination.
2755 * Then use TA, and calculate the final result
2757 * Not reading from TA? Store the first result in TA to avoid overwriting the
2758 * destination if src reg = dst reg
2760 if(strstr(src_name
, "TA"))
2762 shader_addline(buffer
, "SLT %s, %s, -%s;\n", dst_name
, src_name
, src_name
);
2763 shader_addline(buffer
, "SLT TA, -%s, %s;\n", src_name
, src_name
);
2764 shader_addline(buffer
, "ADD %s, %s, -TA;\n", dst_name
, dst_name
);
2768 shader_addline(buffer
, "SLT TA, -%s, %s;\n", src_name
, src_name
);
2769 shader_addline(buffer
, "SLT %s, %s, -%s;\n", dst_name
, src_name
, src_name
);
2770 shader_addline(buffer
, "ADD %s, TA, -%s;\n", dst_name
, dst_name
);
2775 static void shader_hw_dsy(const struct wined3d_shader_instruction
*ins
)
2777 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2783 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
);
2784 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src
);
2785 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2787 shader_addline(buffer
, "DDY %s, %s;\n", dst
, src
);
2788 shader_addline(buffer
, "MUL%s %s, %s, ycorrection.y;\n", shader_arb_get_modifier(ins
), dst
, dst_name
);
2791 static void shader_hw_pow(const struct wined3d_shader_instruction
*ins
)
2793 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2794 char src0
[50], src1
[50], dst
[50];
2795 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
2796 BOOL need_abs
= FALSE
;
2797 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2798 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2800 /* POW operates on the absolute value of the input */
2801 src0_copy
.modifiers
= abs_modifier(src0_copy
.modifiers
, &need_abs
);
2803 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
);
2804 shader_arb_get_src_param(ins
, &src0_copy
, 0, src0
);
2805 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src1
);
2808 shader_addline(buffer
, "ABS TA.x, %s;\n", src0
);
2810 shader_addline(buffer
, "MOV TA.x, %s;\n", src0
);
2812 if (priv
->target_version
>= NV2
)
2814 shader_addline(buffer
, "MOVC TA.y, %s;\n", src1
);
2815 shader_addline(buffer
, "POW%s %s, TA.x, TA.y;\n", shader_arb_get_modifier(ins
), dst
);
2816 shader_addline(buffer
, "MOV %s (EQ.y), %s;\n", dst
, one
);
2820 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2821 const char *flt_eps
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_EPS
);
2823 shader_addline(buffer
, "ABS TA.y, %s;\n", src1
);
2824 shader_addline(buffer
, "SGE TA.y, -TA.y, %s;\n", zero
);
2825 /* Possibly add flt_eps to avoid getting float special values */
2826 shader_addline(buffer
, "MAD TA.z, TA.y, %s, %s;\n", flt_eps
, src1
);
2827 shader_addline(buffer
, "POW%s TA.x, TA.x, TA.z;\n", shader_arb_get_modifier(ins
));
2828 shader_addline(buffer
, "MAD TA.x, -TA.x, TA.y, TA.x;\n");
2829 shader_addline(buffer
, "MAD %s, TA.y, %s, TA.x;\n", dst
, one
);
2833 static void shader_hw_loop(const struct wined3d_shader_instruction
*ins
)
2835 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2837 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2840 shader_arb_get_src_param(ins
, &ins
->src
[1], 0, src_name
);
2844 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2845 struct list
*e
= list_head(&priv
->control_frames
);
2846 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2848 if(priv
->loop_depth
> 1) shader_addline(buffer
, "PUSHA aL;\n");
2849 /* The constant loader makes sure to load -1 into iX.w */
2850 shader_addline(buffer
, "ARLC aL, %s.xywz;\n", src_name
);
2851 shader_addline(buffer
, "BRA loop_%u_end (LE.x);\n", control_frame
->no
.loop
);
2852 shader_addline(buffer
, "loop_%u_start:\n", control_frame
->no
.loop
);
2856 shader_addline(buffer
, "LOOP %s;\n", src_name
);
2860 static void shader_hw_rep(const struct wined3d_shader_instruction
*ins
)
2862 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2864 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2866 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
);
2868 /* The constant loader makes sure to load -1 into iX.w */
2871 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2872 struct list
*e
= list_head(&priv
->control_frames
);
2873 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2875 if(priv
->loop_depth
> 1) shader_addline(buffer
, "PUSHA aL;\n");
2877 shader_addline(buffer
, "ARLC aL, %s.xywz;\n", src_name
);
2878 shader_addline(buffer
, "BRA loop_%u_end (LE.x);\n", control_frame
->no
.loop
);
2879 shader_addline(buffer
, "loop_%u_start:\n", control_frame
->no
.loop
);
2883 shader_addline(buffer
, "REP %s;\n", src_name
);
2887 static void shader_hw_endloop(const struct wined3d_shader_instruction
*ins
)
2889 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2890 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2894 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2895 struct list
*e
= list_head(&priv
->control_frames
);
2896 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2898 shader_addline(buffer
, "ARAC aL.xy, aL;\n");
2899 shader_addline(buffer
, "BRA loop_%u_start (GT.x);\n", control_frame
->no
.loop
);
2900 shader_addline(buffer
, "loop_%u_end:\n", control_frame
->no
.loop
);
2902 if(priv
->loop_depth
> 1) shader_addline(buffer
, "POPA aL;\n");
2906 shader_addline(buffer
, "ENDLOOP;\n");
2910 static void shader_hw_endrep(const struct wined3d_shader_instruction
*ins
)
2912 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2913 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2917 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2918 struct list
*e
= list_head(&priv
->control_frames
);
2919 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2921 shader_addline(buffer
, "ARAC aL.xy, aL;\n");
2922 shader_addline(buffer
, "BRA loop_%u_start (GT.x);\n", control_frame
->no
.loop
);
2923 shader_addline(buffer
, "loop_%u_end:\n", control_frame
->no
.loop
);
2925 if(priv
->loop_depth
> 1) shader_addline(buffer
, "POPA aL;\n");
2929 shader_addline(buffer
, "ENDREP;\n");
2933 static const struct control_frame
*find_last_loop(const struct shader_arb_ctx_priv
*priv
)
2935 struct control_frame
*control_frame
;
2937 LIST_FOR_EACH_ENTRY(control_frame
, &priv
->control_frames
, struct control_frame
, entry
)
2939 if(control_frame
->type
== LOOP
|| control_frame
->type
== REP
) return control_frame
;
2941 ERR("Could not find loop for break\n");
2945 static void shader_hw_break(const struct wined3d_shader_instruction
*ins
)
2947 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2948 const struct control_frame
*control_frame
= find_last_loop(ins
->ctx
->backend_data
);
2949 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2953 shader_addline(buffer
, "BRA loop_%u_end;\n", control_frame
->no
.loop
);
2957 shader_addline(buffer
, "BRK;\n");
2961 static const char *get_compare(enum wined3d_shader_rel_op op
)
2965 case WINED3D_SHADER_REL_OP_GT
: return "GT";
2966 case WINED3D_SHADER_REL_OP_EQ
: return "EQ";
2967 case WINED3D_SHADER_REL_OP_GE
: return "GE";
2968 case WINED3D_SHADER_REL_OP_LT
: return "LT";
2969 case WINED3D_SHADER_REL_OP_NE
: return "NE";
2970 case WINED3D_SHADER_REL_OP_LE
: return "LE";
2972 FIXME("Unrecognized operator %#x.\n", op
);
2977 static enum wined3d_shader_rel_op
invert_compare(enum wined3d_shader_rel_op op
)
2981 case WINED3D_SHADER_REL_OP_GT
: return WINED3D_SHADER_REL_OP_LE
;
2982 case WINED3D_SHADER_REL_OP_EQ
: return WINED3D_SHADER_REL_OP_NE
;
2983 case WINED3D_SHADER_REL_OP_GE
: return WINED3D_SHADER_REL_OP_LT
;
2984 case WINED3D_SHADER_REL_OP_LT
: return WINED3D_SHADER_REL_OP_GE
;
2985 case WINED3D_SHADER_REL_OP_NE
: return WINED3D_SHADER_REL_OP_EQ
;
2986 case WINED3D_SHADER_REL_OP_LE
: return WINED3D_SHADER_REL_OP_GT
;
2988 FIXME("Unrecognized operator %#x.\n", op
);
2993 static void shader_hw_breakc(const struct wined3d_shader_instruction
*ins
)
2995 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2996 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2997 const struct control_frame
*control_frame
= find_last_loop(ins
->ctx
->backend_data
);
3000 const char *comp
= get_compare(ins
->flags
);
3002 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
3003 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
3007 /* SUBC CC, src0, src1" works only in pixel shaders, so use TA to throw
3008 * away the subtraction result
3010 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3011 shader_addline(buffer
, "BRA loop_%u_end (%s.x);\n", control_frame
->no
.loop
, comp
);
3015 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3016 shader_addline(buffer
, "BRK (%s.x);\n", comp
);
3020 static void shader_hw_ifc(const struct wined3d_shader_instruction
*ins
)
3022 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3023 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3024 struct list
*e
= list_head(&priv
->control_frames
);
3025 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3029 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3031 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
3032 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
3036 /* Invert the flag. We jump to the else label if the condition is NOT true */
3037 comp
= get_compare(invert_compare(ins
->flags
));
3038 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3039 shader_addline(buffer
, "BRA ifc_%u_else (%s.x);\n", control_frame
->no
.ifc
, comp
);
3043 comp
= get_compare(ins
->flags
);
3044 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3045 shader_addline(buffer
, "IF %s.x;\n", comp
);
3049 static void shader_hw_else(const struct wined3d_shader_instruction
*ins
)
3051 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3052 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3053 struct list
*e
= list_head(&priv
->control_frames
);
3054 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3055 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3059 shader_addline(buffer
, "BRA ifc_%u_endif;\n", control_frame
->no
.ifc
);
3060 shader_addline(buffer
, "ifc_%u_else:\n", control_frame
->no
.ifc
);
3061 control_frame
->had_else
= TRUE
;
3065 shader_addline(buffer
, "ELSE;\n");
3069 static void shader_hw_endif(const struct wined3d_shader_instruction
*ins
)
3071 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3072 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3073 struct list
*e
= list_head(&priv
->control_frames
);
3074 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3075 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3079 if(control_frame
->had_else
)
3081 shader_addline(buffer
, "ifc_%u_endif:\n", control_frame
->no
.ifc
);
3085 shader_addline(buffer
, "#No else branch. else is endif\n");
3086 shader_addline(buffer
, "ifc_%u_else:\n", control_frame
->no
.ifc
);
3091 shader_addline(buffer
, "ENDIF;\n");
3095 static void shader_hw_texldd(const struct wined3d_shader_instruction
*ins
)
3097 DWORD sampler_idx
= ins
->src
[1].reg
.idx
[0].offset
;
3099 char reg_src
[3][40];
3100 WORD flags
= TEX_DERIV
;
3102 shader_arb_get_dst_param(ins
, &ins
->dst
[0], reg_dest
);
3103 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_src
[0]);
3104 shader_arb_get_src_param(ins
, &ins
->src
[2], 1, reg_src
[1]);
3105 shader_arb_get_src_param(ins
, &ins
->src
[3], 2, reg_src
[2]);
3107 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) flags
|= TEX_PROJ
;
3108 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) flags
|= TEX_BIAS
;
3110 shader_hw_sample(ins
, sampler_idx
, reg_dest
, reg_src
[0], flags
, reg_src
[1], reg_src
[2]);
3113 static void shader_hw_texldl(const struct wined3d_shader_instruction
*ins
)
3115 DWORD sampler_idx
= ins
->src
[1].reg
.idx
[0].offset
;
3118 WORD flags
= TEX_LOD
;
3120 shader_arb_get_dst_param(ins
, &ins
->dst
[0], reg_dest
);
3121 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_coord
);
3123 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) flags
|= TEX_PROJ
;
3124 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) flags
|= TEX_BIAS
;
3126 shader_hw_sample(ins
, sampler_idx
, reg_dest
, reg_coord
, flags
, NULL
, NULL
);
3129 static void shader_hw_label(const struct wined3d_shader_instruction
*ins
)
3131 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3132 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3134 priv
->in_main_func
= FALSE
;
3135 /* Call instructions activate the NV extensions, not labels and rets. If there is an uncalled
3136 * subroutine, don't generate a label that will make GL complain
3138 if(priv
->target_version
== ARB
) return;
3140 shader_addline(buffer
, "l%u:\n", ins
->src
[0].reg
.idx
[0].offset
);
3143 static void vshader_add_footer(struct shader_arb_ctx_priv
*priv_ctx
,
3144 const struct arb_vshader_private
*shader_data
, const struct arb_vs_compile_args
*args
,
3145 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
,
3146 struct wined3d_shader_buffer
*buffer
)
3150 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
3151 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
3152 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
3153 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
3155 if (args
->super
.fog_src
== VS_FOG_Z
)
3157 shader_addline(buffer
, "MOV result.fogcoord, TMP_OUT.z;\n");
3163 /* posFixup.x is always 1.0, so we can safely use it */
3164 shader_addline(buffer
, "ADD result.fogcoord, posFixup.x, -posFixup.x;\n");
3168 /* Clamp fogcoord */
3169 const char *zero
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ZERO
);
3170 const char *one
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ONE
);
3172 shader_addline(buffer
, "MIN TMP_FOGCOORD.x, TMP_FOGCOORD.x, %s;\n", one
);
3173 shader_addline(buffer
, "MAX result.fogcoord.x, TMP_FOGCOORD.x, %s;\n", zero
);
3177 /* Clipplanes are always stored without y inversion */
3178 if (use_nv_clip(gl_info
) && priv_ctx
->target_version
>= NV2
)
3180 if (args
->super
.clip_enabled
)
3182 for (i
= 0; i
< priv_ctx
->vs_clipplanes
; i
++)
3184 shader_addline(buffer
, "DP4 result.clip[%u].x, TMP_OUT, state.clip[%u].plane;\n", i
, i
);
3188 else if (args
->clip
.boolclip
.clip_texcoord
)
3190 unsigned int cur_clip
= 0;
3191 char component
[4] = {'x', 'y', 'z', 'w'};
3192 const char *zero
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_ZERO
);
3194 for (i
= 0; i
< gl_info
->limits
.clipplanes
; ++i
)
3196 if (args
->clip
.boolclip
.clipplane_mask
& (1 << i
))
3198 shader_addline(buffer
, "DP4 TA.%c, TMP_OUT, state.clip[%u].plane;\n",
3199 component
[cur_clip
++], i
);
3205 shader_addline(buffer
, "MOV TA, %s;\n", zero
);
3208 shader_addline(buffer
, "MOV TA.yzw, %s;\n", zero
);
3211 shader_addline(buffer
, "MOV TA.zw, %s;\n", zero
);
3214 shader_addline(buffer
, "MOV TA.w, %s;\n", zero
);
3217 shader_addline(buffer
, "MOV result.texcoord[%u], TA;\n",
3218 args
->clip
.boolclip
.clip_texcoord
- 1);
3221 /* Write the final position.
3223 * OpenGL coordinates specify the center of the pixel while d3d coords specify
3224 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
3225 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
3226 * contains 1.0 to allow a mad, but arb vs swizzles are too restricted for that.
3228 shader_addline(buffer
, "MUL TA, posFixup, TMP_OUT.w;\n");
3229 shader_addline(buffer
, "ADD TMP_OUT.x, TMP_OUT.x, TA.z;\n");
3230 shader_addline(buffer
, "MAD TMP_OUT.y, TMP_OUT.y, posFixup.y, TA.w;\n");
3232 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
3233 * and the glsl equivalent
3235 if (need_helper_const(shader_data
, reg_maps
, gl_info
))
3237 const char *two
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_TWO
);
3238 shader_addline(buffer
, "MAD TMP_OUT.z, TMP_OUT.z, %s, -TMP_OUT.w;\n", two
);
3242 shader_addline(buffer
, "ADD TMP_OUT.z, TMP_OUT.z, TMP_OUT.z;\n");
3243 shader_addline(buffer
, "ADD TMP_OUT.z, TMP_OUT.z, -TMP_OUT.w;\n");
3246 shader_addline(buffer
, "MOV result.position, TMP_OUT;\n");
3248 priv_ctx
->footer_written
= TRUE
;
3251 static void shader_hw_ret(const struct wined3d_shader_instruction
*ins
)
3253 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3254 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3255 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
3256 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3258 if(priv
->target_version
== ARB
) return;
3262 if (priv
->in_main_func
) vshader_add_footer(priv
, shader
->backend_data
,
3263 priv
->cur_vs_args
, ins
->ctx
->reg_maps
, ins
->ctx
->gl_info
, buffer
);
3266 shader_addline(buffer
, "RET;\n");
3269 static void shader_hw_call(const struct wined3d_shader_instruction
*ins
)
3271 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3272 shader_addline(buffer
, "CAL l%u;\n", ins
->src
[0].reg
.idx
[0].offset
);
3275 /* Context activation is done by the caller. */
3276 static GLuint
create_arb_blt_vertex_program(const struct wined3d_gl_info
*gl_info
)
3278 GLuint program_id
= 0;
3281 const char *blt_vprogram
=
3283 "PARAM c[1] = { { 1, 0.5 } };\n"
3284 "MOV result.position, vertex.position;\n"
3285 "MOV result.color, c[0].x;\n"
3286 "MOV result.texcoord[0], vertex.texcoord[0];\n"
3289 GL_EXTCALL(glGenProgramsARB(1, &program_id
));
3290 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, program_id
));
3291 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
3292 strlen(blt_vprogram
), blt_vprogram
));
3293 checkGLcall("glProgramStringARB()");
3295 gl_info
->gl_ops
.gl
.p_glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
3298 FIXME("Vertex program error at position %d: %s\n\n", pos
,
3299 debugstr_a((const char *)gl_info
->gl_ops
.gl
.p_glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
3300 shader_arb_dump_program_source(blt_vprogram
);
3306 GL_EXTCALL(glGetProgramivARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
3307 checkGLcall("glGetProgramivARB()");
3308 if (!native
) WARN("Program exceeds native resource limits.\n");
3314 /* Context activation is done by the caller. */
3315 static GLuint
create_arb_blt_fragment_program(const struct wined3d_gl_info
*gl_info
,
3316 enum tex_types tex_type
, BOOL masked
)
3318 GLuint program_id
= 0;
3319 const char *fprogram
;
3322 static const char * const blt_fprograms_full
[tex_type_count
] =
3329 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
3330 "MOV result.depth.z, R0.x;\n"
3337 "TEX R0.x, fragment.texcoord[0], texture[0], CUBE;\n"
3338 "MOV result.depth.z, R0.x;\n"
3343 "TEX R0.x, fragment.texcoord[0], texture[0], RECT;\n"
3344 "MOV result.depth.z, R0.x;\n"
3348 static const char * const blt_fprograms_masked
[tex_type_count
] =
3354 "PARAM mask = program.local[0];\n"
3356 "SLT R0.xy, fragment.position, mask.zwzw;\n"
3357 "MUL R0.x, R0.x, R0.y;\n"
3359 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
3360 "MOV result.depth.z, R0.x;\n"
3366 "PARAM mask = program.local[0];\n"
3368 "SLT R0.xy, fragment.position, mask.zwzw;\n"
3369 "MUL R0.x, R0.x, R0.y;\n"
3371 "TEX R0.x, fragment.texcoord[0], texture[0], CUBE;\n"
3372 "MOV result.depth.z, R0.x;\n"
3376 "PARAM mask = program.local[0];\n"
3378 "SLT R0.xy, fragment.position, mask.zwzw;\n"
3379 "MUL R0.x, R0.x, R0.y;\n"
3381 "TEX R0.x, fragment.texcoord[0], texture[0], RECT;\n"
3382 "MOV result.depth.z, R0.x;\n"
3386 fprogram
= masked
? blt_fprograms_masked
[tex_type
] : blt_fprograms_full
[tex_type
];
3389 FIXME("tex_type %#x not supported, falling back to tex_2d\n", tex_type
);
3391 fprogram
= masked
? blt_fprograms_masked
[tex_type
] : blt_fprograms_full
[tex_type
];
3394 GL_EXTCALL(glGenProgramsARB(1, &program_id
));
3395 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, program_id
));
3396 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
, strlen(fprogram
), fprogram
));
3397 checkGLcall("glProgramStringARB()");
3399 gl_info
->gl_ops
.gl
.p_glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
3402 FIXME("Fragment program error at position %d: %s\n\n", pos
,
3403 debugstr_a((const char *)gl_info
->gl_ops
.gl
.p_glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
3404 shader_arb_dump_program_source(fprogram
);
3410 GL_EXTCALL(glGetProgramivARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
3411 checkGLcall("glGetProgramivARB()");
3412 if (!native
) WARN("Program exceeds native resource limits.\n");
3418 static void arbfp_add_sRGB_correction(struct wined3d_shader_buffer
*buffer
, const char *fragcolor
,
3419 const char *tmp1
, const char *tmp2
, const char *tmp3
, const char *tmp4
, BOOL condcode
)
3421 /* Perform sRGB write correction. See GLX_EXT_framebuffer_sRGB */
3425 /* Sigh. MOVC CC doesn't work, so use one of the temps as dummy dest */
3426 shader_addline(buffer
, "SUBC %s, %s.x, srgb_consts1.x;\n", tmp1
, fragcolor
);
3427 /* Calculate the > 0.0031308 case */
3428 shader_addline(buffer
, "POW %s.x (GE), %s.x, srgb_consts0.x;\n", fragcolor
, fragcolor
);
3429 shader_addline(buffer
, "POW %s.y (GE), %s.y, srgb_consts0.x;\n", fragcolor
, fragcolor
);
3430 shader_addline(buffer
, "POW %s.z (GE), %s.z, srgb_consts0.x;\n", fragcolor
, fragcolor
);
3431 shader_addline(buffer
, "MUL %s.xyz (GE), %s, srgb_consts0.y;\n", fragcolor
, fragcolor
);
3432 shader_addline(buffer
, "SUB %s.xyz (GE), %s, srgb_consts0.z;\n", fragcolor
, fragcolor
);
3433 /* Calculate the < case */
3434 shader_addline(buffer
, "MUL %s.xyz (LT), srgb_consts0.w, %s;\n", fragcolor
, fragcolor
);
3438 /* Calculate the > 0.0031308 case */
3439 shader_addline(buffer
, "POW %s.x, %s.x, srgb_consts0.x;\n", tmp1
, fragcolor
);
3440 shader_addline(buffer
, "POW %s.y, %s.y, srgb_consts0.x;\n", tmp1
, fragcolor
);
3441 shader_addline(buffer
, "POW %s.z, %s.z, srgb_consts0.x;\n", tmp1
, fragcolor
);
3442 shader_addline(buffer
, "MUL %s, %s, srgb_consts0.y;\n", tmp1
, tmp1
);
3443 shader_addline(buffer
, "SUB %s, %s, srgb_consts0.z;\n", tmp1
, tmp1
);
3444 /* Calculate the < case */
3445 shader_addline(buffer
, "MUL %s, srgb_consts0.w, %s;\n", tmp2
, fragcolor
);
3446 /* Get 1.0 / 0.0 masks for > 0.0031308 and < 0.0031308 */
3447 shader_addline(buffer
, "SLT %s, srgb_consts1.x, %s;\n", tmp3
, fragcolor
);
3448 shader_addline(buffer
, "SGE %s, srgb_consts1.x, %s;\n", tmp4
, fragcolor
);
3449 /* Store the components > 0.0031308 in the destination */
3450 shader_addline(buffer
, "MUL %s.xyz, %s, %s;\n", fragcolor
, tmp1
, tmp3
);
3451 /* Add the components that are < 0.0031308 */
3452 shader_addline(buffer
, "MAD %s.xyz, %s, %s, %s;\n", fragcolor
, tmp2
, tmp4
, fragcolor
);
3453 /* Move everything into result.color at once. Nvidia hardware cannot handle partial
3454 * result.color writes(.rgb first, then .a), or handle overwriting already written
3455 * components. The assembler uses a temporary register in this case, which is usually
3456 * not allocated from one of our registers that were used earlier.
3459 /* [0.0;1.0] clamping. Not needed, this is done implicitly */
3462 static const DWORD
*find_loop_control_values(const struct wined3d_shader
*shader
, DWORD idx
)
3464 const struct wined3d_shader_lconst
*constant
;
3466 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsI
, struct wined3d_shader_lconst
, entry
)
3468 if (constant
->idx
== idx
)
3470 return constant
->value
;
3476 static void init_ps_input(const struct wined3d_shader
*shader
,
3477 const struct arb_ps_compile_args
*args
, struct shader_arb_ctx_priv
*priv
)
3479 static const char * const texcoords
[8] =
3481 "fragment.texcoord[0]", "fragment.texcoord[1]", "fragment.texcoord[2]", "fragment.texcoord[3]",
3482 "fragment.texcoord[4]", "fragment.texcoord[5]", "fragment.texcoord[6]", "fragment.texcoord[7]"
3485 const struct wined3d_shader_signature_element
*sig
= shader
->input_signature
;
3486 const char *semantic_name
;
3489 switch(args
->super
.vp_mode
)
3491 case pretransformed
:
3493 /* The pixelshader has to collect the varyings on its own. In any case properly load
3494 * color0 and color1. In the case of pretransformed vertices also load texcoords. Set
3495 * other attribs to 0.0.
3497 * For fixedfunction this behavior is correct, according to the tests. For pretransformed
3498 * we'd either need a replacement shader that can load other attribs like BINORMAL, or
3499 * load the texcoord attrib pointers to match the pixel shader signature
3501 for(i
= 0; i
< MAX_REG_INPUT
; i
++)
3503 semantic_name
= sig
[i
].semantic_name
;
3504 semantic_idx
= sig
[i
].semantic_idx
;
3505 if (!semantic_name
) continue;
3507 if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_COLOR
))
3509 if (!semantic_idx
) priv
->ps_input
[i
] = "fragment.color.primary";
3510 else if(semantic_idx
== 1) priv
->ps_input
[i
] = "fragment.color.secondary";
3511 else priv
->ps_input
[i
] = "0.0";
3513 else if(args
->super
.vp_mode
== fixedfunction
)
3515 priv
->ps_input
[i
] = "0.0";
3517 else if(shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
3519 if(semantic_idx
< 8) priv
->ps_input
[i
] = texcoords
[semantic_idx
];
3520 else priv
->ps_input
[i
] = "0.0";
3522 else if(shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_FOG
))
3524 if (!semantic_idx
) priv
->ps_input
[i
] = "fragment.fogcoord";
3525 else priv
->ps_input
[i
] = "0.0";
3529 priv
->ps_input
[i
] = "0.0";
3532 TRACE("v%u, semantic %s%u is %s\n", i
, semantic_name
, semantic_idx
, priv
->ps_input
[i
]);
3537 /* That one is easy. The vertex shaders provide v0-v7 in fragment.texcoord and v8 and v9 in
3540 for(i
= 0; i
< 8; i
++)
3542 priv
->ps_input
[i
] = texcoords
[i
];
3544 priv
->ps_input
[8] = "fragment.color.primary";
3545 priv
->ps_input
[9] = "fragment.color.secondary";
3550 static void arbfp_add_linear_fog(struct wined3d_shader_buffer
*buffer
,
3551 const char *fragcolor
, const char *tmp
)
3553 shader_addline(buffer
, "SUB %s.x, state.fog.params.z, fragment.fogcoord.x;\n", tmp
);
3554 shader_addline(buffer
, "MUL_SAT %s.x, %s.x, state.fog.params.w;\n", tmp
, tmp
);
3555 shader_addline(buffer
, "LRP %s.rgb, %s.x, %s, state.fog.color;\n", fragcolor
, tmp
, fragcolor
);
3558 /* Context activation is done by the caller. */
3559 static GLuint
shader_arb_generate_pshader(const struct wined3d_shader
*shader
,
3560 const struct wined3d_gl_info
*gl_info
, struct wined3d_shader_buffer
*buffer
,
3561 const struct arb_ps_compile_args
*args
, struct arb_ps_compiled_shader
*compiled
)
3563 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
3564 const DWORD
*function
= shader
->function
;
3567 DWORD next_local
= 0;
3568 struct shader_arb_ctx_priv priv_ctx
;
3569 BOOL dcl_td
= FALSE
;
3570 BOOL want_nv_prog
= FALSE
;
3571 struct arb_pshader_private
*shader_priv
= shader
->backend_data
;
3574 BOOL custom_linear_fog
= FALSE
;
3578 unsigned int i
, found
= 0;
3580 for (i
= 0, map
= reg_maps
->temporary
; map
; map
>>= 1, ++i
)
3583 || (shader
->u
.ps
.color0_mov
&& i
== shader
->u
.ps
.color0_reg
)
3584 || (reg_maps
->shader_version
.major
< 2 && !i
))
3587 sprintf(srgbtmp
[found
], "R%u", i
);
3589 if (found
== 4) break;
3594 sprintf(srgbtmp
[0], "TA");
3595 sprintf(srgbtmp
[1], "TB");
3596 sprintf(srgbtmp
[2], "TC");
3597 sprintf(srgbtmp
[3], "TD");
3601 sprintf(srgbtmp
[1], "TA");
3602 sprintf(srgbtmp
[2], "TB");
3603 sprintf(srgbtmp
[3], "TC");
3606 sprintf(srgbtmp
[2], "TA");
3607 sprintf(srgbtmp
[3], "TB");
3610 sprintf(srgbtmp
[3], "TA");
3616 /* Create the hw ARB shader */
3617 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
3618 priv_ctx
.cur_ps_args
= args
;
3619 priv_ctx
.compiled_fprog
= compiled
;
3620 priv_ctx
.cur_np2fixup_info
= &compiled
->np2fixup_info
;
3621 init_ps_input(shader
, args
, &priv_ctx
);
3622 list_init(&priv_ctx
.control_frames
);
3623 priv_ctx
.ps_post_process
= args
->super
.srgb_correction
;
3625 /* Avoid enabling NV_fragment_program* if we do not need it.
3627 * Enabling GL_NV_fragment_program_option causes the driver to occupy a temporary register,
3628 * and it slows down the shader execution noticeably(about 5%). Usually our instruction emulation
3629 * is faster than what we gain from using higher native instructions. There are some things though
3630 * that cannot be emulated. In that case enable the extensions.
3631 * If the extension is enabled, instruction handlers that support both ways will use it.
3633 * Testing shows no performance difference between OPTION NV_fragment_program2 and NV_fragment_program.
3634 * So enable the best we can get.
3636 if(reg_maps
->usesdsx
|| reg_maps
->usesdsy
|| reg_maps
->loop_depth
> 0 || reg_maps
->usestexldd
||
3637 reg_maps
->usestexldl
|| reg_maps
->usesfacing
|| reg_maps
->usesifc
|| reg_maps
->usescall
)
3639 want_nv_prog
= TRUE
;
3642 shader_addline(buffer
, "!!ARBfp1.0\n");
3643 if (want_nv_prog
&& gl_info
->supported
[NV_FRAGMENT_PROGRAM2
])
3645 shader_addline(buffer
, "OPTION NV_fragment_program2;\n");
3646 priv_ctx
.target_version
= NV3
;
3648 else if (want_nv_prog
&& gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
])
3650 shader_addline(buffer
, "OPTION NV_fragment_program;\n");
3651 priv_ctx
.target_version
= NV2
;
3655 /* This is an error - either we're advertising the wrong shader version, or aren't enforcing some
3658 ERR("The shader requires instructions that are not available in plain GL_ARB_fragment_program\n");
3661 priv_ctx
.target_version
= ARB
;
3664 if (reg_maps
->rt_mask
> 1)
3666 shader_addline(buffer
, "OPTION ARB_draw_buffers;\n");
3669 if (reg_maps
->shader_version
.major
< 3)
3671 switch (args
->super
.fog
)
3673 case WINED3D_FFP_PS_FOG_OFF
:
3675 case WINED3D_FFP_PS_FOG_LINEAR
:
3676 if (gl_info
->quirks
& WINED3D_QUIRK_BROKEN_ARB_FOG
)
3678 custom_linear_fog
= TRUE
;
3679 priv_ctx
.ps_post_process
= TRUE
;
3682 shader_addline(buffer
, "OPTION ARB_fog_linear;\n");
3684 case WINED3D_FFP_PS_FOG_EXP
:
3685 shader_addline(buffer
, "OPTION ARB_fog_exp;\n");
3687 case WINED3D_FFP_PS_FOG_EXP2
:
3688 shader_addline(buffer
, "OPTION ARB_fog_exp2;\n");
3693 /* For now always declare the temps. At least the Nvidia assembler optimizes completely
3694 * unused temps away(but occupies them for the whole shader if they're used once). Always
3695 * declaring them avoids tricky bookkeeping work
3697 shader_addline(buffer
, "TEMP TA;\n"); /* Used for modifiers */
3698 shader_addline(buffer
, "TEMP TB;\n"); /* Used for modifiers */
3699 shader_addline(buffer
, "TEMP TC;\n"); /* Used for modifiers */
3700 if(dcl_td
) shader_addline(buffer
, "TEMP TD;\n"); /* Used for sRGB writing */
3701 shader_addline(buffer
, "PARAM coefdiv = { 0.5, 0.25, 0.125, 0.0625 };\n");
3702 shader_addline(buffer
, "PARAM coefmul = { 2, 4, 8, 16 };\n");
3703 wined3d_ftoa(eps
, ftoa_tmp
);
3704 shader_addline(buffer
, "PARAM ps_helper_const = { 0.0, 1.0, %s, 0.0 };\n", ftoa_tmp
);
3706 if (reg_maps
->shader_version
.major
< 2)
3708 strcpy(fragcolor
, "R0");
3712 if (priv_ctx
.ps_post_process
)
3714 if (shader
->u
.ps
.color0_mov
)
3716 sprintf(fragcolor
, "R%u", shader
->u
.ps
.color0_reg
);
3720 shader_addline(buffer
, "TEMP TMP_COLOR;\n");
3721 strcpy(fragcolor
, "TMP_COLOR");
3724 strcpy(fragcolor
, "result.color");
3728 if (args
->super
.srgb_correction
)
3730 shader_addline(buffer
, "PARAM srgb_consts0 = ");
3731 shader_arb_append_imm_vec4(buffer
, wined3d_srgb_const0
);
3732 shader_addline(buffer
, ";\n");
3733 shader_addline(buffer
, "PARAM srgb_consts1 = ");
3734 shader_arb_append_imm_vec4(buffer
, wined3d_srgb_const1
);
3735 shader_addline(buffer
, ";\n");
3738 /* Base Declarations */
3739 shader_generate_arb_declarations(shader
, reg_maps
, buffer
, gl_info
, NULL
, &priv_ctx
);
3741 for (i
= 0, map
= reg_maps
->bumpmat
; map
; map
>>= 1, ++i
)
3743 unsigned char bump_const
;
3745 if (!(map
& 1)) continue;
3747 bump_const
= compiled
->numbumpenvmatconsts
;
3748 compiled
->bumpenvmatconst
[bump_const
].const_num
= WINED3D_CONST_NUM_UNUSED
;
3749 compiled
->bumpenvmatconst
[bump_const
].texunit
= i
;
3750 compiled
->luminanceconst
[bump_const
].const_num
= WINED3D_CONST_NUM_UNUSED
;
3751 compiled
->luminanceconst
[bump_const
].texunit
= i
;
3753 /* We can fit the constants into the constant limit for sure because texbem, texbeml, bem and beml are only supported
3754 * in 1.x shaders, and GL_ARB_fragment_program has a constant limit of 24 constants. So in the worst case we're loading
3755 * 8 shader constants, 8 bump matrices and 8 luminance parameters and are perfectly fine. (No NP2 fixup on bumpmapped
3756 * textures due to conditional NP2 restrictions)
3758 * Use local constants to load the bump env parameters, not program.env. This avoids collisions with d3d constants of
3759 * shaders in newer shader models. Since the bump env parameters have to share their space with NP2 fixup constants,
3760 * their location is shader dependent anyway and they cannot be loaded globally.
3762 compiled
->bumpenvmatconst
[bump_const
].const_num
= next_local
++;
3763 shader_addline(buffer
, "PARAM bumpenvmat%d = program.local[%d];\n",
3764 i
, compiled
->bumpenvmatconst
[bump_const
].const_num
);
3765 compiled
->numbumpenvmatconsts
= bump_const
+ 1;
3767 if (!(reg_maps
->luminanceparams
& (1 << i
))) continue;
3769 compiled
->luminanceconst
[bump_const
].const_num
= next_local
++;
3770 shader_addline(buffer
, "PARAM luminance%d = program.local[%d];\n",
3771 i
, compiled
->luminanceconst
[bump_const
].const_num
);
3774 for(i
= 0; i
< MAX_CONST_I
; i
++)
3776 compiled
->int_consts
[i
] = WINED3D_CONST_NUM_UNUSED
;
3777 if (reg_maps
->integer_constants
& (1 << i
) && priv_ctx
.target_version
>= NV2
)
3779 const DWORD
*control_values
= find_loop_control_values(shader
, i
);
3783 shader_addline(buffer
, "PARAM I%u = {%u, %u, %u, -1};\n", i
,
3784 control_values
[0], control_values
[1], control_values
[2]);
3788 compiled
->int_consts
[i
] = next_local
;
3789 compiled
->num_int_consts
++;
3790 shader_addline(buffer
, "PARAM I%u = program.local[%u];\n", i
, next_local
++);
3795 if(reg_maps
->vpos
|| reg_maps
->usesdsy
)
3797 compiled
->ycorrection
= next_local
;
3798 shader_addline(buffer
, "PARAM ycorrection = program.local[%u];\n", next_local
++);
3802 shader_addline(buffer
, "TEMP vpos;\n");
3803 /* ycorrection.x: Backbuffer height(onscreen) or 0(offscreen).
3804 * ycorrection.y: -1.0(onscreen), 1.0(offscreen)
3805 * ycorrection.z: 1.0
3806 * ycorrection.w: 0.0
3808 shader_addline(buffer
, "MAD vpos, fragment.position, ycorrection.zyww, ycorrection.wxww;\n");
3809 shader_addline(buffer
, "FLR vpos.xy, vpos;\n");
3814 compiled
->ycorrection
= WINED3D_CONST_NUM_UNUSED
;
3817 /* Load constants to fixup NP2 texcoords if there are still free constants left:
3818 * Constants (texture dimensions) for the NP2 fixup are loaded as local program parameters. This will consume
3819 * at most 8 (MAX_FRAGMENT_SAMPLERS / 2) parameters, which is highly unlikely, since the application had to
3820 * use 16 NP2 textures at the same time. In case that we run out of constants the fixup is simply not
3821 * applied / activated. This will probably result in wrong rendering of the texture, but will save us from
3822 * shader compilation errors and the subsequent errors when drawing with this shader. */
3823 if (priv_ctx
.cur_ps_args
->super
.np2_fixup
) {
3824 unsigned char cur_fixup_sampler
= 0;
3826 struct arb_ps_np2fixup_info
* const fixup
= priv_ctx
.cur_np2fixup_info
;
3827 const WORD map
= priv_ctx
.cur_ps_args
->super
.np2_fixup
;
3828 const UINT max_lconsts
= gl_info
->limits
.arb_ps_local_constants
;
3830 fixup
->offset
= next_local
;
3831 fixup
->super
.active
= 0;
3833 for (i
= 0; i
< MAX_FRAGMENT_SAMPLERS
; ++i
) {
3834 if (!(map
& (1 << i
))) continue;
3836 if (fixup
->offset
+ (cur_fixup_sampler
>> 1) < max_lconsts
) {
3837 fixup
->super
.active
|= (1 << i
);
3838 fixup
->super
.idx
[i
] = cur_fixup_sampler
++;
3840 FIXME("No free constant found to load NP2 fixup data into shader. "
3841 "Sampling from this texture will probably look wrong.\n");
3846 fixup
->super
.num_consts
= (cur_fixup_sampler
+ 1) >> 1;
3847 if (fixup
->super
.num_consts
) {
3848 shader_addline(buffer
, "PARAM np2fixup[%u] = { program.env[%u..%u] };\n",
3849 fixup
->super
.num_consts
, fixup
->offset
, fixup
->super
.num_consts
+ fixup
->offset
- 1);
3853 if (shader_priv
->clipplane_emulation
!= ~0U && args
->clip
)
3855 shader_addline(buffer
, "KIL fragment.texcoord[%u];\n", shader_priv
->clipplane_emulation
);
3858 /* Base Shader Body */
3859 shader_generate_main(shader
, buffer
, reg_maps
, function
, &priv_ctx
);
3861 if(args
->super
.srgb_correction
) {
3862 arbfp_add_sRGB_correction(buffer
, fragcolor
, srgbtmp
[0], srgbtmp
[1], srgbtmp
[2], srgbtmp
[3],
3863 priv_ctx
.target_version
>= NV2
);
3866 if (custom_linear_fog
)
3867 arbfp_add_linear_fog(buffer
, fragcolor
, "TA");
3869 if(strcmp(fragcolor
, "result.color")) {
3870 shader_addline(buffer
, "MOV result.color, %s;\n", fragcolor
);
3872 shader_addline(buffer
, "END\n");
3874 /* TODO: change to resource.glObjectHandle or something like that */
3875 GL_EXTCALL(glGenProgramsARB(1, &retval
));
3877 TRACE("Creating a hw pixel shader, prg=%d\n", retval
);
3878 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, retval
));
3880 TRACE("Created hw pixel shader, prg=%d\n", retval
);
3881 /* Create the program and check for errors */
3882 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
3883 buffer
->content_size
, buffer
->buffer
));
3884 checkGLcall("glProgramStringARB()");
3886 gl_info
->gl_ops
.gl
.p_glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &errPos
);
3889 FIXME("HW PixelShader Error at position %d: %s\n\n",
3890 errPos
, debugstr_a((const char *)gl_info
->gl_ops
.gl
.p_glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
3891 shader_arb_dump_program_source(buffer
->buffer
);
3898 GL_EXTCALL(glGetProgramivARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
3899 checkGLcall("glGetProgramivARB()");
3900 if (!native
) WARN("Program exceeds native resource limits.\n");
3906 static int compare_sig(const struct wined3d_shader_signature_element
*sig1
, const struct wined3d_shader_signature_element
*sig2
)
3911 for(i
= 0; i
< MAX_REG_INPUT
; i
++)
3913 if (!sig1
[i
].semantic_name
|| !sig2
[i
].semantic_name
)
3915 /* Compare pointers, not contents. One string is NULL(element does not exist), the other one is not NULL */
3916 if(sig1
[i
].semantic_name
!= sig2
[i
].semantic_name
) return sig1
[i
].semantic_name
< sig2
[i
].semantic_name
? -1 : 1;
3920 if ((ret
= strcmp(sig1
[i
].semantic_name
, sig2
[i
].semantic_name
))) return ret
;
3921 if(sig1
[i
].semantic_idx
!= sig2
[i
].semantic_idx
) return sig1
[i
].semantic_idx
< sig2
[i
].semantic_idx
? -1 : 1;
3922 if(sig1
[i
].sysval_semantic
!= sig2
[i
].sysval_semantic
) return sig1
[i
].sysval_semantic
< sig2
[i
].sysval_semantic
? -1 : 1;
3923 if(sig1
[i
].component_type
!= sig2
[i
].component_type
) return sig1
[i
].component_type
< sig2
[i
].component_type
? -1 : 1;
3924 if(sig1
[i
].register_idx
!= sig2
[i
].register_idx
) return sig1
[i
].register_idx
< sig2
[i
].register_idx
? -1 : 1;
3925 if(sig1
[i
].mask
!= sig2
[i
].mask
) return sig1
[i
].mask
< sig2
[i
].mask
? -1 : 1;
3930 static struct wined3d_shader_signature_element
*clone_sig(const struct wined3d_shader_signature_element
*sig
)
3932 struct wined3d_shader_signature_element
*new;
3936 new = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*new) * MAX_REG_INPUT
);
3937 for(i
= 0; i
< MAX_REG_INPUT
; i
++)
3939 if (!sig
[i
].semantic_name
) continue;
3942 /* Clone the semantic string */
3943 name
= HeapAlloc(GetProcessHeap(), 0, strlen(sig
[i
].semantic_name
) + 1);
3944 strcpy(name
, sig
[i
].semantic_name
);
3945 new[i
].semantic_name
= name
;
3950 static DWORD
find_input_signature(struct shader_arb_priv
*priv
, const struct wined3d_shader_signature_element
*sig
)
3952 struct wine_rb_entry
*entry
= wine_rb_get(&priv
->signature_tree
, sig
);
3953 struct ps_signature
*found_sig
;
3957 found_sig
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
3958 TRACE("Found existing signature %u\n", found_sig
->idx
);
3959 return found_sig
->idx
;
3961 found_sig
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*found_sig
));
3962 found_sig
->sig
= clone_sig(sig
);
3963 found_sig
->idx
= priv
->ps_sig_number
++;
3964 TRACE("New signature stored and assigned number %u\n", found_sig
->idx
);
3965 if(wine_rb_put(&priv
->signature_tree
, sig
, &found_sig
->entry
) == -1)
3967 ERR("Failed to insert program entry.\n");
3969 return found_sig
->idx
;
3972 static void init_output_registers(const struct wined3d_shader
*shader
,
3973 const struct wined3d_shader_signature_element
*ps_input_sig
,
3974 struct shader_arb_ctx_priv
*priv_ctx
, struct arb_vs_compiled_shader
*compiled
)
3977 static const char * const texcoords
[8] =
3979 "result.texcoord[0]", "result.texcoord[1]", "result.texcoord[2]", "result.texcoord[3]",
3980 "result.texcoord[4]", "result.texcoord[5]", "result.texcoord[6]", "result.texcoord[7]"
3982 const char *semantic_name
;
3983 DWORD semantic_idx
, reg_idx
;
3985 /* Write generic input varyings 0 to 7 to result.texcoord[], varying 8 to result.color.primary
3986 * and varying 9 to result.color.secondary
3988 static const char * const decl_idx_to_string
[MAX_REG_INPUT
] =
3990 "result.texcoord[0]", "result.texcoord[1]", "result.texcoord[2]", "result.texcoord[3]",
3991 "result.texcoord[4]", "result.texcoord[5]", "result.texcoord[6]", "result.texcoord[7]",
3992 "result.color.primary", "result.color.secondary"
3997 TRACE("Pixel shader uses builtin varyings\n");
3998 /* Map builtins to builtins */
3999 for(i
= 0; i
< 8; i
++)
4001 priv_ctx
->texcrd_output
[i
] = texcoords
[i
];
4003 priv_ctx
->color_output
[0] = "result.color.primary";
4004 priv_ctx
->color_output
[1] = "result.color.secondary";
4005 priv_ctx
->fog_output
= "TMP_FOGCOORD";
4007 /* Map declared regs to builtins. Use "TA" to /dev/null unread output */
4008 for (i
= 0; i
< (sizeof(shader
->output_signature
) / sizeof(*shader
->output_signature
)); ++i
)
4010 semantic_name
= shader
->output_signature
[i
].semantic_name
;
4011 if (!semantic_name
) continue;
4013 if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_POSITION
))
4015 TRACE("o%u is TMP_OUT\n", i
);
4016 if (!shader
->output_signature
[i
].semantic_idx
) priv_ctx
->vs_output
[i
] = "TMP_OUT";
4017 else priv_ctx
->vs_output
[i
] = "TA";
4019 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_PSIZE
))
4021 TRACE("o%u is result.pointsize\n", i
);
4022 if (!shader
->output_signature
[i
].semantic_idx
) priv_ctx
->vs_output
[i
] = "result.pointsize";
4023 else priv_ctx
->vs_output
[i
] = "TA";
4025 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_COLOR
))
4027 TRACE("o%u is result.color.?, idx %u\n", i
, shader
->output_signature
[i
].semantic_idx
);
4028 if (!shader
->output_signature
[i
].semantic_idx
)
4029 priv_ctx
->vs_output
[i
] = "result.color.primary";
4030 else if (shader
->output_signature
[i
].semantic_idx
== 1)
4031 priv_ctx
->vs_output
[i
] = "result.color.secondary";
4032 else priv_ctx
->vs_output
[i
] = "TA";
4034 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
4036 TRACE("o%u is %s\n", i
, texcoords
[shader
->output_signature
[i
].semantic_idx
]);
4037 if (shader
->output_signature
[i
].semantic_idx
>= 8) priv_ctx
->vs_output
[i
] = "TA";
4038 else priv_ctx
->vs_output
[i
] = texcoords
[shader
->output_signature
[i
].semantic_idx
];
4040 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_FOG
))
4042 TRACE("o%u is result.fogcoord\n", i
);
4043 if (shader
->output_signature
[i
].semantic_idx
> 0) priv_ctx
->vs_output
[i
] = "TA";
4044 else priv_ctx
->vs_output
[i
] = "result.fogcoord";
4048 priv_ctx
->vs_output
[i
] = "TA";
4054 TRACE("Pixel shader uses declared varyings\n");
4056 /* Map builtin to declared. /dev/null the results by default to the TA temp reg */
4057 for(i
= 0; i
< 8; i
++)
4059 priv_ctx
->texcrd_output
[i
] = "TA";
4061 priv_ctx
->color_output
[0] = "TA";
4062 priv_ctx
->color_output
[1] = "TA";
4063 priv_ctx
->fog_output
= "TA";
4065 for(i
= 0; i
< MAX_REG_INPUT
; i
++)
4067 semantic_name
= ps_input_sig
[i
].semantic_name
;
4068 semantic_idx
= ps_input_sig
[i
].semantic_idx
;
4069 reg_idx
= ps_input_sig
[i
].register_idx
;
4070 if (!semantic_name
) continue;
4072 /* If a declared input register is not written by builtin arguments, don't write to it.
4073 * GL_NV_vertex_program makes sure the input defaults to 0.0, which is correct with D3D
4075 * Don't care about POSITION and PSIZE here - this is a builtin vertex shader, position goes
4076 * to TMP_OUT in any case
4078 if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
4080 if (semantic_idx
< 8)
4081 priv_ctx
->texcrd_output
[semantic_idx
] = decl_idx_to_string
[reg_idx
];
4083 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_COLOR
))
4085 if (semantic_idx
< 2)
4086 priv_ctx
->color_output
[semantic_idx
] = decl_idx_to_string
[reg_idx
];
4088 else if(shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_FOG
))
4091 priv_ctx
->fog_output
= decl_idx_to_string
[reg_idx
];
4098 if (!strcmp(decl_idx_to_string
[reg_idx
], "result.color.primary")
4099 || !strcmp(decl_idx_to_string
[reg_idx
], "result.color.secondary"))
4101 compiled
->need_color_unclamp
= TRUE
;
4105 /* Map declared to declared */
4106 for (i
= 0; i
< (sizeof(shader
->output_signature
) / sizeof(*shader
->output_signature
)); ++i
)
4108 /* Write unread output to TA to throw them away */
4109 priv_ctx
->vs_output
[i
] = "TA";
4110 semantic_name
= shader
->output_signature
[i
].semantic_name
;
4111 if (!semantic_name
) continue;
4113 if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_POSITION
)
4114 && !shader
->output_signature
[i
].semantic_idx
)
4116 priv_ctx
->vs_output
[i
] = "TMP_OUT";
4119 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_PSIZE
)
4120 && !shader
->output_signature
[i
].semantic_idx
)
4122 priv_ctx
->vs_output
[i
] = "result.pointsize";
4126 for(j
= 0; j
< MAX_REG_INPUT
; j
++)
4128 if (!ps_input_sig
[j
].semantic_name
) continue;
4130 if (!strcmp(ps_input_sig
[j
].semantic_name
, semantic_name
)
4131 && ps_input_sig
[j
].semantic_idx
== shader
->output_signature
[i
].semantic_idx
)
4133 priv_ctx
->vs_output
[i
] = decl_idx_to_string
[ps_input_sig
[j
].register_idx
];
4135 if (!strcmp(priv_ctx
->vs_output
[i
], "result.color.primary")
4136 || !strcmp(priv_ctx
->vs_output
[i
], "result.color.secondary"))
4138 compiled
->need_color_unclamp
= TRUE
;
4145 /* Context activation is done by the caller. */
4146 static GLuint
shader_arb_generate_vshader(const struct wined3d_shader
*shader
,
4147 const struct wined3d_gl_info
*gl_info
, struct wined3d_shader_buffer
*buffer
,
4148 const struct arb_vs_compile_args
*args
, struct arb_vs_compiled_shader
*compiled
,
4149 const struct wined3d_shader_signature_element
*ps_input_sig
)
4151 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
4152 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
4153 struct shader_arb_priv
*priv
= shader
->device
->shader_priv
;
4154 const DWORD
*function
= shader
->function
;
4156 DWORD next_local
= 0;
4157 struct shader_arb_ctx_priv priv_ctx
;
4161 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
4162 priv_ctx
.cur_vs_args
= args
;
4163 list_init(&priv_ctx
.control_frames
);
4164 init_output_registers(shader
, ps_input_sig
, &priv_ctx
, compiled
);
4166 /* Create the hw ARB shader */
4167 shader_addline(buffer
, "!!ARBvp1.0\n");
4169 /* Always enable the NV extension if available. Unlike fragment shaders, there is no
4170 * mesurable performance penalty, and we can always make use of it for clipplanes.
4172 if (gl_info
->supported
[NV_VERTEX_PROGRAM3
])
4174 shader_addline(buffer
, "OPTION NV_vertex_program3;\n");
4175 priv_ctx
.target_version
= NV3
;
4176 shader_addline(buffer
, "ADDRESS aL;\n");
4178 else if (gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
])
4180 shader_addline(buffer
, "OPTION NV_vertex_program2;\n");
4181 priv_ctx
.target_version
= NV2
;
4182 shader_addline(buffer
, "ADDRESS aL;\n");
4184 priv_ctx
.target_version
= ARB
;
4187 shader_addline(buffer
, "TEMP TMP_OUT;\n");
4189 shader_addline(buffer
, "TEMP TMP_FOGCOORD;\n");
4190 if (need_helper_const(shader_data
, reg_maps
, gl_info
))
4193 wined3d_ftoa(eps
, ftoa_tmp
);
4194 shader_addline(buffer
, "PARAM helper_const = { 0.0, 1.0, 2.0, %s};\n", ftoa_tmp
);
4196 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
))
4198 shader_addline(buffer
, "PARAM rel_addr_const = { 0.5, %d.0, 0.0, 0.0 };\n", shader_data
->rel_offset
);
4199 shader_addline(buffer
, "TEMP A0_SHADOW;\n");
4202 shader_addline(buffer
, "TEMP TA;\n");
4203 shader_addline(buffer
, "TEMP TB;\n");
4205 /* Base Declarations */
4206 shader_generate_arb_declarations(shader
, reg_maps
, buffer
, gl_info
,
4207 &priv_ctx
.vs_clipplanes
, &priv_ctx
);
4209 for(i
= 0; i
< MAX_CONST_I
; i
++)
4211 compiled
->int_consts
[i
] = WINED3D_CONST_NUM_UNUSED
;
4212 if(reg_maps
->integer_constants
& (1 << i
) && priv_ctx
.target_version
>= NV2
)
4214 const DWORD
*control_values
= find_loop_control_values(shader
, i
);
4218 shader_addline(buffer
, "PARAM I%u = {%u, %u, %u, -1};\n", i
,
4219 control_values
[0], control_values
[1], control_values
[2]);
4223 compiled
->int_consts
[i
] = next_local
;
4224 compiled
->num_int_consts
++;
4225 shader_addline(buffer
, "PARAM I%u = program.local[%u];\n", i
, next_local
++);
4230 /* We need a constant to fixup the final position */
4231 shader_addline(buffer
, "PARAM posFixup = program.local[%u];\n", next_local
);
4232 compiled
->pos_fixup
= next_local
++;
4234 /* Initialize output parameters. GL_ARB_vertex_program does not require special initialization values
4235 * for output parameters. D3D in theory does not do that either, but some applications depend on a
4236 * proper initialization of the secondary color, and programs using the fixed function pipeline without
4237 * a replacement shader depend on the texcoord.w being set properly.
4239 * GL_NV_vertex_program defines that all output values are initialized to {0.0, 0.0, 0.0, 1.0}. This
4240 * assertion is in effect even when using GL_ARB_vertex_program without any NV specific additions. So
4241 * skip this if NV_vertex_program is supported. Otherwise, initialize the secondary color. For the tex-
4242 * coords, we have a flag in the opengl caps. Many cards do not require the texcoord being set, and
4243 * this can eat a number of instructions, so skip it unless this cap is set as well
4245 if (!gl_info
->supported
[NV_VERTEX_PROGRAM
])
4247 const char *color_init
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_0001
);
4248 shader_addline(buffer
, "MOV result.color.secondary, %s;\n", color_init
);
4250 if (gl_info
->quirks
& WINED3D_QUIRK_SET_TEXCOORD_W
&& !priv
->ffp_proj_control
)
4253 const char *one
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_ONE
);
4254 for(i
= 0; i
< min(8, MAX_REG_TEXCRD
); i
++)
4256 if (reg_maps
->texcoord_mask
[i
] && reg_maps
->texcoord_mask
[i
] != WINED3DSP_WRITEMASK_ALL
)
4257 shader_addline(buffer
, "MOV result.texcoord[%u].w, %s\n", i
, one
);
4262 /* The shader starts with the main function */
4263 priv_ctx
.in_main_func
= TRUE
;
4264 /* Base Shader Body */
4265 shader_generate_main(shader
, buffer
, reg_maps
, function
, &priv_ctx
);
4267 if (!priv_ctx
.footer_written
) vshader_add_footer(&priv_ctx
,
4268 shader_data
, args
, reg_maps
, gl_info
, buffer
);
4270 shader_addline(buffer
, "END\n");
4272 /* TODO: change to resource.glObjectHandle or something like that */
4273 GL_EXTCALL(glGenProgramsARB(1, &ret
));
4275 TRACE("Creating a hw vertex shader, prg=%d\n", ret
);
4276 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, ret
));
4278 TRACE("Created hw vertex shader, prg=%d\n", ret
);
4279 /* Create the program and check for errors */
4280 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
4281 buffer
->content_size
, buffer
->buffer
));
4282 checkGLcall("glProgramStringARB()");
4284 gl_info
->gl_ops
.gl
.p_glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &errPos
);
4287 FIXME("HW VertexShader Error at position %d: %s\n\n",
4288 errPos
, debugstr_a((const char *)gl_info
->gl_ops
.gl
.p_glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
4289 shader_arb_dump_program_source(buffer
->buffer
);
4296 GL_EXTCALL(glGetProgramivARB(GL_VERTEX_PROGRAM_ARB
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
4297 checkGLcall("glGetProgramivARB()");
4298 if (!native
) WARN("Program exceeds native resource limits.\n");
4304 /* Context activation is done by the caller. */
4305 static struct arb_ps_compiled_shader
*find_arb_pshader(struct wined3d_shader
*shader
,
4306 const struct arb_ps_compile_args
*args
)
4308 struct wined3d_device
*device
= shader
->device
;
4309 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
4310 const struct wined3d_d3d_info
*d3d_info
= &device
->adapter
->d3d_info
;
4313 struct arb_ps_compiled_shader
*new_array
;
4314 struct wined3d_shader_buffer buffer
;
4315 struct arb_pshader_private
*shader_data
;
4318 if (!shader
->backend_data
)
4320 struct shader_arb_priv
*priv
= device
->shader_priv
;
4322 shader
->backend_data
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
4323 shader_data
= shader
->backend_data
;
4324 shader_data
->clamp_consts
= shader
->reg_maps
.shader_version
.major
== 1;
4326 if (shader
->reg_maps
.shader_version
.major
< 3)
4327 shader_data
->input_signature_idx
= ~0U;
4329 shader_data
->input_signature_idx
= find_input_signature(priv
, shader
->input_signature
);
4331 TRACE("Shader got assigned input signature index %u\n", shader_data
->input_signature_idx
);
4333 if (!d3d_info
->vs_clipping
)
4334 shader_data
->clipplane_emulation
= shader_find_free_input_register(&shader
->reg_maps
,
4335 d3d_info
->limits
.ffp_blend_stages
- 1);
4337 shader_data
->clipplane_emulation
= ~0U;
4339 shader_data
= shader
->backend_data
;
4341 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4342 * so a linear search is more performant than a hashmap or a binary search
4343 * (cache coherency etc)
4345 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4347 if (!memcmp(&shader_data
->gl_shaders
[i
].args
, args
, sizeof(*args
)))
4348 return &shader_data
->gl_shaders
[i
];
4351 TRACE("No matching GL shader found, compiling a new shader\n");
4352 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4353 if (shader_data
->num_gl_shaders
)
4355 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4356 new_array
= HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, shader_data
->gl_shaders
,
4357 new_size
* sizeof(*shader_data
->gl_shaders
));
4359 new_array
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
->gl_shaders
));
4364 ERR("Out of memory\n");
4367 shader_data
->gl_shaders
= new_array
;
4368 shader_data
->shader_array_size
= new_size
;
4371 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4373 pixelshader_update_samplers(shader
, args
->super
.tex_types
);
4375 if (!shader_buffer_init(&buffer
))
4377 ERR("Failed to initialize shader buffer.\n");
4381 ret
= shader_arb_generate_pshader(shader
, gl_info
, &buffer
, args
,
4382 &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
]);
4383 shader_buffer_free(&buffer
);
4384 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].prgId
= ret
;
4386 return &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++];
4389 static inline BOOL
vs_args_equal(const struct arb_vs_compile_args
*stored
, const struct arb_vs_compile_args
*new,
4390 const DWORD use_map
, BOOL skip_int
) {
4391 if((stored
->super
.swizzle_map
& use_map
) != new->super
.swizzle_map
) return FALSE
;
4392 if(stored
->super
.clip_enabled
!= new->super
.clip_enabled
) return FALSE
;
4393 if(stored
->super
.fog_src
!= new->super
.fog_src
) return FALSE
;
4394 if(stored
->clip
.boolclip_compare
!= new->clip
.boolclip_compare
) return FALSE
;
4395 if(stored
->ps_signature
!= new->ps_signature
) return FALSE
;
4396 if(stored
->vertex
.samplers_compare
!= new->vertex
.samplers_compare
) return FALSE
;
4397 if(skip_int
) return TRUE
;
4399 return !memcmp(stored
->loop_ctrl
, new->loop_ctrl
, sizeof(stored
->loop_ctrl
));
4402 static struct arb_vs_compiled_shader
*find_arb_vshader(struct wined3d_shader
*shader
,
4403 const struct wined3d_gl_info
*gl_info
, DWORD use_map
, const struct arb_vs_compile_args
*args
,
4404 const struct wined3d_shader_signature_element
*ps_input_sig
)
4408 struct arb_vs_compiled_shader
*new_array
;
4409 struct wined3d_shader_buffer buffer
;
4410 struct arb_vshader_private
*shader_data
;
4413 if (!shader
->backend_data
)
4415 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
4417 shader
->backend_data
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
4418 shader_data
= shader
->backend_data
;
4420 if ((gl_info
->quirks
& WINED3D_QUIRK_ARB_VS_OFFSET_LIMIT
)
4421 && reg_maps
->min_rel_offset
<= reg_maps
->max_rel_offset
)
4423 if (reg_maps
->max_rel_offset
- reg_maps
->min_rel_offset
> 127)
4425 FIXME("The difference between the minimum and maximum relative offset is > 127.\n");
4426 FIXME("Which this OpenGL implementation does not support. Try using GLSL.\n");
4427 FIXME("Min: %u, Max: %u.\n", reg_maps
->min_rel_offset
, reg_maps
->max_rel_offset
);
4429 else if (reg_maps
->max_rel_offset
- reg_maps
->min_rel_offset
> 63)
4430 shader_data
->rel_offset
= reg_maps
->min_rel_offset
+ 63;
4431 else if (reg_maps
->max_rel_offset
> 63)
4432 shader_data
->rel_offset
= reg_maps
->min_rel_offset
;
4435 shader_data
= shader
->backend_data
;
4437 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4438 * so a linear search is more performant than a hashmap or a binary search
4439 * (cache coherency etc)
4441 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
4442 if (vs_args_equal(&shader_data
->gl_shaders
[i
].args
, args
,
4443 use_map
, gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]))
4445 return &shader_data
->gl_shaders
[i
];
4449 TRACE("No matching GL shader found, compiling a new shader\n");
4451 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4452 if (shader_data
->num_gl_shaders
)
4454 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4455 new_array
= HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, shader_data
->gl_shaders
,
4456 new_size
* sizeof(*shader_data
->gl_shaders
));
4458 new_array
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
->gl_shaders
));
4463 ERR("Out of memory\n");
4466 shader_data
->gl_shaders
= new_array
;
4467 shader_data
->shader_array_size
= new_size
;
4470 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4472 if (!shader_buffer_init(&buffer
))
4474 ERR("Failed to initialize shader buffer.\n");
4478 ret
= shader_arb_generate_vshader(shader
, gl_info
, &buffer
, args
,
4479 &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
],
4481 shader_buffer_free(&buffer
);
4482 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].prgId
= ret
;
4484 return &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++];
4487 static void find_arb_ps_compile_args(const struct wined3d_state
*state
,
4488 const struct wined3d_context
*context
, const struct wined3d_shader
*shader
,
4489 struct arb_ps_compile_args
*args
)
4491 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4492 const struct wined3d_d3d_info
*d3d_info
= context
->d3d_info
;
4496 find_ps_compile_args(state
, shader
, context
->stream_info
.position_transformed
, &args
->super
, gl_info
);
4498 /* This forces all local boolean constants to 1 to make them stateblock independent */
4499 args
->bools
= shader
->reg_maps
.local_bool_consts
;
4501 for(i
= 0; i
< MAX_CONST_B
; i
++)
4503 if (state
->ps_consts_b
[i
])
4504 args
->bools
|= ( 1 << i
);
4507 /* Only enable the clip plane emulation KIL if at least one clipplane is enabled. The KIL instruction
4508 * is quite expensive because it forces the driver to disable early Z discards. It is cheaper to
4509 * duplicate the shader than have a no-op KIL instruction in every shader
4511 if (!d3d_info
->vs_clipping
&& use_vs(state
)
4512 && state
->render_states
[WINED3D_RS_CLIPPING
]
4513 && state
->render_states
[WINED3D_RS_CLIPPLANEENABLE
])
4518 /* Skip if unused or local, or supported natively */
4519 int_skip
= ~shader
->reg_maps
.integer_constants
| shader
->reg_maps
.local_int_consts
;
4520 if (int_skip
== 0xffff || gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
])
4522 memset(args
->loop_ctrl
, 0, sizeof(args
->loop_ctrl
));
4526 for(i
= 0; i
< MAX_CONST_I
; i
++)
4528 if(int_skip
& (1 << i
))
4530 args
->loop_ctrl
[i
][0] = 0;
4531 args
->loop_ctrl
[i
][1] = 0;
4532 args
->loop_ctrl
[i
][2] = 0;
4536 args
->loop_ctrl
[i
][0] = state
->ps_consts_i
[i
* 4];
4537 args
->loop_ctrl
[i
][1] = state
->ps_consts_i
[i
* 4 + 1];
4538 args
->loop_ctrl
[i
][2] = state
->ps_consts_i
[i
* 4 + 2];
4543 static void find_arb_vs_compile_args(const struct wined3d_state
*state
,
4544 const struct wined3d_context
*context
, const struct wined3d_shader
*shader
,
4545 struct arb_vs_compile_args
*args
)
4547 const struct wined3d_device
*device
= shader
->device
;
4548 const struct wined3d_adapter
*adapter
= device
->adapter
;
4549 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4550 const struct wined3d_d3d_info
*d3d_info
= context
->d3d_info
;
4554 find_vs_compile_args(state
, shader
, context
->stream_info
.swizzle_map
, &args
->super
);
4556 args
->clip
.boolclip_compare
= 0;
4559 const struct wined3d_shader
*ps
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
4560 const struct arb_pshader_private
*shader_priv
= ps
->backend_data
;
4561 args
->ps_signature
= shader_priv
->input_signature_idx
;
4563 args
->clip
.boolclip
.clip_texcoord
= shader_priv
->clipplane_emulation
+ 1;
4567 args
->ps_signature
= ~0;
4568 if (!d3d_info
->vs_clipping
&& adapter
->fragment_pipe
== &arbfp_fragment_pipeline
)
4569 args
->clip
.boolclip
.clip_texcoord
= ffp_clip_emul(context
) ? d3d_info
->limits
.ffp_blend_stages
: 0;
4570 /* Otherwise: Setting boolclip_compare set clip_texcoord to 0 */
4573 if (args
->clip
.boolclip
.clip_texcoord
)
4575 if (state
->render_states
[WINED3D_RS_CLIPPING
])
4576 args
->clip
.boolclip
.clipplane_mask
= (unsigned char)state
->render_states
[WINED3D_RS_CLIPPLANEENABLE
];
4577 /* clipplane_mask was set to 0 by setting boolclip_compare to 0 */
4580 /* This forces all local boolean constants to 1 to make them stateblock independent */
4581 args
->clip
.boolclip
.bools
= shader
->reg_maps
.local_bool_consts
;
4582 /* TODO: Figure out if it would be better to store bool constants as bitmasks in the stateblock */
4583 for(i
= 0; i
< MAX_CONST_B
; i
++)
4585 if (state
->vs_consts_b
[i
])
4586 args
->clip
.boolclip
.bools
|= ( 1 << i
);
4589 args
->vertex
.samplers
[0] = context
->tex_unit_map
[MAX_FRAGMENT_SAMPLERS
+ 0];
4590 args
->vertex
.samplers
[1] = context
->tex_unit_map
[MAX_FRAGMENT_SAMPLERS
+ 1];
4591 args
->vertex
.samplers
[2] = context
->tex_unit_map
[MAX_FRAGMENT_SAMPLERS
+ 2];
4592 args
->vertex
.samplers
[3] = 0;
4594 /* Skip if unused or local */
4595 int_skip
= ~shader
->reg_maps
.integer_constants
| shader
->reg_maps
.local_int_consts
;
4596 /* This is about flow control, not clipping. */
4597 if (int_skip
== 0xffff || gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
])
4599 memset(args
->loop_ctrl
, 0, sizeof(args
->loop_ctrl
));
4603 for(i
= 0; i
< MAX_CONST_I
; i
++)
4605 if(int_skip
& (1 << i
))
4607 args
->loop_ctrl
[i
][0] = 0;
4608 args
->loop_ctrl
[i
][1] = 0;
4609 args
->loop_ctrl
[i
][2] = 0;
4613 args
->loop_ctrl
[i
][0] = state
->vs_consts_i
[i
* 4];
4614 args
->loop_ctrl
[i
][1] = state
->vs_consts_i
[i
* 4 + 1];
4615 args
->loop_ctrl
[i
][2] = state
->vs_consts_i
[i
* 4 + 2];
4620 /* Context activation is done by the caller. */
4621 static void shader_arb_select(void *shader_priv
, struct wined3d_context
*context
,
4622 const struct wined3d_state
*state
)
4624 struct shader_arb_priv
*priv
= shader_priv
;
4625 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4628 /* Deal with pixel shaders first so the vertex shader arg function has the input signature ready */
4631 struct wined3d_shader
*ps
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
4632 struct arb_ps_compile_args compile_args
;
4633 struct arb_ps_compiled_shader
*compiled
;
4635 TRACE("Using pixel shader %p.\n", ps
);
4636 find_arb_ps_compile_args(state
, context
, ps
, &compile_args
);
4637 compiled
= find_arb_pshader(ps
, &compile_args
);
4638 priv
->current_fprogram_id
= compiled
->prgId
;
4639 priv
->compiled_fprog
= compiled
;
4641 /* Bind the fragment program */
4642 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, priv
->current_fprogram_id
));
4643 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->current_fprogram_id);");
4645 if (!priv
->use_arbfp_fixed_func
)
4646 priv
->fragment_pipe
->enable_extension(gl_info
, FALSE
);
4648 /* Enable OpenGL fragment programs. */
4649 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
4650 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
4652 TRACE("Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", priv
->current_fprogram_id
);
4654 /* Pixel Shader 1.x constants are clamped to [-1;1], Pixel Shader 2.0 constants are not. If switching between
4655 * a 1.x and newer shader, reload the first 8 constants
4657 if (priv
->last_ps_const_clamped
!= ((struct arb_pshader_private
*)ps
->backend_data
)->clamp_consts
)
4659 priv
->last_ps_const_clamped
= ((struct arb_pshader_private
*)ps
->backend_data
)->clamp_consts
;
4660 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, 8);
4661 for(i
= 0; i
< 8; i
++)
4663 priv
->pshader_const_dirty
[i
] = 1;
4665 /* Also takes care of loading local constants */
4666 shader_arb_load_constants_internal(shader_priv
, context
, state
, TRUE
, FALSE
, TRUE
);
4670 UINT rt_height
= state
->fb
->render_targets
[0]->resource
.height
;
4671 shader_arb_ps_local_constants(compiled
, context
, state
, rt_height
);
4674 /* Force constant reloading for the NP2 fixup (see comment in shader_glsl_select for more info) */
4675 if (compiled
->np2fixup_info
.super
.active
)
4676 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_NP2_FIXUP
;
4678 if (ps
->load_local_constsF
)
4679 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_F
;
4683 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
] && !priv
->use_arbfp_fixed_func
)
4685 /* Disable only if we're not using arbfp fixed function fragment
4686 * processing. If this is used, keep GL_FRAGMENT_PROGRAM_ARB
4687 * enabled, and the fixed function pipeline will bind the fixed
4688 * function replacement shader. */
4689 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4690 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4691 priv
->current_fprogram_id
= 0;
4693 priv
->fragment_pipe
->enable_extension(gl_info
, TRUE
);
4698 struct wined3d_shader
*vs
= state
->shader
[WINED3D_SHADER_TYPE_VERTEX
];
4699 struct arb_vs_compile_args compile_args
;
4700 struct arb_vs_compiled_shader
*compiled
;
4701 const struct wined3d_shader_signature_element
*ps_input_sig
;
4703 TRACE("Using vertex shader %p\n", vs
);
4704 find_arb_vs_compile_args(state
, context
, vs
, &compile_args
);
4706 /* Instead of searching for the signature in the signature list, read the one from the
4707 * current pixel shader. It's maybe not the shader where the signature came from, but it
4708 * is the same signature and faster to find. */
4709 if (compile_args
.ps_signature
== ~0U)
4710 ps_input_sig
= NULL
;
4712 ps_input_sig
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
]->input_signature
;
4714 compiled
= find_arb_vshader(vs
, context
->gl_info
, context
->stream_info
.use_map
,
4715 &compile_args
, ps_input_sig
);
4716 priv
->current_vprogram_id
= compiled
->prgId
;
4717 priv
->compiled_vprog
= compiled
;
4719 /* Bind the vertex program */
4720 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->current_vprogram_id
));
4721 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->current_vprogram_id);");
4723 priv
->vertex_pipe
->vp_enable(gl_info
, FALSE
);
4725 /* Enable OpenGL vertex programs */
4726 gl_info
->gl_ops
.gl
.p_glEnable(GL_VERTEX_PROGRAM_ARB
);
4727 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
4728 TRACE("Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", priv
->current_vprogram_id
);
4729 shader_arb_vs_local_constants(compiled
, context
, state
);
4731 if(priv
->last_vs_color_unclamp
!= compiled
->need_color_unclamp
) {
4732 priv
->last_vs_color_unclamp
= compiled
->need_color_unclamp
;
4734 if (gl_info
->supported
[ARB_COLOR_BUFFER_FLOAT
])
4736 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, !compiled
->need_color_unclamp
));
4737 checkGLcall("glClampColorARB");
4739 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
4743 if (vs
->load_local_constsF
)
4744 context
->constant_update_mask
|= WINED3D_SHADER_CONST_VS_F
;
4748 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
4750 priv
->current_vprogram_id
= 0;
4751 gl_info
->gl_ops
.gl
.p_glDisable(GL_VERTEX_PROGRAM_ARB
);
4752 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4754 priv
->vertex_pipe
->vp_enable(gl_info
, TRUE
);
4759 /* Context activation is done by the caller. */
4760 static void shader_arb_disable(void *shader_priv
, struct wined3d_context
*context
)
4762 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4763 struct shader_arb_priv
*priv
= shader_priv
;
4765 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
4767 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4768 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4769 priv
->current_fprogram_id
= 0;
4771 priv
->fragment_pipe
->enable_extension(gl_info
, FALSE
);
4773 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
4775 priv
->current_vprogram_id
= 0;
4776 gl_info
->gl_ops
.gl
.p_glDisable(GL_VERTEX_PROGRAM_ARB
);
4777 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4779 priv
->vertex_pipe
->vp_enable(gl_info
, FALSE
);
4781 if (gl_info
->supported
[ARB_COLOR_BUFFER_FLOAT
] && priv
->last_vs_color_unclamp
)
4783 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, GL_FIXED_ONLY_ARB
));
4784 checkGLcall("glClampColorARB");
4785 priv
->last_vs_color_unclamp
= FALSE
;
4788 context
->shader_update_mask
= (1 << WINED3D_SHADER_TYPE_PIXEL
)
4789 | (1 << WINED3D_SHADER_TYPE_VERTEX
)
4790 | (1 << WINED3D_SHADER_TYPE_GEOMETRY
);
4793 /* Context activation is done by the caller. */
4794 static void shader_arb_select_depth_blt(void *shader_priv
, const struct wined3d_gl_info
*gl_info
,
4795 enum tex_types tex_type
, const SIZE
*ds_mask_size
)
4797 const float mask
[] = {0.0f
, 0.0f
, (float)ds_mask_size
->cx
, (float)ds_mask_size
->cy
};
4798 BOOL masked
= ds_mask_size
->cx
&& ds_mask_size
->cy
;
4799 struct shader_arb_priv
*priv
= shader_priv
;
4800 GLuint
*blt_fprogram
;
4802 if (!priv
->depth_blt_vprogram_id
) priv
->depth_blt_vprogram_id
= create_arb_blt_vertex_program(gl_info
);
4803 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->depth_blt_vprogram_id
));
4804 gl_info
->gl_ops
.gl
.p_glEnable(GL_VERTEX_PROGRAM_ARB
);
4806 blt_fprogram
= masked
? &priv
->depth_blt_fprogram_id_masked
[tex_type
] : &priv
->depth_blt_fprogram_id_full
[tex_type
];
4807 if (!*blt_fprogram
) *blt_fprogram
= create_arb_blt_fragment_program(gl_info
, tex_type
, masked
);
4808 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, *blt_fprogram
));
4809 if (masked
) GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, 0, mask
));
4810 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
4813 /* Context activation is done by the caller. */
4814 static void shader_arb_deselect_depth_blt(void *shader_priv
, const struct wined3d_gl_info
*gl_info
)
4816 struct shader_arb_priv
*priv
= shader_priv
;
4818 if (priv
->current_vprogram_id
) {
4819 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->current_vprogram_id
));
4820 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
4822 TRACE("Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB.\n", priv
->current_vprogram_id
);
4826 gl_info
->gl_ops
.gl
.p_glDisable(GL_VERTEX_PROGRAM_ARB
);
4827 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4830 if (priv
->current_fprogram_id
) {
4831 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, priv
->current_fprogram_id
));
4832 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
4834 TRACE("Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB.\n", priv
->current_fprogram_id
);
4836 else if(!priv
->use_arbfp_fixed_func
)
4838 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4839 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4843 static void shader_arb_destroy(struct wined3d_shader
*shader
)
4845 struct wined3d_device
*device
= shader
->device
;
4846 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
4848 if (shader_is_pshader_version(shader
->reg_maps
.shader_version
.type
))
4850 struct arb_pshader_private
*shader_data
= shader
->backend_data
;
4853 if(!shader_data
) return; /* This can happen if a shader was never compiled */
4855 if (shader_data
->num_gl_shaders
)
4857 struct wined3d_context
*context
= context_acquire(device
, NULL
);
4859 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4861 GL_EXTCALL(glDeleteProgramsARB(1, &shader_data
->gl_shaders
[i
].prgId
));
4862 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader_data->gl_shaders[i].prgId))");
4865 context_release(context
);
4868 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
);
4869 HeapFree(GetProcessHeap(), 0, shader_data
);
4870 shader
->backend_data
= NULL
;
4874 struct arb_vshader_private
*shader_data
= shader
->backend_data
;
4877 if(!shader_data
) return; /* This can happen if a shader was never compiled */
4879 if (shader_data
->num_gl_shaders
)
4881 struct wined3d_context
*context
= context_acquire(device
, NULL
);
4883 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4885 GL_EXTCALL(glDeleteProgramsARB(1, &shader_data
->gl_shaders
[i
].prgId
));
4886 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader_data->gl_shaders[i].prgId))");
4889 context_release(context
);
4892 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
);
4893 HeapFree(GetProcessHeap(), 0, shader_data
);
4894 shader
->backend_data
= NULL
;
4898 static int sig_tree_compare(const void *key
, const struct wine_rb_entry
*entry
)
4900 struct ps_signature
*e
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
4901 return compare_sig(key
, e
->sig
);
4904 static const struct wine_rb_functions sig_tree_functions
=
4912 static HRESULT
shader_arb_alloc(struct wined3d_device
*device
, const struct wined3d_vertex_pipe_ops
*vertex_pipe
,
4913 const struct fragment_pipeline
*fragment_pipe
)
4915 struct shader_arb_priv
*priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*priv
));
4916 struct fragment_caps fragment_caps
;
4917 void *vertex_priv
, *fragment_priv
;
4918 const struct wined3d_d3d_info
*d3d_info
= &device
->adapter
->d3d_info
;
4920 if (!(vertex_priv
= vertex_pipe
->vp_alloc(&arb_program_shader_backend
, priv
)))
4922 ERR("Failed to initialize vertex pipe.\n");
4923 HeapFree(GetProcessHeap(), 0, priv
);
4927 if (!(fragment_priv
= fragment_pipe
->alloc_private(&arb_program_shader_backend
, priv
)))
4929 ERR("Failed to initialize fragment pipe.\n");
4930 vertex_pipe
->vp_free(device
);
4931 HeapFree(GetProcessHeap(), 0, priv
);
4935 priv
->vshader_const_dirty
= HeapAlloc(GetProcessHeap(), 0,
4936 sizeof(*priv
->vshader_const_dirty
) * d3d_info
->limits
.vs_uniform_count
);
4937 if (!priv
->vshader_const_dirty
)
4939 memset(priv
->vshader_const_dirty
, 1,
4940 sizeof(*priv
->vshader_const_dirty
) * d3d_info
->limits
.vs_uniform_count
);
4942 priv
->pshader_const_dirty
= HeapAlloc(GetProcessHeap(), 0,
4943 sizeof(*priv
->pshader_const_dirty
) * d3d_info
->limits
.ps_uniform_count
);
4944 if (!priv
->pshader_const_dirty
)
4946 memset(priv
->pshader_const_dirty
, 1,
4947 sizeof(*priv
->pshader_const_dirty
) * d3d_info
->limits
.ps_uniform_count
);
4949 if(wine_rb_init(&priv
->signature_tree
, &sig_tree_functions
) == -1)
4951 ERR("RB tree init failed\n");
4955 priv
->vertex_pipe
= vertex_pipe
;
4956 priv
->fragment_pipe
= fragment_pipe
;
4957 fragment_pipe
->get_caps(&device
->adapter
->gl_info
, &fragment_caps
);
4958 priv
->ffp_proj_control
= fragment_caps
.wined3d_caps
& WINED3D_FRAGMENT_CAP_PROJ_CONTROL
;
4960 device
->vertex_priv
= vertex_priv
;
4961 device
->fragment_priv
= fragment_priv
;
4962 device
->shader_priv
= priv
;
4967 HeapFree(GetProcessHeap(), 0, priv
->pshader_const_dirty
);
4968 HeapFree(GetProcessHeap(), 0, priv
->vshader_const_dirty
);
4969 fragment_pipe
->free_private(device
);
4970 vertex_pipe
->vp_free(device
);
4971 HeapFree(GetProcessHeap(), 0, priv
);
4972 return E_OUTOFMEMORY
;
4975 static void release_signature(struct wine_rb_entry
*entry
, void *context
)
4977 struct ps_signature
*sig
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
4979 for(i
= 0; i
< MAX_REG_INPUT
; i
++)
4981 HeapFree(GetProcessHeap(), 0, (char *) sig
->sig
[i
].semantic_name
);
4983 HeapFree(GetProcessHeap(), 0, sig
->sig
);
4984 HeapFree(GetProcessHeap(), 0, sig
);
4987 /* Context activation is done by the caller. */
4988 static void shader_arb_free(struct wined3d_device
*device
)
4990 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
4991 struct shader_arb_priv
*priv
= device
->shader_priv
;
4994 if (priv
->depth_blt_vprogram_id
)
4995 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_vprogram_id
));
4997 for (i
= 0; i
< tex_type_count
; ++i
)
4999 if (priv
->depth_blt_fprogram_id_full
[i
])
5001 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_fprogram_id_full
[i
]));
5003 if (priv
->depth_blt_fprogram_id_masked
[i
])
5005 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_fprogram_id_masked
[i
]));
5009 wine_rb_destroy(&priv
->signature_tree
, release_signature
, NULL
);
5010 HeapFree(GetProcessHeap(), 0, priv
->pshader_const_dirty
);
5011 HeapFree(GetProcessHeap(), 0, priv
->vshader_const_dirty
);
5012 priv
->fragment_pipe
->free_private(device
);
5013 priv
->vertex_pipe
->vp_free(device
);
5014 HeapFree(GetProcessHeap(), 0, device
->shader_priv
);
5017 static BOOL
shader_arb_allocate_context_data(struct wined3d_context
*context
)
5022 static void shader_arb_free_context_data(struct wined3d_context
*context
)
5024 struct shader_arb_priv
*priv
;
5026 if (!context
->swapchain
)
5029 priv
= context
->swapchain
->device
->shader_priv
;
5030 if (priv
->last_context
== context
)
5031 priv
->last_context
= NULL
;
5034 static void shader_arb_get_caps(const struct wined3d_gl_info
*gl_info
, struct shader_caps
*caps
)
5036 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
5041 /* 96 is the minimum allowed value of MAX_PROGRAM_ENV_PARAMETERS_ARB
5042 * for vertex programs. If the native limit is less than that it's
5043 * not very useful, and e.g. Mesa swrast returns 0, probably to
5044 * indicate it's a software implementation. */
5045 if (gl_info
->limits
.arb_vs_native_constants
< 96)
5046 vs_consts
= gl_info
->limits
.arb_vs_float_constants
;
5048 vs_consts
= min(gl_info
->limits
.arb_vs_float_constants
, gl_info
->limits
.arb_vs_native_constants
);
5050 if (gl_info
->supported
[NV_VERTEX_PROGRAM3
])
5053 TRACE("Hardware vertex shader version 3.0 enabled (NV_VERTEX_PROGRAM3)\n");
5055 else if (vs_consts
>= 256)
5057 /* Shader Model 2.0 requires at least 256 vertex shader constants */
5059 TRACE("Hardware vertex shader version 2.0 enabled (ARB_PROGRAM)\n");
5064 TRACE("Hardware vertex shader version 1.1 enabled (ARB_PROGRAM)\n");
5066 caps
->vs_version
= min(wined3d_settings
.max_sm_vs
, vs_version
);
5067 caps
->vs_uniform_count
= vs_consts
;
5071 caps
->vs_version
= 0;
5072 caps
->vs_uniform_count
= 0;
5075 caps
->gs_version
= 0;
5077 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
5082 /* Similar as above for vertex programs, but the minimum for fragment
5083 * programs is 24. */
5084 if (gl_info
->limits
.arb_ps_native_constants
< 24)
5085 ps_consts
= gl_info
->limits
.arb_ps_float_constants
;
5087 ps_consts
= min(gl_info
->limits
.arb_ps_float_constants
, gl_info
->limits
.arb_ps_native_constants
);
5089 if (gl_info
->supported
[NV_FRAGMENT_PROGRAM2
])
5092 TRACE("Hardware pixel shader version 3.0 enabled (NV_FRAGMENT_PROGRAM2)\n");
5094 else if (ps_consts
>= 32)
5096 /* Shader Model 2.0 requires at least 32 pixel shader constants */
5098 TRACE("Hardware pixel shader version 2.0 enabled (ARB_PROGRAM)\n");
5103 TRACE("Hardware pixel shader version 1.4 enabled (ARB_PROGRAM)\n");
5105 caps
->ps_version
= min(wined3d_settings
.max_sm_ps
, ps_version
);
5106 caps
->ps_uniform_count
= ps_consts
;
5107 caps
->ps_1x_max_value
= 8.0f
;
5111 caps
->ps_version
= 0;
5112 caps
->ps_uniform_count
= 0;
5113 caps
->ps_1x_max_value
= 0.0f
;
5116 caps
->wined3d_caps
= WINED3D_SHADER_CAP_SRGB_WRITE
;
5117 if (use_nv_clip(gl_info
))
5118 caps
->wined3d_caps
|= WINED3D_SHADER_CAP_VS_CLIPPING
;
5121 static BOOL
shader_arb_color_fixup_supported(struct color_fixup_desc fixup
)
5123 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
5125 TRACE("Checking support for color_fixup:\n");
5126 dump_color_fixup_desc(fixup
);
5129 /* We support everything except complex conversions. */
5130 if (!is_complex_fixup(fixup
))
5136 TRACE("[FAILED]\n");
5140 static void shader_arb_add_instruction_modifiers(const struct wined3d_shader_instruction
*ins
) {
5142 char write_mask
[20], regstr
[50];
5143 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
5144 BOOL is_color
= FALSE
;
5145 const struct wined3d_shader_dst_param
*dst
;
5147 if (!ins
->dst_count
) return;
5151 if (!shift
) return; /* Saturate alone is handled by the instructions */
5153 shader_arb_get_write_mask(ins
, dst
, write_mask
);
5154 shader_arb_get_register_name(ins
, &dst
->reg
, regstr
, &is_color
);
5156 /* Generate a line that does the output modifier computation
5157 * FIXME: _SAT vs shift? _SAT alone is already handled in the instructions, if this
5158 * maps problems in e.g. _d4_sat modify shader_arb_get_modifier
5160 shader_addline(buffer
, "MUL%s %s%s, %s, %s;\n", shader_arb_get_modifier(ins
),
5161 regstr
, write_mask
, regstr
, shift_tab
[shift
]);
5164 static const SHADER_HANDLER shader_arb_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
5166 /* WINED3DSIH_ABS */ shader_hw_map2gl
,
5167 /* WINED3DSIH_ADD */ shader_hw_map2gl
,
5168 /* WINED3DSIH_AND */ NULL
,
5169 /* WINED3DSIH_BEM */ pshader_hw_bem
,
5170 /* WINED3DSIH_BREAK */ shader_hw_break
,
5171 /* WINED3DSIH_BREAKC */ shader_hw_breakc
,
5172 /* WINED3DSIH_BREAKP */ NULL
,
5173 /* WINED3DSIH_CALL */ shader_hw_call
,
5174 /* WINED3DSIH_CALLNZ */ NULL
,
5175 /* WINED3DSIH_CMP */ pshader_hw_cmp
,
5176 /* WINED3DSIH_CND */ pshader_hw_cnd
,
5177 /* WINED3DSIH_CRS */ shader_hw_map2gl
,
5178 /* WINED3DSIH_CUT */ NULL
,
5179 /* WINED3DSIH_DCL */ shader_hw_nop
,
5180 /* WINED3DSIH_DCL_CONSTANT_BUFFER */ shader_hw_nop
,
5181 /* WINED3DSIH_DCL_INPUT_PRIMITIVE */ shader_hw_nop
,
5182 /* WINED3DSIH_DCL_OUTPUT_TOPOLOGY */ shader_hw_nop
,
5183 /* WINED3DSIH_DCL_VERTICES_OUT */ shader_hw_nop
,
5184 /* WINED3DSIH_DEF */ shader_hw_nop
,
5185 /* WINED3DSIH_DEFB */ shader_hw_nop
,
5186 /* WINED3DSIH_DEFI */ shader_hw_nop
,
5187 /* WINED3DSIH_DIV */ NULL
,
5188 /* WINED3DSIH_DP2ADD */ pshader_hw_dp2add
,
5189 /* WINED3DSIH_DP3 */ shader_hw_map2gl
,
5190 /* WINED3DSIH_DP4 */ shader_hw_map2gl
,
5191 /* WINED3DSIH_DST */ shader_hw_map2gl
,
5192 /* WINED3DSIH_DSX */ shader_hw_map2gl
,
5193 /* WINED3DSIH_DSY */ shader_hw_dsy
,
5194 /* WINED3DSIH_ELSE */ shader_hw_else
,
5195 /* WINED3DSIH_EMIT */ NULL
,
5196 /* WINED3DSIH_ENDIF */ shader_hw_endif
,
5197 /* WINED3DSIH_ENDLOOP */ shader_hw_endloop
,
5198 /* WINED3DSIH_ENDREP */ shader_hw_endrep
,
5199 /* WINED3DSIH_EQ */ NULL
,
5200 /* WINED3DSIH_EXP */ shader_hw_scalar_op
,
5201 /* WINED3DSIH_EXPP */ shader_hw_scalar_op
,
5202 /* WINED3DSIH_FRC */ shader_hw_map2gl
,
5203 /* WINED3DSIH_FTOI */ NULL
,
5204 /* WINED3DSIH_GE */ NULL
,
5205 /* WINED3DSIH_IADD */ NULL
,
5206 /* WINED3DSIH_IEQ */ NULL
,
5207 /* WINED3DSIH_IF */ NULL
/* Hardcoded into the shader */,
5208 /* WINED3DSIH_IFC */ shader_hw_ifc
,
5209 /* WINED3DSIH_IGE */ NULL
,
5210 /* WINED3DSIH_IMUL */ NULL
,
5211 /* WINED3DSIH_ITOF */ NULL
,
5212 /* WINED3DSIH_LABEL */ shader_hw_label
,
5213 /* WINED3DSIH_LD */ NULL
,
5214 /* WINED3DSIH_LIT */ shader_hw_map2gl
,
5215 /* WINED3DSIH_LOG */ shader_hw_scalar_op
,
5216 /* WINED3DSIH_LOGP */ shader_hw_scalar_op
,
5217 /* WINED3DSIH_LOOP */ shader_hw_loop
,
5218 /* WINED3DSIH_LRP */ shader_hw_lrp
,
5219 /* WINED3DSIH_LT */ NULL
,
5220 /* WINED3DSIH_M3x2 */ shader_hw_mnxn
,
5221 /* WINED3DSIH_M3x3 */ shader_hw_mnxn
,
5222 /* WINED3DSIH_M3x4 */ shader_hw_mnxn
,
5223 /* WINED3DSIH_M4x3 */ shader_hw_mnxn
,
5224 /* WINED3DSIH_M4x4 */ shader_hw_mnxn
,
5225 /* WINED3DSIH_MAD */ shader_hw_map2gl
,
5226 /* WINED3DSIH_MAX */ shader_hw_map2gl
,
5227 /* WINED3DSIH_MIN */ shader_hw_map2gl
,
5228 /* WINED3DSIH_MOV */ shader_hw_mov
,
5229 /* WINED3DSIH_MOVA */ shader_hw_mov
,
5230 /* WINED3DSIH_MOVC */ NULL
,
5231 /* WINED3DSIH_MUL */ shader_hw_map2gl
,
5232 /* WINED3DSIH_NOP */ shader_hw_nop
,
5233 /* WINED3DSIH_NRM */ shader_hw_nrm
,
5234 /* WINED3DSIH_PHASE */ shader_hw_nop
,
5235 /* WINED3DSIH_POW */ shader_hw_pow
,
5236 /* WINED3DSIH_RCP */ shader_hw_scalar_op
,
5237 /* WINED3DSIH_REP */ shader_hw_rep
,
5238 /* WINED3DSIH_RET */ shader_hw_ret
,
5239 /* WINED3DSIH_ROUND_NI */ NULL
,
5240 /* WINED3DSIH_RSQ */ shader_hw_scalar_op
,
5241 /* WINED3DSIH_SAMPLE */ NULL
,
5242 /* WINED3DSIH_SAMPLE_GRAD */ NULL
,
5243 /* WINED3DSIH_SAMPLE_LOD */ NULL
,
5244 /* WINED3DSIH_SETP */ NULL
,
5245 /* WINED3DSIH_SGE */ shader_hw_map2gl
,
5246 /* WINED3DSIH_SGN */ shader_hw_sgn
,
5247 /* WINED3DSIH_SINCOS */ shader_hw_sincos
,
5248 /* WINED3DSIH_SLT */ shader_hw_map2gl
,
5249 /* WINED3DSIH_SQRT */ NULL
,
5250 /* WINED3DSIH_SUB */ shader_hw_map2gl
,
5251 /* WINED3DSIH_TEX */ pshader_hw_tex
,
5252 /* WINED3DSIH_TEXBEM */ pshader_hw_texbem
,
5253 /* WINED3DSIH_TEXBEML */ pshader_hw_texbem
,
5254 /* WINED3DSIH_TEXCOORD */ pshader_hw_texcoord
,
5255 /* WINED3DSIH_TEXDEPTH */ pshader_hw_texdepth
,
5256 /* WINED3DSIH_TEXDP3 */ pshader_hw_texdp3
,
5257 /* WINED3DSIH_TEXDP3TEX */ pshader_hw_texdp3tex
,
5258 /* WINED3DSIH_TEXKILL */ pshader_hw_texkill
,
5259 /* WINED3DSIH_TEXLDD */ shader_hw_texldd
,
5260 /* WINED3DSIH_TEXLDL */ shader_hw_texldl
,
5261 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_hw_texm3x2depth
,
5262 /* WINED3DSIH_TEXM3x2PAD */ pshader_hw_texm3x2pad
,
5263 /* WINED3DSIH_TEXM3x2TEX */ pshader_hw_texm3x2tex
,
5264 /* WINED3DSIH_TEXM3x3 */ pshader_hw_texm3x3
,
5265 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
5266 /* WINED3DSIH_TEXM3x3PAD */ pshader_hw_texm3x3pad
,
5267 /* WINED3DSIH_TEXM3x3SPEC */ pshader_hw_texm3x3spec
,
5268 /* WINED3DSIH_TEXM3x3TEX */ pshader_hw_texm3x3tex
,
5269 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_hw_texm3x3vspec
,
5270 /* WINED3DSIH_TEXREG2AR */ pshader_hw_texreg2ar
,
5271 /* WINED3DSIH_TEXREG2GB */ pshader_hw_texreg2gb
,
5272 /* WINED3DSIH_TEXREG2RGB */ pshader_hw_texreg2rgb
,
5273 /* WINED3DSIH_UDIV */ NULL
,
5274 /* WINED3DSIH_USHR */ NULL
,
5275 /* WINED3DSIH_UTOF */ NULL
,
5276 /* WINED3DSIH_XOR */ NULL
,
5279 static BOOL
get_bool_const(const struct wined3d_shader_instruction
*ins
,
5280 const struct wined3d_shader
*shader
, DWORD idx
)
5282 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
5283 BOOL vshader
= shader_is_vshader_version(reg_maps
->shader_version
.type
);
5284 const struct wined3d_shader_lconst
*constant
;
5286 WORD flag
= (1 << idx
);
5287 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5289 if (reg_maps
->local_bool_consts
& flag
)
5291 /* What good is an if(bool) with a hardcoded local constant? I don't know, but handle it */
5292 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsB
, struct wined3d_shader_lconst
, entry
)
5294 if (constant
->idx
== idx
)
5296 return constant
->value
[0];
5299 ERR("Local constant not found\n");
5304 if(vshader
) bools
= priv
->cur_vs_args
->clip
.boolclip
.bools
;
5305 else bools
= priv
->cur_ps_args
->bools
;
5306 return bools
& flag
;
5310 static void get_loop_control_const(const struct wined3d_shader_instruction
*ins
,
5311 const struct wined3d_shader
*shader
, UINT idx
, struct wined3d_shader_loop_control
*loop_control
)
5313 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
5314 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5316 /* Integer constants can either be a local constant, or they can be stored in the shader
5317 * type specific compile args. */
5318 if (reg_maps
->local_int_consts
& (1 << idx
))
5320 const struct wined3d_shader_lconst
*constant
;
5322 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsI
, struct wined3d_shader_lconst
, entry
)
5324 if (constant
->idx
== idx
)
5326 loop_control
->count
= constant
->value
[0];
5327 loop_control
->start
= constant
->value
[1];
5328 /* Step is signed. */
5329 loop_control
->step
= (int)constant
->value
[2];
5333 /* If this happens the flag was set incorrectly */
5334 ERR("Local constant not found\n");
5335 loop_control
->count
= 0;
5336 loop_control
->start
= 0;
5337 loop_control
->step
= 0;
5341 switch (reg_maps
->shader_version
.type
)
5343 case WINED3D_SHADER_TYPE_VERTEX
:
5344 /* Count and aL start value are unsigned */
5345 loop_control
->count
= priv
->cur_vs_args
->loop_ctrl
[idx
][0];
5346 loop_control
->start
= priv
->cur_vs_args
->loop_ctrl
[idx
][1];
5347 /* Step is signed. */
5348 loop_control
->step
= ((char)priv
->cur_vs_args
->loop_ctrl
[idx
][2]);
5351 case WINED3D_SHADER_TYPE_PIXEL
:
5352 loop_control
->count
= priv
->cur_ps_args
->loop_ctrl
[idx
][0];
5353 loop_control
->start
= priv
->cur_ps_args
->loop_ctrl
[idx
][1];
5354 loop_control
->step
= ((char)priv
->cur_ps_args
->loop_ctrl
[idx
][2]);
5358 FIXME("Unhandled shader type %#x.\n", reg_maps
->shader_version
.type
);
5363 static void record_instruction(struct list
*list
, const struct wined3d_shader_instruction
*ins
)
5366 struct wined3d_shader_dst_param
*dst_param
= NULL
;
5367 struct wined3d_shader_src_param
*src_param
= NULL
, *rel_addr
= NULL
;
5368 struct recorded_instruction
*rec
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*rec
));
5371 ERR("Out of memory\n");
5376 dst_param
= HeapAlloc(GetProcessHeap(), 0, sizeof(*dst_param
));
5377 if(!dst_param
) goto free
;
5378 *dst_param
= *ins
->dst
;
5379 if (ins
->dst
->reg
.idx
[0].rel_addr
)
5381 rel_addr
= HeapAlloc(GetProcessHeap(), 0, sizeof(*rel_addr
));
5384 *rel_addr
= *ins
->dst
->reg
.idx
[0].rel_addr
;
5385 dst_param
->reg
.idx
[0].rel_addr
= rel_addr
;
5387 rec
->ins
.dst
= dst_param
;
5389 src_param
= HeapAlloc(GetProcessHeap(), 0, sizeof(*src_param
) * ins
->src_count
);
5392 for (i
= 0; i
< ins
->src_count
; ++i
)
5394 src_param
[i
] = ins
->src
[i
];
5395 if (ins
->src
[i
].reg
.idx
[0].rel_addr
)
5397 rel_addr
= HeapAlloc(GetProcessHeap(), 0, sizeof(*rel_addr
));
5400 *rel_addr
= *ins
->src
[i
].reg
.idx
[0].rel_addr
;
5401 src_param
[i
].reg
.idx
[0].rel_addr
= rel_addr
;
5404 rec
->ins
.src
= src_param
;
5405 list_add_tail(list
, &rec
->entry
);
5409 ERR("Out of memory\n");
5412 HeapFree(GetProcessHeap(), 0, (void *)dst_param
->reg
.idx
[0].rel_addr
);
5413 HeapFree(GetProcessHeap(), 0, dst_param
);
5417 for(i
= 0; i
< ins
->src_count
; i
++)
5419 HeapFree(GetProcessHeap(), 0, (void *)src_param
[i
].reg
.idx
[0].rel_addr
);
5421 HeapFree(GetProcessHeap(), 0, src_param
);
5423 HeapFree(GetProcessHeap(), 0, rec
);
5426 static void free_recorded_instruction(struct list
*list
)
5428 struct recorded_instruction
*rec_ins
, *entry2
;
5431 LIST_FOR_EACH_ENTRY_SAFE(rec_ins
, entry2
, list
, struct recorded_instruction
, entry
)
5433 list_remove(&rec_ins
->entry
);
5434 if (rec_ins
->ins
.dst
)
5436 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.dst
->reg
.idx
[0].rel_addr
);
5437 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.dst
);
5439 if (rec_ins
->ins
.src
)
5441 for (i
= 0; i
< rec_ins
->ins
.src_count
; ++i
)
5443 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.src
[i
].reg
.idx
[0].rel_addr
);
5445 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.src
);
5447 HeapFree(GetProcessHeap(), 0, rec_ins
);
5451 static void pop_control_frame(const struct wined3d_shader_instruction
*ins
)
5453 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5454 struct control_frame
*control_frame
;
5456 if (ins
->handler_idx
== WINED3DSIH_ENDLOOP
|| ins
->handler_idx
== WINED3DSIH_ENDREP
)
5458 struct list
*e
= list_head(&priv
->control_frames
);
5459 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5460 list_remove(&control_frame
->entry
);
5461 HeapFree(GetProcessHeap(), 0, control_frame
);
5464 else if (ins
->handler_idx
== WINED3DSIH_ENDIF
)
5466 /* Non-ifc ENDIFs were already handled previously. */
5467 struct list
*e
= list_head(&priv
->control_frames
);
5468 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5469 list_remove(&control_frame
->entry
);
5470 HeapFree(GetProcessHeap(), 0, control_frame
);
5474 static void shader_arb_handle_instruction(const struct wined3d_shader_instruction
*ins
) {
5475 SHADER_HANDLER hw_fct
;
5476 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5477 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
5478 struct control_frame
*control_frame
;
5479 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
5482 if(ins
->handler_idx
== WINED3DSIH_LOOP
|| ins
->handler_idx
== WINED3DSIH_REP
)
5484 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
5485 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5487 if(ins
->handler_idx
== WINED3DSIH_LOOP
) control_frame
->type
= LOOP
;
5488 if(ins
->handler_idx
== WINED3DSIH_REP
) control_frame
->type
= REP
;
5490 if(priv
->target_version
>= NV2
)
5492 control_frame
->no
.loop
= priv
->num_loops
++;
5497 /* Don't bother recording when we're in a not used if branch */
5503 if(!priv
->recording
)
5505 list_init(&priv
->record
);
5506 priv
->recording
= TRUE
;
5507 control_frame
->outer_loop
= TRUE
;
5508 get_loop_control_const(ins
, shader
, ins
->src
[0].reg
.idx
[0].offset
, &control_frame
->loop_control
);
5509 return; /* Instruction is handled */
5511 /* Record this loop in the outer loop's recording */
5514 else if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
|| ins
->handler_idx
== WINED3DSIH_ENDREP
)
5516 if(priv
->target_version
>= NV2
)
5518 /* Nothing to do. The control frame is popped after the HW instr handler */
5522 struct list
*e
= list_head(&priv
->control_frames
);
5523 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5524 list_remove(&control_frame
->entry
);
5526 if(control_frame
->outer_loop
)
5528 unsigned int iteration
;
5532 /* Turn off recording before playback */
5533 priv
->recording
= FALSE
;
5535 /* Move the recorded instructions to a separate list and get them out of the private data
5536 * structure. If there are nested loops, the shader_arb_handle_instruction below will
5537 * be recorded again, thus priv->record might be overwritten
5540 list_move_tail(©
, &priv
->record
);
5541 list_init(&priv
->record
);
5543 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5545 shader_addline(buffer
, "#unrolling loop: %u iterations, aL=%u, inc %d\n",
5546 control_frame
->loop_control
.count
, control_frame
->loop_control
.start
,
5547 control_frame
->loop_control
.step
);
5548 aL
= control_frame
->loop_control
.start
;
5552 shader_addline(buffer
, "#unrolling rep: %u iterations\n", control_frame
->loop_control
.count
);
5555 for (iteration
= 0; iteration
< control_frame
->loop_control
.count
; ++iteration
)
5557 struct recorded_instruction
*rec_ins
;
5558 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5561 shader_addline(buffer
, "#Iteration %u, aL=%d\n", iteration
, aL
);
5565 shader_addline(buffer
, "#Iteration %u\n", iteration
);
5568 LIST_FOR_EACH_ENTRY(rec_ins
, ©
, struct recorded_instruction
, entry
)
5570 shader_arb_handle_instruction(&rec_ins
->ins
);
5573 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5575 aL
+= control_frame
->loop_control
.step
;
5578 shader_addline(buffer
, "#end loop/rep\n");
5580 free_recorded_instruction(©
);
5581 HeapFree(GetProcessHeap(), 0, control_frame
);
5582 return; /* Instruction is handled */
5586 /* This is a nested loop. Proceed to the normal recording function */
5587 HeapFree(GetProcessHeap(), 0, control_frame
);
5594 record_instruction(&priv
->record
, ins
);
5599 if(ins
->handler_idx
== WINED3DSIH_IF
)
5601 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
5602 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5603 control_frame
->type
= IF
;
5605 bool_const
= get_bool_const(ins
, shader
, ins
->src
[0].reg
.idx
[0].offset
);
5606 if (ins
->src
[0].modifiers
== WINED3DSPSM_NOT
)
5607 bool_const
= !bool_const
;
5608 if (!priv
->muted
&& !bool_const
)
5610 shader_addline(buffer
, "#if(FALSE){\n");
5612 control_frame
->muting
= TRUE
;
5614 else shader_addline(buffer
, "#if(TRUE) {\n");
5616 return; /* Instruction is handled */
5618 else if(ins
->handler_idx
== WINED3DSIH_IFC
)
5620 /* IF(bool) and if_cond(a, b) use the same ELSE and ENDIF tokens */
5621 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
5622 control_frame
->type
= IFC
;
5623 control_frame
->no
.ifc
= priv
->num_ifcs
++;
5624 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5626 else if(ins
->handler_idx
== WINED3DSIH_ELSE
)
5628 struct list
*e
= list_head(&priv
->control_frames
);
5629 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5631 if(control_frame
->type
== IF
)
5633 shader_addline(buffer
, "#} else {\n");
5634 if(!priv
->muted
&& !control_frame
->muting
)
5637 control_frame
->muting
= TRUE
;
5639 else if(control_frame
->muting
) priv
->muted
= FALSE
;
5640 return; /* Instruction is handled. */
5642 /* In case of an ifc, generate a HW shader instruction */
5643 if (control_frame
->type
!= IFC
)
5644 ERR("Control frame does not match.\n");
5646 else if(ins
->handler_idx
== WINED3DSIH_ENDIF
)
5648 struct list
*e
= list_head(&priv
->control_frames
);
5649 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5651 if(control_frame
->type
== IF
)
5653 shader_addline(buffer
, "#} endif\n");
5654 if(control_frame
->muting
) priv
->muted
= FALSE
;
5655 list_remove(&control_frame
->entry
);
5656 HeapFree(GetProcessHeap(), 0, control_frame
);
5657 return; /* Instruction is handled */
5659 /* In case of an ifc, generate a HW shader instruction */
5660 if (control_frame
->type
!= IFC
)
5661 ERR("Control frame does not match.\n");
5666 pop_control_frame(ins
);
5670 /* Select handler */
5671 hw_fct
= shader_arb_instruction_handler_table
[ins
->handler_idx
];
5673 /* Unhandled opcode */
5676 FIXME("Backend can't handle opcode %#x\n", ins
->handler_idx
);
5681 pop_control_frame(ins
);
5683 shader_arb_add_instruction_modifiers(ins
);
5686 static BOOL
shader_arb_has_ffp_proj_control(void *shader_priv
)
5688 struct shader_arb_priv
*priv
= shader_priv
;
5690 return priv
->ffp_proj_control
;
5693 const struct wined3d_shader_backend_ops arb_program_shader_backend
=
5695 shader_arb_handle_instruction
,
5698 shader_arb_select_depth_blt
,
5699 shader_arb_deselect_depth_blt
,
5700 shader_arb_update_float_vertex_constants
,
5701 shader_arb_update_float_pixel_constants
,
5702 shader_arb_load_constants
,
5706 shader_arb_allocate_context_data
,
5707 shader_arb_free_context_data
,
5708 shader_arb_get_caps
,
5709 shader_arb_color_fixup_supported
,
5710 shader_arb_has_ffp_proj_control
,
5713 /* ARB_fragment_program fixed function pipeline replacement definitions */
5714 #define ARB_FFP_CONST_TFACTOR 0
5715 #define ARB_FFP_CONST_SPECULAR_ENABLE ((ARB_FFP_CONST_TFACTOR) + 1)
5716 #define ARB_FFP_CONST_CONSTANT(i) ((ARB_FFP_CONST_SPECULAR_ENABLE) + 1 + i)
5717 #define ARB_FFP_CONST_BUMPMAT(i) ((ARB_FFP_CONST_CONSTANT(7)) + 1 + i)
5718 #define ARB_FFP_CONST_LUMINANCE(i) ((ARB_FFP_CONST_BUMPMAT(7)) + 1 + i)
5720 struct arbfp_ffp_desc
5722 struct ffp_frag_desc parent
;
5726 /* Context activation is done by the caller. */
5727 static void arbfp_enable(const struct wined3d_gl_info
*gl_info
, BOOL enable
)
5731 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
5732 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
5736 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
5737 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
5741 static void *arbfp_alloc(const struct wined3d_shader_backend_ops
*shader_backend
, void *shader_priv
)
5743 struct shader_arb_priv
*priv
;
5745 /* Share private data between the shader backend and the pipeline
5746 * replacement, if both are the arb implementation. This is needed to
5747 * figure out whether ARBfp should be disabled if no pixel shader is bound
5749 if (shader_backend
== &arb_program_shader_backend
)
5751 else if (!(priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*priv
))))
5754 if (wine_rb_init(&priv
->fragment_shaders
, &wined3d_ffp_frag_program_rb_functions
) == -1)
5756 ERR("Failed to initialize rbtree.\n");
5757 if (priv
!= shader_priv
)
5758 HeapFree(GetProcessHeap(), 0, priv
);
5761 priv
->use_arbfp_fixed_func
= TRUE
;
5766 /* Context activation is done by the caller. */
5767 static void arbfp_free_ffpshader(struct wine_rb_entry
*entry
, void *context
)
5769 const struct wined3d_gl_info
*gl_info
= context
;
5770 struct arbfp_ffp_desc
*entry_arb
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_ffp_desc
, parent
.entry
);
5772 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb
->shader
));
5773 checkGLcall("glDeleteProgramsARB(1, &entry_arb->shader)");
5774 HeapFree(GetProcessHeap(), 0, entry_arb
);
5777 /* Context activation is done by the caller. */
5778 static void arbfp_free(struct wined3d_device
*device
)
5780 struct shader_arb_priv
*priv
= device
->fragment_priv
;
5782 wine_rb_destroy(&priv
->fragment_shaders
, arbfp_free_ffpshader
, &device
->adapter
->gl_info
);
5783 priv
->use_arbfp_fixed_func
= FALSE
;
5785 if (device
->shader_backend
!= &arb_program_shader_backend
)
5787 HeapFree(GetProcessHeap(), 0, device
->fragment_priv
);
5791 static void arbfp_get_caps(const struct wined3d_gl_info
*gl_info
, struct fragment_caps
*caps
)
5793 caps
->wined3d_caps
= WINED3D_FRAGMENT_CAP_PROJ_CONTROL
5794 | WINED3D_FRAGMENT_CAP_SRGB_WRITE
;
5795 caps
->PrimitiveMiscCaps
= WINED3DPMISCCAPS_TSSARGTEMP
;
5796 caps
->TextureOpCaps
= WINED3DTEXOPCAPS_DISABLE
|
5797 WINED3DTEXOPCAPS_SELECTARG1
|
5798 WINED3DTEXOPCAPS_SELECTARG2
|
5799 WINED3DTEXOPCAPS_MODULATE4X
|
5800 WINED3DTEXOPCAPS_MODULATE2X
|
5801 WINED3DTEXOPCAPS_MODULATE
|
5802 WINED3DTEXOPCAPS_ADDSIGNED2X
|
5803 WINED3DTEXOPCAPS_ADDSIGNED
|
5804 WINED3DTEXOPCAPS_ADD
|
5805 WINED3DTEXOPCAPS_SUBTRACT
|
5806 WINED3DTEXOPCAPS_ADDSMOOTH
|
5807 WINED3DTEXOPCAPS_BLENDCURRENTALPHA
|
5808 WINED3DTEXOPCAPS_BLENDFACTORALPHA
|
5809 WINED3DTEXOPCAPS_BLENDTEXTUREALPHA
|
5810 WINED3DTEXOPCAPS_BLENDDIFFUSEALPHA
|
5811 WINED3DTEXOPCAPS_BLENDTEXTUREALPHAPM
|
5812 WINED3DTEXOPCAPS_MODULATEALPHA_ADDCOLOR
|
5813 WINED3DTEXOPCAPS_MODULATECOLOR_ADDALPHA
|
5814 WINED3DTEXOPCAPS_MODULATEINVCOLOR_ADDALPHA
|
5815 WINED3DTEXOPCAPS_MODULATEINVALPHA_ADDCOLOR
|
5816 WINED3DTEXOPCAPS_DOTPRODUCT3
|
5817 WINED3DTEXOPCAPS_MULTIPLYADD
|
5818 WINED3DTEXOPCAPS_LERP
|
5819 WINED3DTEXOPCAPS_BUMPENVMAP
|
5820 WINED3DTEXOPCAPS_BUMPENVMAPLUMINANCE
;
5822 /* TODO: Implement WINED3DTEXOPCAPS_PREMODULATE */
5824 caps
->MaxTextureBlendStages
= 8;
5825 caps
->MaxSimultaneousTextures
= min(gl_info
->limits
.fragment_samplers
, 8);
5828 static void state_texfactor_arbfp(struct wined3d_context
*context
,
5829 const struct wined3d_state
*state
, DWORD state_id
)
5831 struct wined3d_device
*device
= context
->swapchain
->device
;
5832 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5835 /* Don't load the parameter if we're using an arbfp pixel shader,
5836 * otherwise we'll overwrite application provided constants. */
5837 if (device
->shader_backend
== &arb_program_shader_backend
)
5839 struct shader_arb_priv
*priv
;
5841 if (use_ps(state
)) return;
5843 priv
= device
->shader_priv
;
5844 priv
->pshader_const_dirty
[ARB_FFP_CONST_TFACTOR
] = 1;
5845 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_TFACTOR
+ 1);
5848 D3DCOLORTOGLFLOAT4(state
->render_states
[WINED3D_RS_TEXTUREFACTOR
], col
);
5849 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_TFACTOR
, col
));
5850 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_TFACTOR, col)");
5853 static void state_arb_specularenable(struct wined3d_context
*context
,
5854 const struct wined3d_state
*state
, DWORD state_id
)
5856 struct wined3d_device
*device
= context
->swapchain
->device
;
5857 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5860 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
5861 * application provided constants
5863 if (device
->shader_backend
== &arb_program_shader_backend
)
5865 struct shader_arb_priv
*priv
;
5867 if (use_ps(state
)) return;
5869 priv
= device
->shader_priv
;
5870 priv
->pshader_const_dirty
[ARB_FFP_CONST_SPECULAR_ENABLE
] = 1;
5871 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_SPECULAR_ENABLE
+ 1);
5874 if (state
->render_states
[WINED3D_RS_SPECULARENABLE
])
5876 /* The specular color has no alpha */
5877 col
[0] = 1.0f
; col
[1] = 1.0f
;
5878 col
[2] = 1.0f
; col
[3] = 0.0f
;
5880 col
[0] = 0.0f
; col
[1] = 0.0f
;
5881 col
[2] = 0.0f
; col
[3] = 0.0f
;
5883 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_SPECULAR_ENABLE
, col
));
5884 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_SPECULAR_ENABLE, col)");
5887 static void set_bumpmat_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
5889 DWORD stage
= (state_id
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
5890 struct wined3d_device
*device
= context
->swapchain
->device
;
5891 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5894 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_BUMP_ENV
;
5896 if (device
->shader_backend
== &arb_program_shader_backend
)
5898 struct shader_arb_priv
*priv
= device
->shader_priv
;
5900 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants. */
5904 priv
->pshader_const_dirty
[ARB_FFP_CONST_BUMPMAT(stage
)] = 1;
5905 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_BUMPMAT(stage
) + 1);
5908 mat
[0][0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT00
]);
5909 mat
[0][1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT01
]);
5910 mat
[1][0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT10
]);
5911 mat
[1][1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT11
]);
5913 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_BUMPMAT(stage
), &mat
[0][0]));
5914 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_BUMPMAT(stage), &mat[0][0])");
5917 static void tex_bumpenvlum_arbfp(struct wined3d_context
*context
,
5918 const struct wined3d_state
*state
, DWORD state_id
)
5920 DWORD stage
= (state_id
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
5921 struct wined3d_device
*device
= context
->swapchain
->device
;
5922 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5925 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_BUMP_ENV
;
5927 if (device
->shader_backend
== &arb_program_shader_backend
)
5929 struct shader_arb_priv
*priv
= device
->shader_priv
;
5931 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants. */
5935 priv
->pshader_const_dirty
[ARB_FFP_CONST_LUMINANCE(stage
)] = 1;
5936 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_LUMINANCE(stage
) + 1);
5939 param
[0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_LSCALE
]);
5940 param
[1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_LOFFSET
]);
5944 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_LUMINANCE(stage
), param
));
5945 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_LUMINANCE(stage), param)");
5948 static const char *get_argreg(struct wined3d_shader_buffer
*buffer
, DWORD argnum
, unsigned int stage
, DWORD arg
)
5952 if(arg
== ARG_UNUSED
) return "unused"; /* This is the marker for unused registers */
5954 switch(arg
& WINED3DTA_SELECTMASK
) {
5955 case WINED3DTA_DIFFUSE
:
5956 ret
= "fragment.color.primary"; break;
5958 case WINED3DTA_CURRENT
:
5959 if (!stage
) ret
= "fragment.color.primary";
5963 case WINED3DTA_TEXTURE
:
5965 case 0: ret
= "tex0"; break;
5966 case 1: ret
= "tex1"; break;
5967 case 2: ret
= "tex2"; break;
5968 case 3: ret
= "tex3"; break;
5969 case 4: ret
= "tex4"; break;
5970 case 5: ret
= "tex5"; break;
5971 case 6: ret
= "tex6"; break;
5972 case 7: ret
= "tex7"; break;
5973 default: ret
= "unknown texture";
5977 case WINED3DTA_TFACTOR
:
5978 ret
= "tfactor"; break;
5980 case WINED3DTA_SPECULAR
:
5981 ret
= "fragment.color.secondary"; break;
5983 case WINED3DTA_TEMP
:
5984 ret
= "tempreg"; break;
5986 case WINED3DTA_CONSTANT
:
5987 FIXME("Implement perstage constants\n");
5989 case 0: ret
= "const0"; break;
5990 case 1: ret
= "const1"; break;
5991 case 2: ret
= "const2"; break;
5992 case 3: ret
= "const3"; break;
5993 case 4: ret
= "const4"; break;
5994 case 5: ret
= "const5"; break;
5995 case 6: ret
= "const6"; break;
5996 case 7: ret
= "const7"; break;
5997 default: ret
= "unknown constant";
6005 if(arg
& WINED3DTA_COMPLEMENT
) {
6006 shader_addline(buffer
, "SUB arg%u, const.x, %s;\n", argnum
, ret
);
6007 if(argnum
== 0) ret
= "arg0";
6008 if(argnum
== 1) ret
= "arg1";
6009 if(argnum
== 2) ret
= "arg2";
6011 if(arg
& WINED3DTA_ALPHAREPLICATE
) {
6012 shader_addline(buffer
, "MOV arg%u, %s.w;\n", argnum
, ret
);
6013 if(argnum
== 0) ret
= "arg0";
6014 if(argnum
== 1) ret
= "arg1";
6015 if(argnum
== 2) ret
= "arg2";
6020 static void gen_ffp_instr(struct wined3d_shader_buffer
*buffer
, unsigned int stage
, BOOL color
,
6021 BOOL alpha
, DWORD dst
, DWORD op
, DWORD dw_arg0
, DWORD dw_arg1
, DWORD dw_arg2
)
6023 const char *dstmask
, *dstreg
, *arg0
, *arg1
, *arg2
;
6024 unsigned int mul
= 1;
6026 if(color
&& alpha
) dstmask
= "";
6027 else if(color
) dstmask
= ".xyz";
6028 else dstmask
= ".w";
6030 if(dst
== tempreg
) dstreg
= "tempreg";
6031 else dstreg
= "ret";
6033 arg0
= get_argreg(buffer
, 0, stage
, dw_arg0
);
6034 arg1
= get_argreg(buffer
, 1, stage
, dw_arg1
);
6035 arg2
= get_argreg(buffer
, 2, stage
, dw_arg2
);
6039 case WINED3D_TOP_DISABLE
:
6041 shader_addline(buffer
, "MOV %s%s, fragment.color.primary;\n", dstreg
, dstmask
);
6044 case WINED3D_TOP_SELECT_ARG2
:
6047 case WINED3D_TOP_SELECT_ARG1
:
6048 shader_addline(buffer
, "MOV %s%s, %s;\n", dstreg
, dstmask
, arg1
);
6051 case WINED3D_TOP_MODULATE_4X
:
6054 case WINED3D_TOP_MODULATE_2X
:
6057 case WINED3D_TOP_MODULATE
:
6058 shader_addline(buffer
, "MUL %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6061 case WINED3D_TOP_ADD_SIGNED_2X
:
6064 case WINED3D_TOP_ADD_SIGNED
:
6065 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
6068 case WINED3D_TOP_ADD
:
6069 shader_addline(buffer
, "ADD_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6072 case WINED3D_TOP_SUBTRACT
:
6073 shader_addline(buffer
, "SUB_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6076 case WINED3D_TOP_ADD_SMOOTH
:
6077 shader_addline(buffer
, "SUB arg1, const.x, %s;\n", arg1
);
6078 shader_addline(buffer
, "MAD_SAT %s%s, arg1, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6081 case WINED3D_TOP_BLEND_CURRENT_ALPHA
:
6082 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_CURRENT
);
6083 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6085 case WINED3D_TOP_BLEND_FACTOR_ALPHA
:
6086 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TFACTOR
);
6087 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6089 case WINED3D_TOP_BLEND_TEXTURE_ALPHA
:
6090 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
6091 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6093 case WINED3D_TOP_BLEND_DIFFUSE_ALPHA
:
6094 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_DIFFUSE
);
6095 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6098 case WINED3D_TOP_BLEND_TEXTURE_ALPHA_PM
:
6099 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
6100 shader_addline(buffer
, "SUB arg0.w, const.x, %s.w;\n", arg0
);
6101 shader_addline(buffer
, "MAD_SAT %s%s, %s, arg0.w, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6104 /* D3DTOP_PREMODULATE ???? */
6106 case WINED3D_TOP_MODULATE_INVALPHA_ADD_COLOR
:
6107 shader_addline(buffer
, "SUB arg0.w, const.x, %s;\n", arg1
);
6108 shader_addline(buffer
, "MAD_SAT %s%s, arg0.w, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6110 case WINED3D_TOP_MODULATE_ALPHA_ADD_COLOR
:
6111 shader_addline(buffer
, "MAD_SAT %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
6113 case WINED3D_TOP_MODULATE_INVCOLOR_ADD_ALPHA
:
6114 shader_addline(buffer
, "SUB arg0, const.x, %s;\n", arg1
);
6115 shader_addline(buffer
, "MAD_SAT %s%s, arg0, %s, %s.w;\n", dstreg
, dstmask
, arg2
, arg1
);
6117 case WINED3D_TOP_MODULATE_COLOR_ADD_ALPHA
:
6118 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s.w;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
6121 case WINED3D_TOP_DOTPRODUCT3
:
6123 shader_addline(buffer
, "SUB arg1, %s, const.w;\n", arg1
);
6124 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
6125 shader_addline(buffer
, "DP3_SAT %s%s, arg1, arg2;\n", dstreg
, dstmask
);
6128 case WINED3D_TOP_MULTIPLY_ADD
:
6129 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg0
);
6132 case WINED3D_TOP_LERP
:
6133 /* The msdn is not quite right here */
6134 shader_addline(buffer
, "LRP %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6137 case WINED3D_TOP_BUMPENVMAP
:
6138 case WINED3D_TOP_BUMPENVMAP_LUMINANCE
:
6139 /* Those are handled in the first pass of the shader(generation pass 1 and 2) already */
6143 FIXME("Unhandled texture op %08x\n", op
);
6147 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.y;\n", dstreg
, dstmask
, dstreg
);
6149 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.z;\n", dstreg
, dstmask
, dstreg
);
6152 static GLuint
gen_arbfp_ffp_shader(const struct ffp_frag_settings
*settings
, const struct wined3d_gl_info
*gl_info
)
6155 struct wined3d_shader_buffer buffer
;
6156 BOOL tex_read
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
6157 BOOL bump_used
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
6158 BOOL luminance_used
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
6159 UINT lowest_disabled_stage
;
6160 const char *textype
;
6162 char colorcor_dst
[8];
6164 DWORD arg0
, arg1
, arg2
;
6165 BOOL tempreg_used
= FALSE
, tfactor_used
= FALSE
;
6167 const char *final_combiner_src
= "ret";
6169 BOOL custom_linear_fog
= FALSE
;
6171 /* Find out which textures are read */
6172 for (stage
= 0; stage
< MAX_TEXTURES
; ++stage
)
6174 if (settings
->op
[stage
].cop
== WINED3D_TOP_DISABLE
)
6176 arg0
= settings
->op
[stage
].carg0
& WINED3DTA_SELECTMASK
;
6177 arg1
= settings
->op
[stage
].carg1
& WINED3DTA_SELECTMASK
;
6178 arg2
= settings
->op
[stage
].carg2
& WINED3DTA_SELECTMASK
;
6179 if(arg0
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6180 if(arg1
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6181 if(arg2
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6183 if (settings
->op
[stage
].cop
== WINED3D_TOP_BLEND_TEXTURE_ALPHA
)
6184 tex_read
[stage
] = TRUE
;
6185 if (settings
->op
[stage
].cop
== WINED3D_TOP_BLEND_TEXTURE_ALPHA_PM
)
6186 tex_read
[stage
] = TRUE
;
6187 if (settings
->op
[stage
].cop
== WINED3D_TOP_BUMPENVMAP
)
6189 bump_used
[stage
] = TRUE
;
6190 tex_read
[stage
] = TRUE
;
6192 if (settings
->op
[stage
].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
)
6194 bump_used
[stage
] = TRUE
;
6195 tex_read
[stage
] = TRUE
;
6196 luminance_used
[stage
] = TRUE
;
6198 else if (settings
->op
[stage
].cop
== WINED3D_TOP_BLEND_FACTOR_ALPHA
)
6200 tfactor_used
= TRUE
;
6203 if(arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
) {
6204 tfactor_used
= TRUE
;
6207 if(settings
->op
[stage
].dst
== tempreg
) tempreg_used
= TRUE
;
6208 if(arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
) {
6209 tempreg_used
= TRUE
;
6212 if (settings
->op
[stage
].aop
== WINED3D_TOP_DISABLE
)
6214 arg0
= settings
->op
[stage
].aarg0
& WINED3DTA_SELECTMASK
;
6215 arg1
= settings
->op
[stage
].aarg1
& WINED3DTA_SELECTMASK
;
6216 arg2
= settings
->op
[stage
].aarg2
& WINED3DTA_SELECTMASK
;
6217 if(arg0
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6218 if(arg1
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6219 if(arg2
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6221 if(arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
) {
6222 tempreg_used
= TRUE
;
6224 if(arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
) {
6225 tfactor_used
= TRUE
;
6228 lowest_disabled_stage
= stage
;
6231 if (!shader_buffer_init(&buffer
))
6233 ERR("Failed to initialize shader buffer.\n");
6237 shader_addline(&buffer
, "!!ARBfp1.0\n");
6239 switch (settings
->fog
)
6241 case WINED3D_FFP_PS_FOG_OFF
: break;
6242 case WINED3D_FFP_PS_FOG_LINEAR
:
6243 if (gl_info
->quirks
& WINED3D_QUIRK_BROKEN_ARB_FOG
)
6245 custom_linear_fog
= TRUE
;
6248 shader_addline(&buffer
, "OPTION ARB_fog_linear;\n");
6251 case WINED3D_FFP_PS_FOG_EXP
: shader_addline(&buffer
, "OPTION ARB_fog_exp;\n"); break;
6252 case WINED3D_FFP_PS_FOG_EXP2
: shader_addline(&buffer
, "OPTION ARB_fog_exp2;\n"); break;
6253 default: FIXME("Unexpected fog setting %d\n", settings
->fog
);
6256 shader_addline(&buffer
, "PARAM const = {1, 2, 4, 0.5};\n");
6257 shader_addline(&buffer
, "TEMP TMP;\n");
6258 shader_addline(&buffer
, "TEMP ret;\n");
6259 if(tempreg_used
|| settings
->sRGB_write
) shader_addline(&buffer
, "TEMP tempreg;\n");
6260 shader_addline(&buffer
, "TEMP arg0;\n");
6261 shader_addline(&buffer
, "TEMP arg1;\n");
6262 shader_addline(&buffer
, "TEMP arg2;\n");
6263 for(stage
= 0; stage
< MAX_TEXTURES
; stage
++) {
6264 if(!tex_read
[stage
]) continue;
6265 shader_addline(&buffer
, "TEMP tex%u;\n", stage
);
6266 if(!bump_used
[stage
]) continue;
6267 shader_addline(&buffer
, "PARAM bumpmat%u = program.env[%u];\n", stage
, ARB_FFP_CONST_BUMPMAT(stage
));
6268 if(!luminance_used
[stage
]) continue;
6269 shader_addline(&buffer
, "PARAM luminance%u = program.env[%u];\n", stage
, ARB_FFP_CONST_LUMINANCE(stage
));
6272 shader_addline(&buffer
, "PARAM tfactor = program.env[%u];\n", ARB_FFP_CONST_TFACTOR
);
6274 shader_addline(&buffer
, "PARAM specular_enable = program.env[%u];\n", ARB_FFP_CONST_SPECULAR_ENABLE
);
6276 if (settings
->sRGB_write
)
6278 shader_addline(&buffer
, "PARAM srgb_consts0 = ");
6279 shader_arb_append_imm_vec4(&buffer
, wined3d_srgb_const0
);
6280 shader_addline(&buffer
, ";\n");
6281 shader_addline(&buffer
, "PARAM srgb_consts1 = ");
6282 shader_arb_append_imm_vec4(&buffer
, wined3d_srgb_const1
);
6283 shader_addline(&buffer
, ";\n");
6286 if (lowest_disabled_stage
< 7 && settings
->emul_clipplanes
)
6287 shader_addline(&buffer
, "KIL fragment.texcoord[7];\n");
6289 /* Generate texture sampling instructions) */
6290 for (stage
= 0; stage
< MAX_TEXTURES
&& settings
->op
[stage
].cop
!= WINED3D_TOP_DISABLE
; ++stage
)
6292 if (!tex_read
[stage
])
6295 switch(settings
->op
[stage
].tex_type
) {
6296 case tex_1d
: textype
= "1D"; break;
6297 case tex_2d
: textype
= "2D"; break;
6298 case tex_3d
: textype
= "3D"; break;
6299 case tex_cube
: textype
= "CUBE"; break;
6300 case tex_rect
: textype
= "RECT"; break;
6301 default: textype
= "unexpected_textype"; break;
6304 if(settings
->op
[stage
].projected
== proj_none
) {
6306 } else if(settings
->op
[stage
].projected
== proj_count4
||
6307 settings
->op
[stage
].projected
== proj_count3
) {
6310 FIXME("Unexpected projection mode %d\n", settings
->op
[stage
].projected
);
6315 && (settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP
6316 || settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
))
6318 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, x, z, 0, 0;\n", stage
- 1);
6319 shader_addline(&buffer
, "DP3 ret.x, arg1, tex%u;\n", stage
- 1);
6320 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, y, w, 0, 0;\n", stage
- 1);
6321 shader_addline(&buffer
, "DP3 ret.y, arg1, tex%u;\n", stage
- 1);
6323 /* with projective textures, texbem only divides the static texture coord, not the displacement,
6324 * so multiply the displacement with the dividing parameter before passing it to TXP
6326 if (settings
->op
[stage
].projected
!= proj_none
) {
6327 if(settings
->op
[stage
].projected
== proj_count4
) {
6328 shader_addline(&buffer
, "MOV ret.w, fragment.texcoord[%u].w;\n", stage
);
6329 shader_addline(&buffer
, "MUL ret.xyz, ret, fragment.texcoord[%u].w, fragment.texcoord[%u];\n", stage
, stage
);
6331 shader_addline(&buffer
, "MOV ret.w, fragment.texcoord[%u].z;\n", stage
);
6332 shader_addline(&buffer
, "MAD ret.xyz, ret, fragment.texcoord[%u].z, fragment.texcoord[%u];\n", stage
, stage
);
6335 shader_addline(&buffer
, "ADD ret, ret, fragment.texcoord[%u];\n", stage
);
6338 shader_addline(&buffer
, "%s tex%u, ret, texture[%u], %s;\n",
6339 instr
, stage
, stage
, textype
);
6340 if (settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
)
6342 shader_addline(&buffer
, "MAD_SAT ret.x, tex%u.z, luminance%u.x, luminance%u.y;\n",
6343 stage
- 1, stage
- 1, stage
- 1);
6344 shader_addline(&buffer
, "MUL tex%u, tex%u, ret.x;\n", stage
, stage
);
6346 } else if(settings
->op
[stage
].projected
== proj_count3
) {
6347 shader_addline(&buffer
, "MOV ret, fragment.texcoord[%u];\n", stage
);
6348 shader_addline(&buffer
, "MOV ret.w, ret.z;\n");
6349 shader_addline(&buffer
, "%s tex%u, ret, texture[%u], %s;\n",
6350 instr
, stage
, stage
, textype
);
6352 shader_addline(&buffer
, "%s tex%u, fragment.texcoord[%u], texture[%u], %s;\n",
6353 instr
, stage
, stage
, stage
, textype
);
6356 sprintf(colorcor_dst
, "tex%u", stage
);
6357 gen_color_correction(&buffer
, colorcor_dst
, WINED3DSP_WRITEMASK_ALL
, "const.x", "const.y",
6358 settings
->op
[stage
].color_fixup
);
6361 /* Generate the main shader */
6362 for (stage
= 0; stage
< MAX_TEXTURES
; ++stage
)
6364 if (settings
->op
[stage
].cop
== WINED3D_TOP_DISABLE
)
6367 final_combiner_src
= "fragment.color.primary";
6371 if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG1
6372 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG1
)
6373 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
;
6374 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG1
6375 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG2
)
6376 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg2
;
6377 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG2
6378 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG1
)
6379 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg1
;
6380 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG2
6381 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG2
)
6382 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
6384 op_equal
= settings
->op
[stage
].aop
== settings
->op
[stage
].cop
6385 && settings
->op
[stage
].carg0
== settings
->op
[stage
].aarg0
6386 && settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
6387 && settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
6389 if (settings
->op
[stage
].aop
== WINED3D_TOP_DISABLE
)
6391 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].dst
,
6392 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6393 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6395 shader_addline(&buffer
, "MOV ret.w, fragment.color.primary.w;\n");
6399 gen_ffp_instr(&buffer
, stage
, TRUE
, TRUE
, settings
->op
[stage
].dst
,
6400 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6401 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6403 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].dst
,
6404 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6405 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6406 gen_ffp_instr(&buffer
, stage
, FALSE
, TRUE
, settings
->op
[stage
].dst
,
6407 settings
->op
[stage
].aop
, settings
->op
[stage
].aarg0
,
6408 settings
->op
[stage
].aarg1
, settings
->op
[stage
].aarg2
);
6412 if (settings
->sRGB_write
|| custom_linear_fog
)
6414 shader_addline(&buffer
, "MAD ret, fragment.color.secondary, specular_enable, %s;\n", final_combiner_src
);
6415 if (settings
->sRGB_write
)
6416 arbfp_add_sRGB_correction(&buffer
, "ret", "arg0", "arg1", "arg2", "tempreg", FALSE
);
6417 if (custom_linear_fog
)
6418 arbfp_add_linear_fog(&buffer
, "ret", "arg0");
6419 shader_addline(&buffer
, "MOV result.color, ret;\n");
6423 shader_addline(&buffer
, "MAD result.color, fragment.color.secondary, specular_enable, %s;\n",
6424 final_combiner_src
);
6428 shader_addline(&buffer
, "END\n");
6430 /* Generate the shader */
6431 GL_EXTCALL(glGenProgramsARB(1, &ret
));
6432 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, ret
));
6433 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
6434 strlen(buffer
.buffer
), buffer
.buffer
));
6435 checkGLcall("glProgramStringARB()");
6437 gl_info
->gl_ops
.gl
.p_glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
6440 FIXME("Fragment program error at position %d: %s\n\n", pos
,
6441 debugstr_a((const char *)gl_info
->gl_ops
.gl
.p_glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
6442 shader_arb_dump_program_source(buffer
.buffer
);
6448 GL_EXTCALL(glGetProgramivARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
6449 checkGLcall("glGetProgramivARB()");
6450 if (!native
) WARN("Program exceeds native resource limits.\n");
6453 shader_buffer_free(&buffer
);
6457 static void fragment_prog_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6459 const struct wined3d_device
*device
= context
->swapchain
->device
;
6460 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
6461 struct shader_arb_priv
*priv
= device
->fragment_priv
;
6462 BOOL use_pshader
= use_ps(state
);
6463 struct ffp_frag_settings settings
;
6464 const struct arbfp_ffp_desc
*desc
;
6467 TRACE("context %p, state %p, state_id %#x.\n", context
, state
, state_id
);
6469 if (isStateDirty(context
, STATE_RENDER(WINED3D_RS_FOGENABLE
)))
6471 if (!use_pshader
&& device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
)
6473 /* Reload fixed function constants since they collide with the
6474 * pixel shader constants. */
6475 for (i
= 0; i
< MAX_TEXTURES
; ++i
)
6477 set_bumpmat_arbfp(context
, state
, STATE_TEXTURESTAGE(i
, WINED3D_TSS_BUMPENV_MAT00
));
6479 state_texfactor_arbfp(context
, state
, STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
));
6480 state_arb_specularenable(context
, state
, STATE_RENDER(WINED3D_RS_SPECULARENABLE
));
6482 else if (use_pshader
)
6484 context
->shader_update_mask
|= 1 << WINED3D_SHADER_TYPE_PIXEL
;
6491 /* Find or create a shader implementing the fixed function pipeline
6492 * settings, then activate it. */
6493 gen_ffp_frag_op(context
, state
, &settings
, FALSE
);
6494 desc
= (const struct arbfp_ffp_desc
*)find_ffp_frag_shader(&priv
->fragment_shaders
, &settings
);
6496 struct arbfp_ffp_desc
*new_desc
= HeapAlloc(GetProcessHeap(), 0, sizeof(*new_desc
));
6499 ERR("Out of memory\n");
6503 new_desc
->parent
.settings
= settings
;
6504 new_desc
->shader
= gen_arbfp_ffp_shader(&settings
, gl_info
);
6505 add_ffp_frag_shader(&priv
->fragment_shaders
, &new_desc
->parent
);
6506 TRACE("Allocated fixed function replacement shader descriptor %p\n", new_desc
);
6510 /* Now activate the replacement program. GL_FRAGMENT_PROGRAM_ARB is already active (however, note the
6511 * comment above the shader_select call below). If e.g. GLSL is active, the shader_select call will
6514 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, desc
->shader
));
6515 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, desc->shader)");
6516 priv
->current_fprogram_id
= desc
->shader
;
6518 if (device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
)
6520 /* Reload fixed function constants since they collide with the
6521 * pixel shader constants. */
6522 for (i
= 0; i
< MAX_TEXTURES
; ++i
)
6524 set_bumpmat_arbfp(context
, state
, STATE_TEXTURESTAGE(i
, WINED3D_TSS_BUMPENV_MAT00
));
6526 state_texfactor_arbfp(context
, state
, STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
));
6527 state_arb_specularenable(context
, state
, STATE_RENDER(WINED3D_RS_SPECULARENABLE
));
6529 context
->last_was_pshader
= FALSE
;
6531 context
->last_was_pshader
= TRUE
;
6534 context
->shader_update_mask
|= 1 << WINED3D_SHADER_TYPE_PIXEL
;
6537 /* We can't link the fog states to the fragment state directly since the
6538 * vertex pipeline links them to FOGENABLE. A different linking in different
6539 * pipeline parts can't be expressed in the combined state table, so we need
6540 * to handle that with a forwarding function. The other invisible side effect
6541 * is that changing the fog start and fog end (which links to FOGENABLE in
6542 * vertex) results in the fragment_prog_arbfp function being called because
6543 * FOGENABLE is dirty, which calls this function here. */
6544 static void state_arbfp_fog(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6546 enum fogsource new_source
;
6547 DWORD fogstart
= state
->render_states
[WINED3D_RS_FOGSTART
];
6548 DWORD fogend
= state
->render_states
[WINED3D_RS_FOGEND
];
6550 TRACE("context %p, state %p, state_id %#x.\n", context
, state
, state_id
);
6552 if (!isStateDirty(context
, STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
)))
6553 fragment_prog_arbfp(context
, state
, state_id
);
6555 if (!state
->render_states
[WINED3D_RS_FOGENABLE
])
6558 if (state
->render_states
[WINED3D_RS_FOGTABLEMODE
] == WINED3D_FOG_NONE
)
6562 new_source
= FOGSOURCE_VS
;
6566 if (state
->render_states
[WINED3D_RS_FOGVERTEXMODE
] == WINED3D_FOG_NONE
|| context
->last_was_rhw
)
6567 new_source
= FOGSOURCE_COORD
;
6569 new_source
= FOGSOURCE_FFP
;
6574 new_source
= FOGSOURCE_FFP
;
6577 if (new_source
!= context
->fog_source
|| fogstart
== fogend
)
6579 context
->fog_source
= new_source
;
6580 state_fogstartend(context
, state
, STATE_RENDER(WINED3D_RS_FOGSTART
));
6584 static void textransform(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6586 if (!isStateDirty(context
, STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
)))
6587 fragment_prog_arbfp(context
, state
, state_id
);
6590 static const struct StateEntryTemplate arbfp_fragmentstate_template
[] =
6592 {STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
), { STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
), state_texfactor_arbfp
}, WINED3D_GL_EXT_NONE
},
6593 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6594 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6595 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6596 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6597 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6598 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6599 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6600 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6601 {STATE_TEXTURESTAGE(0, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6602 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6603 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6604 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6605 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6606 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6607 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6608 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6609 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6610 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6611 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6612 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6613 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6614 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6615 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6616 {STATE_TEXTURESTAGE(1, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6617 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6618 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6619 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6620 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6621 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6622 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6623 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6624 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6625 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6626 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6627 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6628 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6629 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6630 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6631 {STATE_TEXTURESTAGE(2, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6632 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6633 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6634 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6635 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6636 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6637 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6638 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6639 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6640 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6641 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6642 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6643 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6644 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6645 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6646 {STATE_TEXTURESTAGE(3, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6647 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6648 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6649 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6650 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6651 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6652 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6653 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6654 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6655 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6656 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6657 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6658 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6659 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6660 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6661 {STATE_TEXTURESTAGE(4, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6662 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6663 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6664 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6665 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6666 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6667 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6668 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6669 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6670 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6671 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6672 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6673 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6674 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6675 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6676 {STATE_TEXTURESTAGE(5, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6677 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6678 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6679 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6680 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6681 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6682 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6683 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6684 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6685 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6686 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6687 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6688 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6689 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6690 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6691 {STATE_TEXTURESTAGE(6, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6692 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6693 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6694 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6695 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6696 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6697 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6698 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6699 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6700 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6701 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6702 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6703 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6704 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6705 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6706 {STATE_TEXTURESTAGE(7, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6707 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6708 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6709 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6710 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6711 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6712 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6713 {STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
6714 {STATE_RENDER(WINED3D_RS_FOGENABLE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), state_arbfp_fog
}, WINED3D_GL_EXT_NONE
},
6715 {STATE_RENDER(WINED3D_RS_FOGTABLEMODE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6716 {STATE_RENDER(WINED3D_RS_FOGVERTEXMODE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6717 {STATE_RENDER(WINED3D_RS_FOGSTART
), { STATE_RENDER(WINED3D_RS_FOGSTART
), state_fogstartend
}, WINED3D_GL_EXT_NONE
},
6718 {STATE_RENDER(WINED3D_RS_FOGEND
), { STATE_RENDER(WINED3D_RS_FOGSTART
), NULL
}, WINED3D_GL_EXT_NONE
},
6719 {STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), { STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), state_srgbwrite
}, ARB_FRAMEBUFFER_SRGB
},
6720 {STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6721 {STATE_RENDER(WINED3D_RS_FOGCOLOR
), { STATE_RENDER(WINED3D_RS_FOGCOLOR
), state_fogcolor
}, WINED3D_GL_EXT_NONE
},
6722 {STATE_RENDER(WINED3D_RS_FOGDENSITY
), { STATE_RENDER(WINED3D_RS_FOGDENSITY
), state_fogdensity
}, WINED3D_GL_EXT_NONE
},
6723 {STATE_TEXTURESTAGE(0,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(0, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6724 {STATE_TEXTURESTAGE(1,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(1, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6725 {STATE_TEXTURESTAGE(2,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(2, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6726 {STATE_TEXTURESTAGE(3,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(3, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6727 {STATE_TEXTURESTAGE(4,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(4, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6728 {STATE_TEXTURESTAGE(5,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(5, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6729 {STATE_TEXTURESTAGE(6,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(6, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6730 {STATE_TEXTURESTAGE(7,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(7, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6731 {STATE_RENDER(WINED3D_RS_SPECULARENABLE
), { STATE_RENDER(WINED3D_RS_SPECULARENABLE
), state_arb_specularenable
}, WINED3D_GL_EXT_NONE
},
6732 {0 /* Terminate */, { 0, 0 }, WINED3D_GL_EXT_NONE
},
6735 const struct fragment_pipeline arbfp_fragment_pipeline
= {
6740 shader_arb_color_fixup_supported
,
6741 arbfp_fragmentstate_template
,
6744 struct arbfp_blit_type
6746 enum complex_fixup fixup
;
6750 struct arbfp_blit_desc
6753 struct arbfp_blit_type type
;
6754 struct wine_rb_entry entry
;
6757 struct arbfp_blit_priv
6759 struct wine_rb_tree shaders
;
6760 GLuint palette_texture
;
6763 static int arbfp_blit_type_compare(const void *key
, const struct wine_rb_entry
*entry
)
6765 const struct arbfp_blit_type
*ka
= key
;
6766 const struct arbfp_blit_type
*kb
= &WINE_RB_ENTRY_VALUE(entry
, const struct arbfp_blit_desc
, entry
)->type
;
6768 if (ka
->fixup
!= kb
->fixup
)
6769 return ka
->fixup
< kb
->fixup
? -1 : 1;
6770 if (ka
->textype
!= kb
->textype
)
6771 return ka
->textype
< kb
->textype
? -1 : 1;
6775 /* Context activation is done by the caller. */
6776 static void arbfp_free_blit_shader(struct wine_rb_entry
*entry
, void *context
)
6778 const struct wined3d_gl_info
*gl_info
= context
;
6779 struct arbfp_blit_desc
*entry_arb
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_blit_desc
, entry
);
6781 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb
->shader
));
6782 checkGLcall("glDeleteProgramsARB(1, &entry_arb->shader)");
6783 HeapFree(GetProcessHeap(), 0, entry_arb
);
6786 const struct wine_rb_functions wined3d_arbfp_blit_rb_functions
=
6791 arbfp_blit_type_compare
,
6794 static HRESULT
arbfp_blit_alloc(struct wined3d_device
*device
)
6796 struct arbfp_blit_priv
*priv
;
6798 if (!(priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*priv
))))
6799 return E_OUTOFMEMORY
;
6801 if (wine_rb_init(&priv
->shaders
, &wined3d_arbfp_blit_rb_functions
) == -1)
6803 ERR("Failed to initialize rbtree.\n");
6804 HeapFree(GetProcessHeap(), 0, priv
);
6805 return E_OUTOFMEMORY
;
6808 device
->blit_priv
= priv
;
6813 /* Context activation is done by the caller. */
6814 static void arbfp_blit_free(struct wined3d_device
*device
)
6816 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
6817 struct arbfp_blit_priv
*priv
= device
->blit_priv
;
6819 wine_rb_destroy(&priv
->shaders
, arbfp_free_blit_shader
, &device
->adapter
->gl_info
);
6820 checkGLcall("Delete blit programs");
6822 if (priv
->palette_texture
)
6823 gl_info
->gl_ops
.gl
.p_glDeleteTextures(1, &priv
->palette_texture
);
6825 HeapFree(GetProcessHeap(), 0, device
->blit_priv
);
6826 device
->blit_priv
= NULL
;
6829 static BOOL
gen_planar_yuv_read(struct wined3d_shader_buffer
*buffer
, enum complex_fixup fixup
,
6830 GLenum textype
, char *luminance
)
6833 const char *tex
, *texinstr
;
6835 if (fixup
== COMPLEX_FIXUP_UYVY
) {
6843 case GL_TEXTURE_2D
: tex
= "2D"; texinstr
= "TXP"; break;
6844 case GL_TEXTURE_RECTANGLE_ARB
: tex
= "RECT"; texinstr
= "TEX"; break;
6846 /* This is more tricky than just replacing the texture type - we have to navigate
6847 * properly in the texture to find the correct chroma values
6849 FIXME("Implement yuv correction for non-2d, non-rect textures\n");
6853 /* First we have to read the chroma values. This means we need at least two pixels(no filtering),
6854 * or 4 pixels(with filtering). To get the unmodified chromas, we have to rid ourselves of the
6855 * filtering when we sample the texture.
6857 * These are the rules for reading the chroma:
6863 * So we have to get the sampling x position in non-normalized coordinates in integers
6865 if(textype
!= GL_TEXTURE_RECTANGLE_ARB
) {
6866 shader_addline(buffer
, "MUL texcrd.xy, fragment.texcoord[0], size.x;\n");
6867 shader_addline(buffer
, "MOV texcrd.w, size.x;\n");
6869 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
6871 /* We must not allow filtering between pixel x and x+1, this would mix U and V
6872 * Vertical filtering is ok. However, bear in mind that the pixel center is at
6875 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
6876 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, coef.y;\n");
6878 /* Divide the x coordinate by 0.5 and get the fraction. This gives 0.25 and 0.75 for the
6879 * even and odd pixels respectively
6881 shader_addline(buffer
, "MUL texcrd2, texcrd, coef.y;\n");
6882 shader_addline(buffer
, "FRC texcrd2, texcrd2;\n");
6884 /* Sample Pixel 1 */
6885 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
6887 /* Put the value into either of the chroma values */
6888 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
6889 shader_addline(buffer
, "MUL chroma.x, luminance.%c, temp.x;\n", chroma
);
6890 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
6891 shader_addline(buffer
, "MUL chroma.y, luminance.%c, temp.x;\n", chroma
);
6893 /* Sample pixel 2. If we read an even pixel(SLT above returned 1), sample
6894 * the pixel right to the current one. Otherwise, sample the left pixel.
6895 * Bias and scale the SLT result to -1;1 and add it to the texcrd.x.
6897 shader_addline(buffer
, "MAD temp.x, temp.x, coef.z, -coef.x;\n");
6898 shader_addline(buffer
, "ADD texcrd.x, texcrd, temp.x;\n");
6899 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
6901 /* Put the value into the other chroma */
6902 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
6903 shader_addline(buffer
, "MAD chroma.y, luminance.%c, temp.x, chroma.y;\n", chroma
);
6904 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
6905 shader_addline(buffer
, "MAD chroma.x, luminance.%c, temp.x, chroma.x;\n", chroma
);
6907 /* TODO: If filtering is enabled, sample a 2nd pair of pixels left or right of
6908 * the current one and lerp the two U and V values
6911 /* This gives the correctly filtered luminance value */
6912 shader_addline(buffer
, "TEX luminance, fragment.texcoord[0], texture[0], %s;\n", tex
);
6917 static BOOL
gen_yv12_read(struct wined3d_shader_buffer
*buffer
, GLenum textype
, char *luminance
)
6920 static const float yv12_coef
[]
6921 = {2.0f
/ 3.0f
, 1.0f
/ 6.0f
, (2.0f
/ 3.0f
) + (1.0f
/ 6.0f
), 1.0f
/ 3.0f
};
6924 case GL_TEXTURE_2D
: tex
= "2D"; break;
6925 case GL_TEXTURE_RECTANGLE_ARB
: tex
= "RECT"; break;
6927 FIXME("Implement yv12 correction for non-2d, non-rect textures\n");
6931 /* YV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
6932 * V and a (W/2)x(H/2) U plane, each with 8 bit per pixel. So the effective
6933 * bitdepth is 12 bits per pixel. Since the U and V planes have only half the
6934 * pitch of the luminance plane, the packing into the gl texture is a bit
6935 * unfortunate. If the whole texture is interpreted as luminance data it looks
6936 * approximately like this:
6938 * +----------------------------------+----
6950 * +----------------+-----------------+----
6952 * | V even rows | V odd rows |
6954 * +----------------+------------------ -
6956 * | U even rows | U odd rows |
6958 * +----------------+-----------------+----
6962 * So it appears as if there are 4 chroma images, but in fact the odd rows
6963 * in the chroma images are in the same row as the even ones. So its is
6964 * kinda tricky to read
6966 * When reading from rectangle textures, keep in mind that the input y coordinates
6967 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height
6969 shader_addline(buffer
, "PARAM yv12_coef = ");
6970 shader_arb_append_imm_vec4(buffer
, yv12_coef
);
6971 shader_addline(buffer
, ";\n");
6973 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
6974 /* the chroma planes have only half the width */
6975 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, coef.y;\n");
6977 /* The first value is between 2/3 and 5/6th of the texture's height, so scale+bias
6978 * the coordinate. Also read the right side of the image when reading odd lines
6980 * Don't forget to clamp the y values in into the range, otherwise we'll get filtering
6983 if(textype
== GL_TEXTURE_2D
) {
6985 shader_addline(buffer
, "RCP chroma.w, size.y;\n");
6987 shader_addline(buffer
, "MUL texcrd2.y, texcrd.y, size.y;\n");
6989 shader_addline(buffer
, "FLR texcrd2.y, texcrd2.y;\n");
6990 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, yv12_coef.y, yv12_coef.x;\n");
6992 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
6993 shader_addline(buffer
, "ADD texcrd2.x, texcrd2.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
6994 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
6995 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
6996 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
6998 /* clamp, keep the half pixel origin in mind */
6999 shader_addline(buffer
, "MAD temp.y, coef.y, chroma.w, yv12_coef.x;\n");
7000 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7001 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.z;\n");
7002 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7004 /* Read from [size - size+size/4] */
7005 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
7006 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, coef.w, size.y;\n");
7008 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
7009 shader_addline(buffer
, "ADD texcrd2.x, texcrd.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
7010 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
7011 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
7012 shader_addline(buffer
, "MUL texcrd2.x, texcrd2.x, size.x;\n");
7013 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
7015 /* Make sure to read exactly from the pixel center */
7016 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
7017 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, coef.y;\n");
7020 shader_addline(buffer
, "MAD temp.y, size.y, coef.w, size.y;\n");
7021 shader_addline(buffer
, "ADD temp.y, temp.y, -coef.y;\n");
7022 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7023 shader_addline(buffer
, "ADD temp.y, size.y, coef.y;\n");
7024 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7026 /* Read the texture, put the result into the output register */
7027 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7028 shader_addline(buffer
, "MOV chroma.x, temp.w;\n");
7030 /* The other chroma value is 1/6th of the texture lower, from 5/6th to 6/6th
7031 * No need to clamp because we're just reusing the already clamped value from above
7033 if(textype
== GL_TEXTURE_2D
) {
7034 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, yv12_coef.y;\n");
7036 shader_addline(buffer
, "MAD texcrd.y, size.y, coef.w, texcrd.y;\n");
7038 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7039 shader_addline(buffer
, "MOV chroma.y, temp.w;\n");
7041 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
7042 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
7043 * values due to filtering
7045 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7046 if(textype
== GL_TEXTURE_2D
) {
7047 /* Multiply the y coordinate by 2/3 and clamp it */
7048 shader_addline(buffer
, "MUL texcrd.y, texcrd.y, yv12_coef.x;\n");
7049 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.x;\n");
7050 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7051 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7053 /* Reading from texture_rectangles is pretty straightforward, just use the unmodified
7054 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
7057 shader_addline(buffer
, "ADD temp.x, size.y, -coef.y;\n");
7058 shader_addline(buffer
, "MIN texcrd.y, texcrd.y, size.x;\n");
7059 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7066 static BOOL
gen_nv12_read(struct wined3d_shader_buffer
*buffer
, GLenum textype
,
7070 static const float nv12_coef
[]
7071 = {2.0f
/ 3.0f
, 1.0f
/ 3.0f
, 1.0f
, 1.0f
};
7078 case GL_TEXTURE_RECTANGLE_ARB
:
7082 FIXME("Implement nv12 correction for non-2d, non-rect textures\n");
7086 /* NV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
7087 * sized plane where each component is an UV pair. So the effective
7088 * bitdepth is 12 bits per pixel If the whole texture is interpreted as luminance
7089 * data it looks approximately like this:
7091 * +----------------------------------+----
7103 * +----------------------------------+----
7104 * |UVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUV|
7105 * |UVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUV|
7111 * +----------------------------------+----
7113 * When reading from rectangle textures, keep in mind that the input y coordinates
7114 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height. */
7116 shader_addline(buffer
, "PARAM nv12_coef = ");
7117 shader_arb_append_imm_vec4(buffer
, nv12_coef
);
7118 shader_addline(buffer
, ";\n");
7120 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7121 /* We only have half the number of chroma pixels. */
7122 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, coef.y;\n");
7124 if (textype
== GL_TEXTURE_2D
)
7126 shader_addline(buffer
, "RCP chroma.w, size.x;\n");
7127 shader_addline(buffer
, "RCP chroma.z, size.y;\n");
7129 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, nv12_coef.y, nv12_coef.x;\n");
7131 /* We must not allow filtering horizontally, this would mix U and V.
7132 * Vertical filtering is ok. However, bear in mind that the pixel center is at
7133 * 0.5, so add 0.5. */
7135 /* Convert to non-normalized coordinates so we can find the
7136 * individual pixel. */
7137 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, size.x;\n");
7138 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
7139 /* Multiply by 2 since chroma components are stored in UV pixel pairs,
7140 * add 0.5 to hit the center of the pixel. */
7141 shader_addline(buffer
, "MAD texcrd.x, texcrd.x, coef.z, coef.y;\n");
7143 /* Convert back to normalized coordinates. */
7144 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, chroma.w;\n");
7146 /* Clamp, keep the half pixel origin in mind. */
7147 shader_addline(buffer
, "MAD temp.y, coef.y, chroma.z, nv12_coef.x;\n");
7148 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7149 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.z, nv12_coef.z;\n");
7150 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7154 /* Read from [size - size+size/2] */
7155 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, coef.y, size.y;\n");
7157 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
7158 /* Multiply by 2 since chroma components are stored in UV pixel pairs,
7159 * add 0.5 to hit the center of the pixel. */
7160 shader_addline(buffer
, "MAD texcrd.x, texcrd.x, coef.z, coef.y;\n");
7163 shader_addline(buffer
, "MAD temp.y, size.y, coef.y, size.y;\n");
7164 shader_addline(buffer
, "ADD temp.y, temp.y, -coef.y;\n");
7165 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7166 shader_addline(buffer
, "ADD temp.y, size.y, coef.y;\n");
7167 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7169 /* Read the texture, put the result into the output register. */
7170 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7171 shader_addline(buffer
, "MOV chroma.y, temp.w;\n");
7173 if (textype
== GL_TEXTURE_2D
)
7176 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, chroma.w;\n");
7181 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, coef.x;\n");
7183 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7184 shader_addline(buffer
, "MOV chroma.x, temp.w;\n");
7186 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
7187 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
7188 * values due to filtering. */
7189 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7190 if (textype
== GL_TEXTURE_2D
)
7192 /* Multiply the y coordinate by 2/3 and clamp it */
7193 shader_addline(buffer
, "MUL texcrd.y, texcrd.y, nv12_coef.x;\n");
7194 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, nv12_coef.x;\n");
7195 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7196 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7200 /* Reading from texture_rectangles is pretty straightforward, just use the unmodified
7201 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
7204 shader_addline(buffer
, "ADD temp.x, size.y, -coef.y;\n");
7205 shader_addline(buffer
, "MIN texcrd.y, texcrd.y, size.x;\n");
7206 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7213 /* Context activation is done by the caller. */
7214 static GLuint
gen_p8_shader(struct arbfp_blit_priv
*priv
,
7215 const struct wined3d_gl_info
*gl_info
, GLenum textype
)
7218 struct wined3d_shader_buffer buffer
;
7222 if (!shader_buffer_init(&buffer
))
7224 ERR("Failed to initialize shader buffer.\n");
7228 GL_EXTCALL(glGenProgramsARB(1, &shader
));
7229 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7232 shader_buffer_free(&buffer
);
7236 shader_addline(&buffer
, "!!ARBfp1.0\n");
7237 shader_addline(&buffer
, "TEMP index;\n");
7239 /* { 255/256, 0.5/255*255/256, 0, 0 } */
7240 shader_addline(&buffer
, "PARAM constants = { 0.996, 0.00195, 0, 0 };\n");
7242 /* The alpha-component contains the palette index */
7243 if(textype
== GL_TEXTURE_RECTANGLE_ARB
)
7244 shader_addline(&buffer
, "TXP index, fragment.texcoord[0], texture[0], RECT;\n");
7246 shader_addline(&buffer
, "TEX index, fragment.texcoord[0], texture[0], 2D;\n");
7248 /* Scale the index by 255/256 and add a bias of '0.5' in order to sample in the middle */
7249 shader_addline(&buffer
, "MAD index.a, index.a, constants.x, constants.y;\n");
7251 /* Use the alpha-component as an index in the palette to get the final color */
7252 shader_addline(&buffer
, "TEX result.color, index.a, texture[1], 1D;\n");
7253 shader_addline(&buffer
, "END\n");
7255 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
7256 strlen(buffer
.buffer
), buffer
.buffer
));
7257 checkGLcall("glProgramStringARB()");
7259 gl_info
->gl_ops
.gl
.p_glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
7262 FIXME("Fragment program error at position %d: %s\n\n", pos
,
7263 debugstr_a((const char *)gl_info
->gl_ops
.gl
.p_glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
7264 shader_arb_dump_program_source(buffer
.buffer
);
7267 shader_buffer_free(&buffer
);
7272 /* Context activation is done by the caller. */
7273 static void upload_palette(const struct wined3d_surface
*surface
, struct wined3d_context
*context
)
7276 struct wined3d_device
*device
= surface
->resource
.device
;
7277 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
7278 struct arbfp_blit_priv
*priv
= device
->blit_priv
;
7279 BOOL colorkey
= !!(surface
->container
->color_key_flags
& WINEDDSD_CKSRCBLT
);
7281 d3dfmt_p8_init_palette(surface
, table
, colorkey
);
7283 if (!priv
->palette_texture
)
7284 gl_info
->gl_ops
.gl
.p_glGenTextures(1, &priv
->palette_texture
);
7286 GL_EXTCALL(glActiveTextureARB(GL_TEXTURE1
));
7287 gl_info
->gl_ops
.gl
.p_glBindTexture(GL_TEXTURE_1D
, priv
->palette_texture
);
7289 gl_info
->gl_ops
.gl
.p_glTexEnvi(GL_TEXTURE_ENV
, GL_TEXTURE_ENV_MODE
, GL_REPLACE
);
7291 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_MIN_FILTER
, GL_NEAREST
);
7292 /* Make sure we have discrete color levels. */
7293 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_MAG_FILTER
, GL_NEAREST
);
7294 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_WRAP_S
, GL_CLAMP_TO_EDGE
);
7295 /* Upload the palette */
7296 /* TODO: avoid unneeded uploads in the future by adding some SFLAG_PALETTE_DIRTY mechanism */
7297 gl_info
->gl_ops
.gl
.p_glTexImage1D(GL_TEXTURE_1D
, 0, GL_RGBA
, 256, 0, GL_RGBA
, GL_UNSIGNED_BYTE
, table
);
7299 /* Switch back to unit 0 in which the 2D texture will be stored. */
7300 context_active_texture(context
, gl_info
, 0);
7303 /* Context activation is done by the caller. */
7304 static GLuint
gen_yuv_shader(struct arbfp_blit_priv
*priv
, const struct wined3d_gl_info
*gl_info
,
7305 enum complex_fixup yuv_fixup
, GLenum textype
)
7308 struct wined3d_shader_buffer buffer
;
7309 char luminance_component
;
7313 if (!shader_buffer_init(&buffer
))
7315 ERR("Failed to initialize shader buffer.\n");
7319 GL_EXTCALL(glGenProgramsARB(1, &shader
));
7320 checkGLcall("GL_EXTCALL(glGenProgramsARB(1, &shader))");
7321 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7322 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
7325 shader_buffer_free(&buffer
);
7329 /* The YUY2 and UYVY formats contain two pixels packed into a 32 bit macropixel,
7330 * giving effectively 16 bit per pixel. The color consists of a luminance(Y) and
7331 * two chroma(U and V) values. Each macropixel has two luminance values, one for
7332 * each single pixel it contains, and one U and one V value shared between both
7335 * The data is loaded into an A8L8 texture. With YUY2, the luminance component
7336 * contains the luminance and alpha the chroma. With UYVY it is vice versa. Thus
7337 * take the format into account when generating the read swizzles
7339 * Reading the Y value is straightforward - just sample the texture. The hardware
7340 * takes care of filtering in the horizontal and vertical direction.
7342 * Reading the U and V values is harder. We have to avoid filtering horizontally,
7343 * because that would mix the U and V values of one pixel or two adjacent pixels.
7344 * Thus floor the texture coordinate and add 0.5 to get an unfiltered read,
7345 * regardless of the filtering setting. Vertical filtering works automatically
7346 * though - the U and V values of two rows are mixed nicely.
7348 * Apart of avoiding filtering issues, the code has to know which value it just
7349 * read, and where it can find the other one. To determine this, it checks if
7350 * it sampled an even or odd pixel, and shifts the 2nd read accordingly.
7352 * Handling horizontal filtering of U and V values requires reading a 2nd pair
7353 * of pixels, extracting U and V and mixing them. This is not implemented yet.
7355 * An alternative implementation idea is to load the texture as A8R8G8B8 texture,
7356 * with width / 2. This way one read gives all 3 values, finding U and V is easy
7357 * in an unfiltered situation. Finding the luminance on the other hand requires
7358 * finding out if it is an odd or even pixel. The real drawback of this approach
7359 * is filtering. This would have to be emulated completely in the shader, reading
7360 * up two 2 packed pixels in up to 2 rows and interpolating both horizontally and
7361 * vertically. Beyond that it would require adjustments to the texture handling
7362 * code to deal with the width scaling
7364 shader_addline(&buffer
, "!!ARBfp1.0\n");
7365 shader_addline(&buffer
, "TEMP luminance;\n");
7366 shader_addline(&buffer
, "TEMP temp;\n");
7367 shader_addline(&buffer
, "TEMP chroma;\n");
7368 shader_addline(&buffer
, "TEMP texcrd;\n");
7369 shader_addline(&buffer
, "TEMP texcrd2;\n");
7370 shader_addline(&buffer
, "PARAM coef = {1.0, 0.5, 2.0, 0.25};\n");
7371 shader_addline(&buffer
, "PARAM yuv_coef = {1.403, 0.344, 0.714, 1.770};\n");
7372 shader_addline(&buffer
, "PARAM size = program.local[0];\n");
7376 case COMPLEX_FIXUP_UYVY
:
7377 case COMPLEX_FIXUP_YUY2
:
7378 if (!gen_planar_yuv_read(&buffer
, yuv_fixup
, textype
, &luminance_component
))
7380 shader_buffer_free(&buffer
);
7385 case COMPLEX_FIXUP_YV12
:
7386 if (!gen_yv12_read(&buffer
, textype
, &luminance_component
))
7388 shader_buffer_free(&buffer
);
7393 case COMPLEX_FIXUP_NV12
:
7394 if (!gen_nv12_read(&buffer
, textype
, &luminance_component
))
7396 shader_buffer_free(&buffer
);
7402 FIXME("Unsupported YUV fixup %#x\n", yuv_fixup
);
7403 shader_buffer_free(&buffer
);
7407 /* Calculate the final result. Formula is taken from
7408 * http://www.fourcc.org/fccyvrgb.php. Note that the chroma
7409 * ranges from -0.5 to 0.5
7411 shader_addline(&buffer
, "SUB chroma.xy, chroma, coef.y;\n");
7413 shader_addline(&buffer
, "MAD result.color.x, chroma.x, yuv_coef.x, luminance.%c;\n", luminance_component
);
7414 shader_addline(&buffer
, "MAD temp.x, -chroma.y, yuv_coef.y, luminance.%c;\n", luminance_component
);
7415 shader_addline(&buffer
, "MAD result.color.y, -chroma.x, yuv_coef.z, temp.x;\n");
7416 shader_addline(&buffer
, "MAD result.color.z, chroma.y, yuv_coef.w, luminance.%c;\n", luminance_component
);
7417 shader_addline(&buffer
, "END\n");
7419 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
7420 strlen(buffer
.buffer
), buffer
.buffer
));
7421 checkGLcall("glProgramStringARB()");
7423 gl_info
->gl_ops
.gl
.p_glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
7426 FIXME("Fragment program error at position %d: %s\n\n", pos
,
7427 debugstr_a((const char *)gl_info
->gl_ops
.gl
.p_glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
7428 shader_arb_dump_program_source(buffer
.buffer
);
7434 GL_EXTCALL(glGetProgramivARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
7435 checkGLcall("glGetProgramivARB()");
7436 if (!native
) WARN("Program exceeds native resource limits.\n");
7439 shader_buffer_free(&buffer
);
7444 /* Context activation is done by the caller. */
7445 static HRESULT
arbfp_blit_set(void *blit_priv
, struct wined3d_context
*context
, const struct wined3d_surface
*surface
)
7448 float size
[4] = {(float) surface
->pow2Width
, (float) surface
->pow2Height
, 1.0f
, 1.0f
};
7449 struct arbfp_blit_priv
*priv
= blit_priv
;
7450 enum complex_fixup fixup
;
7451 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
7452 GLenum textype
= surface
->container
->target
;
7453 struct wine_rb_entry
*entry
;
7454 struct arbfp_blit_type type
;
7455 struct arbfp_blit_desc
*desc
;
7457 if (surface
->flags
& SFLAG_CONVERTED
)
7459 gl_info
->gl_ops
.gl
.p_glEnable(textype
);
7460 checkGLcall("glEnable(textype)");
7464 if (!is_complex_fixup(surface
->resource
.format
->color_fixup
))
7467 dump_color_fixup_desc(surface
->resource
.format
->color_fixup
);
7468 /* Don't bother setting up a shader for unconverted formats */
7469 gl_info
->gl_ops
.gl
.p_glEnable(textype
);
7470 checkGLcall("glEnable(textype)");
7474 fixup
= get_complex_fixup(surface
->resource
.format
->color_fixup
);
7477 type
.textype
= textype
;
7478 entry
= wine_rb_get(&priv
->shaders
, &type
);
7481 desc
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_blit_desc
, entry
);
7482 shader
= desc
->shader
;
7488 case COMPLEX_FIXUP_P8
:
7489 shader
= gen_p8_shader(priv
, gl_info
, textype
);
7493 shader
= gen_yuv_shader(priv
, gl_info
, fixup
, textype
);
7499 FIXME("Unsupported complex fixup %#x, not setting a shader\n", fixup
);
7500 gl_info
->gl_ops
.gl
.p_glEnable(textype
);
7501 checkGLcall("glEnable(textype)");
7505 desc
= HeapAlloc(GetProcessHeap(), 0, sizeof(*desc
));
7509 desc
->type
.textype
= textype
;
7510 desc
->type
.fixup
= fixup
;
7511 desc
->shader
= shader
;
7512 if (wine_rb_put(&priv
->shaders
, &desc
->type
, &desc
->entry
) == -1)
7515 ERR("Out of memory\n");
7516 GL_EXTCALL(glDeleteProgramsARB(1, &shader
));
7517 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader))");
7518 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, 0));
7519 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, 0)");
7520 HeapFree(GetProcessHeap(), 0, desc
);
7521 return E_OUTOFMEMORY
;
7525 if (fixup
== COMPLEX_FIXUP_P8
)
7526 upload_palette(surface
, context
);
7528 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
7529 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
7530 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7531 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
7532 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, 0, size
));
7533 checkGLcall("glProgramLocalParameter4fvARB");
7538 /* Context activation is done by the caller. */
7539 static void arbfp_blit_unset(const struct wined3d_gl_info
*gl_info
)
7541 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
7542 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
7543 gl_info
->gl_ops
.gl
.p_glDisable(GL_TEXTURE_2D
);
7544 checkGLcall("glDisable(GL_TEXTURE_2D)");
7545 if (gl_info
->supported
[ARB_TEXTURE_CUBE_MAP
])
7547 gl_info
->gl_ops
.gl
.p_glDisable(GL_TEXTURE_CUBE_MAP_ARB
);
7548 checkGLcall("glDisable(GL_TEXTURE_CUBE_MAP_ARB)");
7550 if (gl_info
->supported
[ARB_TEXTURE_RECTANGLE
])
7552 gl_info
->gl_ops
.gl
.p_glDisable(GL_TEXTURE_RECTANGLE_ARB
);
7553 checkGLcall("glDisable(GL_TEXTURE_RECTANGLE_ARB)");
7557 static BOOL
arbfp_blit_supported(const struct wined3d_gl_info
*gl_info
, enum wined3d_blit_op blit_op
,
7558 const RECT
*src_rect
, DWORD src_usage
, enum wined3d_pool src_pool
, const struct wined3d_format
*src_format
,
7559 const RECT
*dst_rect
, DWORD dst_usage
, enum wined3d_pool dst_pool
, const struct wined3d_format
*dst_format
)
7561 enum complex_fixup src_fixup
;
7563 if (!gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
7566 if (blit_op
!= WINED3D_BLIT_OP_COLOR_BLIT
)
7568 TRACE("Unsupported blit_op=%d\n", blit_op
);
7572 if (src_pool
== WINED3D_POOL_SYSTEM_MEM
|| dst_pool
== WINED3D_POOL_SYSTEM_MEM
)
7575 src_fixup
= get_complex_fixup(src_format
->color_fixup
);
7576 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
7578 TRACE("Checking support for fixup:\n");
7579 dump_color_fixup_desc(src_format
->color_fixup
);
7582 if (!is_identity_fixup(dst_format
->color_fixup
))
7584 TRACE("Destination fixups are not supported\n");
7588 if (is_identity_fixup(src_format
->color_fixup
))
7594 /* We only support YUV conversions. */
7595 if (!is_complex_fixup(src_format
->color_fixup
))
7597 TRACE("[FAILED]\n");
7603 case COMPLEX_FIXUP_YUY2
:
7604 case COMPLEX_FIXUP_UYVY
:
7605 case COMPLEX_FIXUP_YV12
:
7606 case COMPLEX_FIXUP_NV12
:
7607 case COMPLEX_FIXUP_P8
:
7612 FIXME("Unsupported YUV fixup %#x\n", src_fixup
);
7613 TRACE("[FAILED]\n");
7618 HRESULT
arbfp_blit_surface(struct wined3d_device
*device
, DWORD filter
,
7619 struct wined3d_surface
*src_surface
, const RECT
*src_rect_in
,
7620 struct wined3d_surface
*dst_surface
, const RECT
*dst_rect_in
)
7622 struct wined3d_context
*context
;
7623 RECT src_rect
= *src_rect_in
;
7624 RECT dst_rect
= *dst_rect_in
;
7626 /* Activate the destination context, set it up for blitting */
7627 context
= context_acquire(device
, dst_surface
);
7629 /* Now load the surface */
7630 if (wined3d_settings
.offscreen_rendering_mode
!= ORM_FBO
7631 && (src_surface
->locations
& (WINED3D_LOCATION_TEXTURE_RGB
| WINED3D_LOCATION_DRAWABLE
))
7632 == WINED3D_LOCATION_DRAWABLE
7633 && !surface_is_offscreen(src_surface
))
7635 /* Without FBO blits transferring from the drawable to the texture is
7636 * expensive, because we have to flip the data in sysmem. Since we can
7637 * flip in the blitter, we don't actually need that flip anyway. So we
7638 * use the surface's texture as scratch texture, and flip the source
7639 * rectangle instead. */
7640 surface_load_fb_texture(src_surface
, FALSE
);
7642 src_rect
.top
= src_surface
->resource
.height
- src_rect
.top
;
7643 src_rect
.bottom
= src_surface
->resource
.height
- src_rect
.bottom
;
7646 wined3d_texture_load(src_surface
->container
, context
, FALSE
);
7648 context_apply_blit_state(context
, device
);
7650 if (!surface_is_offscreen(dst_surface
))
7651 surface_translate_drawable_coords(dst_surface
, context
->win_handle
, &dst_rect
);
7653 arbfp_blit_set(device
->blit_priv
, context
, src_surface
);
7655 /* Draw a textured quad */
7656 draw_textured_quad(src_surface
, context
, &src_rect
, &dst_rect
, filter
);
7658 /* Leave the opengl state valid for blitting */
7659 arbfp_blit_unset(context
->gl_info
);
7661 if (wined3d_settings
.strict_draw_ordering
7662 || (dst_surface
->swapchain
&& (dst_surface
->swapchain
->front_buffer
== dst_surface
)))
7663 context
->gl_info
->gl_ops
.gl
.p_glFlush(); /* Flush to ensure ordering across contexts. */
7665 context_release(context
);
7667 surface_validate_location(dst_surface
, dst_surface
->draw_binding
);
7668 surface_invalidate_location(dst_surface
, ~dst_surface
->draw_binding
);
7673 static HRESULT
arbfp_blit_color_fill(struct wined3d_device
*device
, struct wined3d_surface
*dst_surface
,
7674 const RECT
*dst_rect
, const struct wined3d_color
*color
)
7676 FIXME("Color filling not implemented by arbfp_blit\n");
7677 return WINED3DERR_INVALIDCALL
;
7680 static HRESULT
arbfp_blit_depth_fill(struct wined3d_device
*device
,
7681 struct wined3d_surface
*surface
, const RECT
*rect
, float depth
)
7683 FIXME("Depth filling not implemented by arbfp_blit.\n");
7684 return WINED3DERR_INVALIDCALL
;
7687 const struct blit_shader arbfp_blit
= {
7692 arbfp_blit_supported
,
7693 arbfp_blit_color_fill
,
7694 arbfp_blit_depth_fill
,