2 * Pixel and vertex shaders implementation using ARB_vertex_program
3 * and ARB_fragment_program GL extensions.
5 * Copyright 2002-2003 Jason Edmeades
6 * Copyright 2002-2003 Raphael Junqueira
7 * Copyright 2004 Christian Costa
8 * Copyright 2005 Oliver Stieber
9 * Copyright 2006 Ivan Gyurdiev
10 * Copyright 2006 Jason Green
11 * Copyright 2006 Henri Verbeet
12 * Copyright 2007-2011, 2013-2014 Stefan Dösinger for CodeWeavers
13 * Copyright 2009 Henri Verbeet for CodeWeavers
15 * This library is free software; you can redistribute it and/or
16 * modify it under the terms of the GNU Lesser General Public
17 * License as published by the Free Software Foundation; either
18 * version 2.1 of the License, or (at your option) any later version.
20 * This library is distributed in the hope that it will be useful,
21 * but WITHOUT ANY WARRANTY; without even the implied warranty of
22 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
23 * Lesser General Public License for more details.
25 * You should have received a copy of the GNU Lesser General Public
26 * License along with this library; if not, write to the Free Software
27 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
31 #include "wine/port.h"
36 #include "wined3d_private.h"
38 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader
);
39 WINE_DECLARE_DEBUG_CHANNEL(d3d_constants
);
40 WINE_DECLARE_DEBUG_CHANNEL(d3d
);
41 WINE_DECLARE_DEBUG_CHANNEL(d3d_perf
);
43 static BOOL
shader_is_pshader_version(enum wined3d_shader_type type
)
45 return type
== WINED3D_SHADER_TYPE_PIXEL
;
48 static BOOL
shader_is_vshader_version(enum wined3d_shader_type type
)
50 return type
== WINED3D_SHADER_TYPE_VERTEX
;
53 static const char *get_line(const char **ptr
)
58 if (!(q
= strstr(p
, "\n")))
80 static const char *arb_get_helper_value(enum wined3d_shader_type shader
, enum arb_helper_value value
)
82 if (shader
!= WINED3D_SHADER_TYPE_VERTEX
&& shader
!= WINED3D_SHADER_TYPE_PIXEL
)
84 ERR("Unsupported shader type '%s'.\n.", debug_shader_type(shader
));
88 if (shader
== WINED3D_SHADER_TYPE_PIXEL
)
92 case ARB_ZERO
: return "ps_helper_const.x";
93 case ARB_ONE
: return "ps_helper_const.y";
94 case ARB_TWO
: return "coefmul.x";
95 case ARB_0001
: return "ps_helper_const.xxxy";
96 case ARB_EPS
: return "ps_helper_const.z";
104 case ARB_ZERO
: return "helper_const.x";
105 case ARB_ONE
: return "helper_const.y";
106 case ARB_TWO
: return "helper_const.z";
107 case ARB_EPS
: return "helper_const.w";
108 case ARB_0001
: return "helper_const.xxxy";
109 case ARB_VS_REL_OFFSET
: return "rel_addr_const.y";
112 FIXME("Unmanaged %s shader helper constant requested: %u.\n",
113 shader
== WINED3D_SHADER_TYPE_PIXEL
? "pixel" : "vertex", value
);
116 case ARB_ZERO
: return "0.0";
117 case ARB_ONE
: return "1.0";
118 case ARB_TWO
: return "2.0";
119 case ARB_0001
: return "{0.0, 0.0, 0.0, 1.0}";
120 case ARB_EPS
: return "1e-8";
121 default: return "bad";
125 static inline BOOL
ffp_clip_emul(const struct wined3d_context
*context
)
127 return context
->lowest_disabled_stage
< 7;
130 /* ARB_program_shader private data */
149 struct wined3d_shader_loop_control loop_control
;
153 struct arb_ps_np2fixup_info
155 struct ps_np2fixup_info super
;
156 /* For ARB we need an offset value:
157 * With both GLSL and ARB mode the NP2 fixup information (the texture dimensions) are stored in a
158 * consecutive way (GLSL uses a uniform array). Since ARB doesn't know the notion of a "standalone"
159 * array we need an offset to the index inside the program local parameter array. */
163 struct arb_ps_compile_args
165 struct ps_compile_args super
;
167 WORD clip
; /* only a boolean, use a WORD for alignment */
168 unsigned char loop_ctrl
[MAX_CONST_I
][3];
171 struct stb_const_desc
173 unsigned char texunit
;
177 struct arb_ps_compiled_shader
179 struct arb_ps_compile_args args
;
180 struct arb_ps_np2fixup_info np2fixup_info
;
181 struct stb_const_desc bumpenvmatconst
[MAX_TEXTURES
];
182 struct stb_const_desc luminanceconst
[MAX_TEXTURES
];
183 UINT int_consts
[MAX_CONST_I
];
186 unsigned char numbumpenvmatconsts
;
190 struct arb_vs_compile_args
192 struct vs_compile_args super
;
198 unsigned char clip_texcoord
;
199 unsigned char clipplane_mask
;
201 DWORD boolclip_compare
;
206 unsigned char samplers
[4];
207 DWORD samplers_compare
;
209 unsigned char loop_ctrl
[MAX_CONST_I
][3];
212 struct arb_vs_compiled_shader
214 struct arb_vs_compile_args args
;
216 UINT int_consts
[MAX_CONST_I
];
218 char need_color_unclamp
;
222 struct recorded_instruction
224 struct wined3d_shader_instruction ins
;
228 struct shader_arb_ctx_priv
233 /* plain GL_ARB_vertex_program or GL_ARB_fragment_program */
235 /* GL_NV_vertex_program2_option or GL_NV_fragment_program_option */
237 /* GL_NV_vertex_program3 or GL_NV_fragment_program2 */
241 const struct arb_vs_compile_args
*cur_vs_args
;
242 const struct arb_ps_compile_args
*cur_ps_args
;
243 const struct arb_ps_compiled_shader
*compiled_fprog
;
244 const struct arb_vs_compiled_shader
*compiled_vprog
;
245 struct arb_ps_np2fixup_info
*cur_np2fixup_info
;
246 struct list control_frames
;
250 unsigned int num_loops
, loop_depth
, num_ifcs
;
252 BOOL ps_post_process
;
254 unsigned int vs_clipplanes
;
258 /* For 3.0 vertex shaders */
259 const char *vs_output
[MAX_REG_OUTPUT
];
260 /* For 2.x and earlier vertex shaders */
261 const char *texcrd_output
[8], *color_output
[2], *fog_output
;
263 /* 3.0 pshader input for compatibility with fixed function */
264 const char *ps_input
[MAX_REG_INPUT
];
269 struct wined3d_shader_signature sig
;
271 struct wine_rb_entry entry
;
274 struct arb_pshader_private
{
275 struct arb_ps_compiled_shader
*gl_shaders
;
276 UINT num_gl_shaders
, shader_array_size
;
277 DWORD input_signature_idx
;
278 DWORD clipplane_emulation
;
282 struct arb_vshader_private
{
283 struct arb_vs_compiled_shader
*gl_shaders
;
284 UINT num_gl_shaders
, shader_array_size
;
288 struct shader_arb_priv
290 GLuint current_vprogram_id
;
291 GLuint current_fprogram_id
;
292 const struct arb_ps_compiled_shader
*compiled_fprog
;
293 const struct arb_vs_compiled_shader
*compiled_vprog
;
294 GLuint depth_blt_vprogram_id
;
295 GLuint depth_blt_fprogram_id_full
[WINED3D_GL_RES_TYPE_COUNT
];
296 GLuint depth_blt_fprogram_id_masked
[WINED3D_GL_RES_TYPE_COUNT
];
297 BOOL use_arbfp_fixed_func
;
298 struct wine_rb_tree fragment_shaders
;
299 BOOL last_ps_const_clamped
;
300 BOOL last_vs_color_unclamp
;
302 struct wine_rb_tree signature_tree
;
305 unsigned int highest_dirty_ps_const
, highest_dirty_vs_const
;
306 char *vshader_const_dirty
, *pshader_const_dirty
;
307 const struct wined3d_context
*last_context
;
309 const struct wined3d_vertex_pipe_ops
*vertex_pipe
;
310 const struct fragment_pipeline
*fragment_pipe
;
311 BOOL ffp_proj_control
;
314 /* Context activation for state handlers is done by the caller. */
316 static BOOL
need_rel_addr_const(const struct arb_vshader_private
*shader_data
,
317 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
)
319 if (shader_data
->rel_offset
) return TRUE
;
320 if (!reg_maps
->usesmova
) return FALSE
;
321 return !gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
];
324 /* Returns TRUE if result.clip from GL_NV_vertex_program2 should be used and FALSE otherwise */
325 static inline BOOL
use_nv_clip(const struct wined3d_gl_info
*gl_info
)
327 return gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]
328 && !(gl_info
->quirks
& WINED3D_QUIRK_NV_CLIP_BROKEN
);
331 static BOOL
need_helper_const(const struct arb_vshader_private
*shader_data
,
332 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
)
334 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
)) return TRUE
;
335 if (!gl_info
->supported
[NV_VERTEX_PROGRAM
]) return TRUE
; /* Need to init colors. */
336 if (gl_info
->quirks
& WINED3D_QUIRK_ARB_VS_OFFSET_LIMIT
) return TRUE
; /* Load the immval offset. */
337 if (gl_info
->quirks
& WINED3D_QUIRK_SET_TEXCOORD_W
) return TRUE
; /* Have to init texcoords. */
338 if (!use_nv_clip(gl_info
)) return TRUE
; /* Init the clip texcoord */
339 if (reg_maps
->usesnrm
) return TRUE
; /* 0.0 */
340 if (reg_maps
->usespow
) return TRUE
; /* EPS, 0.0 and 1.0 */
341 if (reg_maps
->fog
) return TRUE
; /* Clamping fog coord, 0.0 and 1.0 */
345 static unsigned int reserved_vs_const(const struct arb_vshader_private
*shader_data
,
346 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
)
348 unsigned int ret
= 1;
349 /* We use one PARAM for the pos fixup, and in some cases one to load
350 * some immediate values into the shader. */
351 if (need_helper_const(shader_data
, reg_maps
, gl_info
)) ++ret
;
352 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
)) ++ret
;
356 /* Loads floating point constants into the currently set ARB_vertex/fragment_program.
357 * When constant_list == NULL, it will load all the constants.
359 * @target_type should be either GL_VERTEX_PROGRAM_ARB (for vertex shaders)
360 * or GL_FRAGMENT_PROGRAM_ARB (for pixel shaders)
362 /* Context activation is done by the caller. */
363 static unsigned int shader_arb_load_constantsF(const struct wined3d_shader
*shader
,
364 const struct wined3d_gl_info
*gl_info
, GLuint target_type
, unsigned int max_constants
,
365 const float *constants
, char *dirty_consts
)
367 struct wined3d_shader_lconst
*lconst
;
371 if (TRACE_ON(d3d_constants
))
373 for(i
= 0; i
< max_constants
; i
++) {
374 if(!dirty_consts
[i
]) continue;
375 TRACE_(d3d_constants
)("Loading constants %i: %f, %f, %f, %f\n", i
,
376 constants
[i
* 4 + 0], constants
[i
* 4 + 1],
377 constants
[i
* 4 + 2], constants
[i
* 4 + 3]);
383 /* In 1.X pixel shaders constants are implicitly clamped in the range [-1;1] */
384 if (target_type
== GL_FRAGMENT_PROGRAM_ARB
&& shader
->reg_maps
.shader_version
.major
== 1)
387 /* ps 1.x supports only 8 constants, clamp only those. When switching between 1.x and higher
388 * shaders, the first 8 constants are marked dirty for reload
390 for(; i
< min(8, max_constants
); i
++) {
391 if(!dirty_consts
[i
]) continue;
395 if (constants
[j
+ 0] > 1.0f
) lcl_const
[0] = 1.0f
;
396 else if (constants
[j
+ 0] < -1.0f
) lcl_const
[0] = -1.0f
;
397 else lcl_const
[0] = constants
[j
+ 0];
399 if (constants
[j
+ 1] > 1.0f
) lcl_const
[1] = 1.0f
;
400 else if (constants
[j
+ 1] < -1.0f
) lcl_const
[1] = -1.0f
;
401 else lcl_const
[1] = constants
[j
+ 1];
403 if (constants
[j
+ 2] > 1.0f
) lcl_const
[2] = 1.0f
;
404 else if (constants
[j
+ 2] < -1.0f
) lcl_const
[2] = -1.0f
;
405 else lcl_const
[2] = constants
[j
+ 2];
407 if (constants
[j
+ 3] > 1.0f
) lcl_const
[3] = 1.0f
;
408 else if (constants
[j
+ 3] < -1.0f
) lcl_const
[3] = -1.0f
;
409 else lcl_const
[3] = constants
[j
+ 3];
411 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, i
, lcl_const
));
414 /* If further constants are dirty, reload them without clamping.
416 * The alternative is not to touch them, but then we cannot reset the dirty constant count
417 * to zero. That's bad for apps that only use PS 1.x shaders, because in that case the code
418 * above would always re-check the first 8 constants since max_constant remains at the init
423 if (gl_info
->supported
[EXT_GPU_PROGRAM_PARAMETERS
])
425 /* TODO: Benchmark if we're better of with finding the dirty constants ourselves,
426 * or just reloading *all* constants at once
428 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type, i, max_constants, constants + (i * 4)));
430 for(; i
< max_constants
; i
++) {
431 if(!dirty_consts
[i
]) continue;
433 /* Find the next block of dirty constants */
436 for(i
++; (i
< max_constants
) && dirty_consts
[i
]; i
++) {
440 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type
, j
, i
- j
, constants
+ (j
* 4)));
443 for(; i
< max_constants
; i
++) {
444 if(dirty_consts
[i
]) {
446 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, i
, constants
+ (i
* 4)));
450 checkGLcall("glProgramEnvParameter4fvARB()");
452 /* Load immediate constants */
453 if (shader
->load_local_constsF
)
455 if (TRACE_ON(d3d_shader
))
457 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
459 GLfloat
* values
= (GLfloat
*)lconst
->value
;
460 TRACE_(d3d_constants
)("Loading local constants %i: %f, %f, %f, %f\n", lconst
->idx
,
461 values
[0], values
[1], values
[2], values
[3]);
464 /* Immediate constants are clamped for 1.X shaders at loading times */
466 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
468 dirty_consts
[lconst
->idx
] = 1; /* Dirtify so the non-immediate constant overwrites it next time */
469 ret
= max(ret
, lconst
->idx
+ 1);
470 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, lconst
->idx
, (GLfloat
*)lconst
->value
));
472 checkGLcall("glProgramEnvParameter4fvARB()");
473 return ret
; /* The loaded immediate constants need reloading for the next shader */
475 return 0; /* No constants are dirty now */
479 /* Loads the texture dimensions for NP2 fixup into the currently set
480 * ARB_[vertex/fragment]_programs. */
481 static void shader_arb_load_np2fixup_constants(const struct arb_ps_np2fixup_info
*fixup
,
482 const struct wined3d_gl_info
*gl_info
, const struct wined3d_state
*state
)
484 GLfloat np2fixup_constants
[4 * MAX_FRAGMENT_SAMPLERS
];
485 WORD active
= fixup
->super
.active
;
491 for (i
= 0; active
; active
>>= 1, ++i
)
493 const struct wined3d_texture
*tex
= state
->textures
[i
];
494 unsigned char idx
= fixup
->super
.idx
[i
];
495 GLfloat
*tex_dim
= &np2fixup_constants
[(idx
>> 1) * 4];
502 ERR("Nonexistent texture is flagged for NP2 texcoord fixup.\n");
508 tex_dim
[2] = tex
->pow2_matrix
[0];
509 tex_dim
[3] = tex
->pow2_matrix
[5];
513 tex_dim
[0] = tex
->pow2_matrix
[0];
514 tex_dim
[1] = tex
->pow2_matrix
[5];
518 for (i
= 0; i
< fixup
->super
.num_consts
; ++i
)
520 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
521 fixup
->offset
+ i
, &np2fixup_constants
[i
* 4]));
525 /* Context activation is done by the caller. */
526 static void shader_arb_ps_local_constants(const struct arb_ps_compiled_shader
*gl_shader
,
527 const struct wined3d_context
*context
, const struct wined3d_state
*state
, UINT rt_height
)
529 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
532 for(i
= 0; i
< gl_shader
->numbumpenvmatconsts
; i
++)
534 int texunit
= gl_shader
->bumpenvmatconst
[i
].texunit
;
536 /* The state manager takes care that this function is always called if the bump env matrix changes */
537 const float *data
= (const float *)&state
->texture_states
[texunit
][WINED3D_TSS_BUMPENV_MAT00
];
538 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
539 gl_shader
->bumpenvmatconst
[i
].const_num
, data
));
541 if (gl_shader
->luminanceconst
[i
].const_num
!= WINED3D_CONST_NUM_UNUSED
)
543 /* WINED3D_TSS_BUMPENVLSCALE and WINED3D_TSS_BUMPENVLOFFSET are next to each other.
544 * point gl to the scale, and load 4 floats. x = scale, y = offset, z and w are junk, we
545 * don't care about them. The pointers are valid for sure because the stateblock is bigger.
546 * (they're WINED3D_TSS_TEXTURETRANSFORMFLAGS and WINED3D_TSS_ADDRESSW, so most likely 0 or NaN
548 const float *scale
= (const float *)&state
->texture_states
[texunit
][WINED3D_TSS_BUMPENV_LSCALE
];
549 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
550 gl_shader
->luminanceconst
[i
].const_num
, scale
));
553 checkGLcall("Load bumpmap consts");
555 if(gl_shader
->ycorrection
!= WINED3D_CONST_NUM_UNUSED
)
557 /* ycorrection.x: Backbuffer height(onscreen) or 0(offscreen).
558 * ycorrection.y: -1.0(onscreen), 1.0(offscreen)
563 val
[0] = context
->render_offscreen
? 0.0f
: (float) rt_height
;
564 val
[1] = context
->render_offscreen
? 1.0f
: -1.0f
;
567 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, gl_shader
->ycorrection
, val
));
568 checkGLcall("y correction loading");
571 if (!gl_shader
->num_int_consts
) return;
573 for(i
= 0; i
< MAX_CONST_I
; i
++)
575 if(gl_shader
->int_consts
[i
] != WINED3D_CONST_NUM_UNUSED
)
578 val
[0] = (float)state
->ps_consts_i
[4 * i
];
579 val
[1] = (float)state
->ps_consts_i
[4 * i
+ 1];
580 val
[2] = (float)state
->ps_consts_i
[4 * i
+ 2];
583 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, gl_shader
->int_consts
[i
], val
));
586 checkGLcall("Load ps int consts");
589 /* Context activation is done by the caller. */
590 static void shader_arb_vs_local_constants(const struct arb_vs_compiled_shader
*gl_shader
,
591 const struct wined3d_context
*context
, const struct wined3d_state
*state
)
593 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
594 float position_fixup
[4];
597 /* Upload the position fixup */
598 shader_get_position_fixup(context
, state
, position_fixup
);
599 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, gl_shader
->pos_fixup
, position_fixup
));
601 if (!gl_shader
->num_int_consts
) return;
603 for(i
= 0; i
< MAX_CONST_I
; i
++)
605 if(gl_shader
->int_consts
[i
] != WINED3D_CONST_NUM_UNUSED
)
608 val
[0] = (float)state
->vs_consts_i
[4 * i
];
609 val
[1] = (float)state
->vs_consts_i
[4 * i
+ 1];
610 val
[2] = (float)state
->vs_consts_i
[4 * i
+ 2];
613 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, gl_shader
->int_consts
[i
], val
));
616 checkGLcall("Load vs int consts");
619 static void shader_arb_select(void *shader_priv
, struct wined3d_context
*context
,
620 const struct wined3d_state
*state
);
623 * Loads the app-supplied constants into the currently set ARB_[vertex/fragment]_programs.
625 * We only support float constants in ARB at the moment, so don't
626 * worry about the Integers or Booleans
628 /* Context activation is done by the caller (state handler). */
629 static void shader_arb_load_constants_internal(struct shader_arb_priv
*priv
,
630 struct wined3d_context
*context
, const struct wined3d_state
*state
,
631 BOOL usePixelShader
, BOOL useVertexShader
, BOOL from_shader_select
)
633 const struct wined3d_d3d_info
*d3d_info
= context
->d3d_info
;
634 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
636 if (!from_shader_select
)
638 const struct wined3d_shader
*vshader
= state
->shader
[WINED3D_SHADER_TYPE_VERTEX
];
639 const struct wined3d_shader
*pshader
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
642 && (vshader
->reg_maps
.boolean_constants
643 || (!gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]
644 && (vshader
->reg_maps
.integer_constants
& ~vshader
->reg_maps
.local_int_consts
))))
646 TRACE("bool/integer vertex shader constants potentially modified, forcing shader reselection.\n");
647 shader_arb_select(priv
, context
, state
);
650 && (pshader
->reg_maps
.boolean_constants
651 || (!gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
]
652 && (pshader
->reg_maps
.integer_constants
& ~pshader
->reg_maps
.local_int_consts
))))
654 TRACE("bool/integer pixel shader constants potentially modified, forcing shader reselection.\n");
655 shader_arb_select(priv
, context
, state
);
659 if (context
!= priv
->last_context
)
661 memset(priv
->vshader_const_dirty
, 1,
662 sizeof(*priv
->vshader_const_dirty
) * d3d_info
->limits
.vs_uniform_count
);
663 priv
->highest_dirty_vs_const
= d3d_info
->limits
.vs_uniform_count
;
665 memset(priv
->pshader_const_dirty
, 1,
666 sizeof(*priv
->pshader_const_dirty
) * d3d_info
->limits
.ps_uniform_count
);
667 priv
->highest_dirty_ps_const
= d3d_info
->limits
.ps_uniform_count
;
669 priv
->last_context
= context
;
674 const struct wined3d_shader
*vshader
= state
->shader
[WINED3D_SHADER_TYPE_VERTEX
];
675 const struct arb_vs_compiled_shader
*gl_shader
= priv
->compiled_vprog
;
677 /* Load DirectX 9 float constants for vertex shader */
678 priv
->highest_dirty_vs_const
= shader_arb_load_constantsF(vshader
, gl_info
, GL_VERTEX_PROGRAM_ARB
,
679 priv
->highest_dirty_vs_const
, state
->vs_consts_f
, priv
->vshader_const_dirty
);
680 shader_arb_vs_local_constants(gl_shader
, context
, state
);
685 const struct wined3d_shader
*pshader
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
686 const struct arb_ps_compiled_shader
*gl_shader
= priv
->compiled_fprog
;
687 UINT rt_height
= state
->fb
->render_targets
[0]->height
;
689 /* Load DirectX 9 float constants for pixel shader */
690 priv
->highest_dirty_ps_const
= shader_arb_load_constantsF(pshader
, gl_info
, GL_FRAGMENT_PROGRAM_ARB
,
691 priv
->highest_dirty_ps_const
, state
->ps_consts_f
, priv
->pshader_const_dirty
);
692 shader_arb_ps_local_constants(gl_shader
, context
, state
, rt_height
);
694 if (context
->constant_update_mask
& WINED3D_SHADER_CONST_PS_NP2_FIXUP
)
695 shader_arb_load_np2fixup_constants(&gl_shader
->np2fixup_info
, gl_info
, state
);
699 static void shader_arb_load_constants(void *shader_priv
, struct wined3d_context
*context
,
700 const struct wined3d_state
*state
)
702 BOOL vs
= use_vs(state
);
703 BOOL ps
= use_ps(state
);
705 shader_arb_load_constants_internal(shader_priv
, context
, state
, ps
, vs
, FALSE
);
708 static void shader_arb_update_float_vertex_constants(struct wined3d_device
*device
, UINT start
, UINT count
)
710 struct wined3d_context
*context
= context_get_current();
711 struct shader_arb_priv
*priv
= device
->shader_priv
;
714 for (i
= 0; i
< device
->context_count
; ++i
)
716 device
->contexts
[i
]->constant_update_mask
|= WINED3D_SHADER_CONST_VS_F
;
719 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
720 * context. On a context switch the old context will be fully dirtified */
721 if (!context
|| context
->swapchain
->device
!= device
) return;
723 memset(priv
->vshader_const_dirty
+ start
, 1, sizeof(*priv
->vshader_const_dirty
) * count
);
724 priv
->highest_dirty_vs_const
= max(priv
->highest_dirty_vs_const
, start
+ count
);
727 static void shader_arb_update_float_pixel_constants(struct wined3d_device
*device
, UINT start
, UINT count
)
729 struct wined3d_context
*context
= context_get_current();
730 struct shader_arb_priv
*priv
= device
->shader_priv
;
733 for (i
= 0; i
< device
->context_count
; ++i
)
735 device
->contexts
[i
]->constant_update_mask
|= WINED3D_SHADER_CONST_PS_F
;
738 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
739 * context. On a context switch the old context will be fully dirtified */
740 if (!context
|| context
->swapchain
->device
!= device
) return;
742 memset(priv
->pshader_const_dirty
+ start
, 1, sizeof(*priv
->pshader_const_dirty
) * count
);
743 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, start
+ count
);
746 static void shader_arb_append_imm_vec4(struct wined3d_string_buffer
*buffer
, const float *values
)
750 wined3d_ftoa(values
[0], str
[0]);
751 wined3d_ftoa(values
[1], str
[1]);
752 wined3d_ftoa(values
[2], str
[2]);
753 wined3d_ftoa(values
[3], str
[3]);
754 shader_addline(buffer
, "{%s, %s, %s, %s}", str
[0], str
[1], str
[2], str
[3]);
757 /* Generate the variable & register declarations for the ARB_vertex_program output target */
758 static void shader_generate_arb_declarations(const struct wined3d_shader
*shader
,
759 const struct wined3d_shader_reg_maps
*reg_maps
, struct wined3d_string_buffer
*buffer
,
760 const struct wined3d_gl_info
*gl_info
, DWORD
*num_clipplanes
,
761 const struct shader_arb_ctx_priv
*ctx
)
764 char pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
765 const struct wined3d_shader_lconst
*lconst
;
766 unsigned max_constantsF
;
769 /* In pixel shaders, all private constants are program local, we don't need anything
770 * from program.env. Thus we can advertise the full set of constants in pixel shaders.
771 * If we need a private constant the GL implementation will squeeze it in somewhere
773 * With vertex shaders we need the posFixup and on some GL implementations 4 helper
774 * immediate values. The posFixup is loaded using program.env for now, so always
775 * subtract one from the number of constants. If the shader uses indirect addressing,
776 * account for the helper const too because we have to declare all available d3d constants
777 * and don't know which are actually used.
781 max_constantsF
= gl_info
->limits
.arb_ps_native_constants
;
782 /* 24 is the minimum MAX_PROGRAM_ENV_PARAMETERS_ARB value. */
783 if (max_constantsF
< 24)
784 max_constantsF
= gl_info
->limits
.arb_ps_float_constants
;
788 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
789 max_constantsF
= gl_info
->limits
.arb_vs_native_constants
;
790 /* 96 is the minimum MAX_PROGRAM_ENV_PARAMETERS_ARB value.
791 * Also prevents max_constantsF from becoming less than 0 and
793 if (max_constantsF
< 96)
794 max_constantsF
= gl_info
->limits
.arb_vs_float_constants
;
796 if (reg_maps
->usesrelconstF
)
798 DWORD highest_constf
= 0, clip_limit
;
800 max_constantsF
-= reserved_vs_const(shader_data
, reg_maps
, gl_info
);
801 max_constantsF
-= wined3d_popcount(reg_maps
->integer_constants
);
802 max_constantsF
-= gl_info
->reserved_arb_constants
;
804 for (i
= 0; i
< shader
->limits
->constant_float
; ++i
)
807 DWORD shift
= i
& 0x1f;
808 if (reg_maps
->constf
[idx
] & (1u << shift
))
812 if(use_nv_clip(gl_info
) && ctx
->target_version
>= NV2
)
814 if(ctx
->cur_vs_args
->super
.clip_enabled
)
815 clip_limit
= gl_info
->limits
.clipplanes
;
821 unsigned int mask
= ctx
->cur_vs_args
->clip
.boolclip
.clipplane_mask
;
822 clip_limit
= min(wined3d_popcount(mask
), 4);
824 *num_clipplanes
= min(clip_limit
, max_constantsF
- highest_constf
- 1);
825 max_constantsF
-= *num_clipplanes
;
826 if(*num_clipplanes
< clip_limit
)
828 WARN("Only %u clipplanes out of %u enabled\n", *num_clipplanes
, gl_info
->limits
.clipplanes
);
833 if (ctx
->target_version
>= NV2
) *num_clipplanes
= gl_info
->limits
.clipplanes
;
834 else *num_clipplanes
= min(gl_info
->limits
.clipplanes
, 4);
838 for (i
= 0, map
= reg_maps
->temporary
; map
; map
>>= 1, ++i
)
840 if (map
& 1) shader_addline(buffer
, "TEMP R%u;\n", i
);
843 for (i
= 0, map
= reg_maps
->address
; map
; map
>>= 1, ++i
)
845 if (map
& 1) shader_addline(buffer
, "ADDRESS A%u;\n", i
);
848 if (pshader
&& reg_maps
->shader_version
.major
== 1 && reg_maps
->shader_version
.minor
<= 3)
850 for (i
= 0, map
= reg_maps
->texcoord
; map
; map
>>= 1, ++i
)
852 if (map
& 1) shader_addline(buffer
, "TEMP T%u;\n", i
);
856 if (!shader
->load_local_constsF
)
858 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
861 value
= (const float *)lconst
->value
;
862 shader_addline(buffer
, "PARAM C%u = ", lconst
->idx
);
863 shader_arb_append_imm_vec4(buffer
, value
);
864 shader_addline(buffer
, ";\n");
868 /* After subtracting privately used constants from the hardware limit(they are loaded as
869 * local constants), make sure the shader doesn't violate the env constant limit
873 max_constantsF
= min(max_constantsF
, gl_info
->limits
.arb_ps_float_constants
);
877 max_constantsF
= min(max_constantsF
, gl_info
->limits
.arb_vs_float_constants
);
880 /* Avoid declaring more constants than needed */
881 max_constantsF
= min(max_constantsF
, shader
->limits
->constant_float
);
883 /* we use the array-based constants array if the local constants are marked for loading,
884 * because then we use indirect addressing, or when the local constant list is empty,
885 * because then we don't know if we're using indirect addressing or not. If we're hardcoding
886 * local constants do not declare the loaded constants as an array because ARB compilers usually
887 * do not optimize unused constants away
889 if (reg_maps
->usesrelconstF
)
891 /* Need to PARAM the environment parameters (constants) so we can use relative addressing */
892 shader_addline(buffer
, "PARAM C[%d] = { program.env[0..%d] };\n",
893 max_constantsF
, max_constantsF
- 1);
897 for (i
= 0; i
< max_constantsF
; ++i
)
901 mask
= 1u << (i
& 0x1fu
);
902 if (!shader_constant_is_local(shader
, i
) && (reg_maps
->constf
[idx
] & mask
))
904 shader_addline(buffer
, "PARAM C%d = program.env[%d];\n",i
, i
);
910 static const char * const shift_tab
[] = {
911 "dummy", /* 0 (none) */
912 "coefmul.x", /* 1 (x2) */
913 "coefmul.y", /* 2 (x4) */
914 "coefmul.z", /* 3 (x8) */
915 "coefmul.w", /* 4 (x16) */
916 "dummy", /* 5 (x32) */
917 "dummy", /* 6 (x64) */
918 "dummy", /* 7 (x128) */
919 "dummy", /* 8 (d256) */
920 "dummy", /* 9 (d128) */
921 "dummy", /* 10 (d64) */
922 "dummy", /* 11 (d32) */
923 "coefdiv.w", /* 12 (d16) */
924 "coefdiv.z", /* 13 (d8) */
925 "coefdiv.y", /* 14 (d4) */
926 "coefdiv.x" /* 15 (d2) */
929 static void shader_arb_get_write_mask(const struct wined3d_shader_instruction
*ins
,
930 const struct wined3d_shader_dst_param
*dst
, char *write_mask
)
932 char *ptr
= write_mask
;
934 if (dst
->write_mask
!= WINED3DSP_WRITEMASK_ALL
)
937 if (dst
->write_mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
938 if (dst
->write_mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
939 if (dst
->write_mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
940 if (dst
->write_mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
946 static void shader_arb_get_swizzle(const struct wined3d_shader_src_param
*param
, BOOL fixup
, char *swizzle_str
)
948 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
949 * but addressed as "rgba". To fix this we need to swap the register's x
950 * and z components. */
951 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
952 char *ptr
= swizzle_str
;
954 /* swizzle bits fields: wwzzyyxx */
955 DWORD swizzle
= param
->swizzle
;
956 DWORD swizzle_x
= swizzle
& 0x03;
957 DWORD swizzle_y
= (swizzle
>> 2) & 0x03;
958 DWORD swizzle_z
= (swizzle
>> 4) & 0x03;
959 DWORD swizzle_w
= (swizzle
>> 6) & 0x03;
961 /* If the swizzle is the default swizzle (ie, "xyzw"), we don't need to
962 * generate a swizzle string. Unless we need to our own swizzling. */
963 if (swizzle
!= WINED3DSP_NOSWIZZLE
|| fixup
)
966 if (swizzle_x
== swizzle_y
&& swizzle_x
== swizzle_z
&& swizzle_x
== swizzle_w
) {
967 *ptr
++ = swizzle_chars
[swizzle_x
];
969 *ptr
++ = swizzle_chars
[swizzle_x
];
970 *ptr
++ = swizzle_chars
[swizzle_y
];
971 *ptr
++ = swizzle_chars
[swizzle_z
];
972 *ptr
++ = swizzle_chars
[swizzle_w
];
979 static void shader_arb_request_a0(const struct wined3d_shader_instruction
*ins
, const char *src
)
981 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
982 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
984 if (!strcmp(priv
->addr_reg
, src
)) return;
986 strcpy(priv
->addr_reg
, src
);
987 shader_addline(buffer
, "ARL A0.x, %s;\n", src
);
990 static void shader_arb_get_src_param(const struct wined3d_shader_instruction
*ins
,
991 const struct wined3d_shader_src_param
*src
, unsigned int tmpreg
, char *outregstr
);
993 static void shader_arb_get_register_name(const struct wined3d_shader_instruction
*ins
,
994 const struct wined3d_shader_register
*reg
, char *register_name
, BOOL
*is_color
)
996 /* oPos, oFog and oPts in D3D */
997 static const char * const rastout_reg_names
[] = {"TMP_OUT", "TMP_FOGCOORD", "result.pointsize"};
998 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
999 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
1000 BOOL pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
1001 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1007 case WINED3DSPR_TEMP
:
1008 sprintf(register_name
, "R%u", reg
->idx
[0].offset
);
1011 case WINED3DSPR_INPUT
:
1014 if (reg_maps
->shader_version
.major
< 3)
1016 if (!reg
->idx
[0].offset
)
1017 strcpy(register_name
, "fragment.color.primary");
1019 strcpy(register_name
, "fragment.color.secondary");
1023 if (reg
->idx
[0].rel_addr
)
1026 shader_arb_get_src_param(ins
, reg
->idx
[0].rel_addr
, 0, rel_reg
);
1028 if (!strcmp(rel_reg
, "**aL_emul**"))
1030 DWORD idx
= ctx
->aL
+ reg
->idx
[0].offset
;
1031 if(idx
< MAX_REG_INPUT
)
1033 strcpy(register_name
, ctx
->ps_input
[idx
]);
1037 ERR("Pixel shader input register out of bounds: %u\n", idx
);
1038 sprintf(register_name
, "out_of_bounds_%u", idx
);
1041 else if (reg_maps
->input_registers
& 0x0300)
1043 /* There are two ways basically:
1045 * 1) Use the unrolling code that is used for loop emulation and unroll the loop.
1046 * That means trouble if the loop also contains a breakc or if the control values
1047 * aren't local constants.
1048 * 2) Generate an if block that checks if aL.y < 8, == 8 or == 9 and selects the
1049 * source dynamically. The trouble is that we cannot simply read aL.y because it
1050 * is an ADDRESS register. We could however push it, load .zw with a value and use
1051 * ADAC to load the condition code register and pop it again afterwards
1053 FIXME("Relative input register addressing with more than 8 registers\n");
1055 /* This is better than nothing for now */
1056 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1058 else if(ctx
->cur_ps_args
->super
.vp_mode
!= vertexshader
)
1060 /* This is problematic because we'd have to consult the ctx->ps_input strings
1061 * for where to find the varying. Some may be "0.0", others can be texcoords or
1062 * colors. This needs either a pipeline replacement to make the vertex shader feed
1063 * proper varyings, or loop unrolling
1065 * For now use the texcoords and hope for the best
1067 FIXME("Non-vertex shader varying input with indirect addressing\n");
1068 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1072 /* D3D supports indirect addressing only with aL in loop registers. The loop instruction
1073 * pulls GL_NV_fragment_program2 in
1075 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1080 if (reg
->idx
[0].offset
< MAX_REG_INPUT
)
1082 strcpy(register_name
, ctx
->ps_input
[reg
->idx
[0].offset
]);
1086 ERR("Pixel shader input register out of bounds: %u\n", reg
->idx
[0].offset
);
1087 sprintf(register_name
, "out_of_bounds_%u", reg
->idx
[0].offset
);
1094 if (ctx
->cur_vs_args
->super
.swizzle_map
& (1u << reg
->idx
[0].offset
))
1096 sprintf(register_name
, "vertex.attrib[%u]", reg
->idx
[0].offset
);
1100 case WINED3DSPR_CONST
:
1101 if (!pshader
&& reg
->idx
[0].rel_addr
)
1103 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1104 UINT rel_offset
= ctx
->target_version
== ARB
? shader_data
->rel_offset
: 0;
1107 if (reg_maps
->shader_version
.major
< 2)
1109 sprintf(rel_reg
, "A0.x");
1113 shader_arb_get_src_param(ins
, reg
->idx
[0].rel_addr
, 0, rel_reg
);
1114 if (ctx
->target_version
== ARB
)
1116 if (!strcmp(rel_reg
, "**aL_emul**"))
1120 shader_arb_request_a0(ins
, rel_reg
);
1121 sprintf(rel_reg
, "A0.x");
1126 sprintf(register_name
, "C[%u]", ctx
->aL
+ reg
->idx
[0].offset
);
1127 else if (reg
->idx
[0].offset
>= rel_offset
)
1128 sprintf(register_name
, "C[%s + %u]", rel_reg
, reg
->idx
[0].offset
- rel_offset
);
1130 sprintf(register_name
, "C[%s - %u]", rel_reg
, rel_offset
- reg
->idx
[0].offset
);
1134 if (reg_maps
->usesrelconstF
)
1135 sprintf(register_name
, "C[%u]", reg
->idx
[0].offset
);
1137 sprintf(register_name
, "C%u", reg
->idx
[0].offset
);
1141 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
1144 if (reg_maps
->shader_version
.major
== 1
1145 && reg_maps
->shader_version
.minor
<= 3)
1146 /* In ps <= 1.3, Tx is a temporary register as destination
1147 * to all instructions, and as source to most instructions.
1148 * For some instructions it is the texcoord input. Those
1149 * instructions know about the special use. */
1150 sprintf(register_name
, "T%u", reg
->idx
[0].offset
);
1152 /* In ps 1.4 and 2.x Tx is always a (read-only) varying. */
1153 sprintf(register_name
, "fragment.texcoord[%u]", reg
->idx
[0].offset
);
1157 if (reg_maps
->shader_version
.major
== 1 || ctx
->target_version
>= NV2
)
1158 sprintf(register_name
, "A%u", reg
->idx
[0].offset
);
1160 sprintf(register_name
, "A%u_SHADOW", reg
->idx
[0].offset
);
1164 case WINED3DSPR_COLOROUT
:
1165 if (ctx
->ps_post_process
&& !reg
->idx
[0].offset
)
1167 strcpy(register_name
, "TMP_COLOR");
1171 if (ctx
->cur_ps_args
->super
.srgb_correction
)
1172 FIXME("sRGB correction on higher render targets.\n");
1173 if (reg_maps
->rt_mask
> 1)
1174 sprintf(register_name
, "result.color[%u]", reg
->idx
[0].offset
);
1176 strcpy(register_name
, "result.color");
1180 case WINED3DSPR_RASTOUT
:
1181 if (reg
->idx
[0].offset
== 1)
1182 sprintf(register_name
, "%s", ctx
->fog_output
);
1184 sprintf(register_name
, "%s", rastout_reg_names
[reg
->idx
[0].offset
]);
1187 case WINED3DSPR_DEPTHOUT
:
1188 strcpy(register_name
, "result.depth");
1191 case WINED3DSPR_ATTROUT
:
1192 /* case WINED3DSPR_OUTPUT: */
1194 sprintf(register_name
, "oD[%u]", reg
->idx
[0].offset
);
1196 strcpy(register_name
, ctx
->color_output
[reg
->idx
[0].offset
]);
1199 case WINED3DSPR_TEXCRDOUT
:
1201 sprintf(register_name
, "oT[%u]", reg
->idx
[0].offset
);
1202 else if (reg_maps
->shader_version
.major
< 3)
1203 strcpy(register_name
, ctx
->texcrd_output
[reg
->idx
[0].offset
]);
1205 strcpy(register_name
, ctx
->vs_output
[reg
->idx
[0].offset
]);
1208 case WINED3DSPR_LOOP
:
1209 if(ctx
->target_version
>= NV2
)
1211 /* Pshader has an implicitly declared loop index counter A0.x that cannot be renamed */
1212 if(pshader
) sprintf(register_name
, "A0.x");
1213 else sprintf(register_name
, "aL.y");
1217 /* Unfortunately this code cannot return the value of ctx->aL here. An immediate value
1218 * would be valid, but if aL is used for indexing(its only use), there's likely an offset,
1219 * thus the result would be something like C[15 + 30], which is not valid in the ARB program
1220 * grammar. So return a marker for the emulated aL and intercept it in constant and varying
1223 sprintf(register_name
, "**aL_emul**");
1228 case WINED3DSPR_CONSTINT
:
1229 sprintf(register_name
, "I%u", reg
->idx
[0].offset
);
1232 case WINED3DSPR_MISCTYPE
:
1233 if (!reg
->idx
[0].offset
)
1234 sprintf(register_name
, "vpos");
1235 else if (reg
->idx
[0].offset
== 1)
1236 sprintf(register_name
, "fragment.facing.x");
1238 FIXME("Unknown MISCTYPE register index %u.\n", reg
->idx
[0].offset
);
1242 FIXME("Unhandled register type %#x[%u].\n", reg
->type
, reg
->idx
[0].offset
);
1243 sprintf(register_name
, "unrecognized_register[%u]", reg
->idx
[0].offset
);
1248 static void shader_arb_get_dst_param(const struct wined3d_shader_instruction
*ins
,
1249 const struct wined3d_shader_dst_param
*wined3d_dst
, char *str
)
1251 char register_name
[255];
1255 shader_arb_get_register_name(ins
, &wined3d_dst
->reg
, register_name
, &is_color
);
1256 strcpy(str
, register_name
);
1258 shader_arb_get_write_mask(ins
, wined3d_dst
, write_mask
);
1259 strcat(str
, write_mask
);
1262 static const char *shader_arb_get_fixup_swizzle(enum fixup_channel_source channel_source
)
1264 switch(channel_source
)
1266 case CHANNEL_SOURCE_ZERO
: return "0";
1267 case CHANNEL_SOURCE_ONE
: return "1";
1268 case CHANNEL_SOURCE_X
: return "x";
1269 case CHANNEL_SOURCE_Y
: return "y";
1270 case CHANNEL_SOURCE_Z
: return "z";
1271 case CHANNEL_SOURCE_W
: return "w";
1273 FIXME("Unhandled channel source %#x\n", channel_source
);
1278 struct color_fixup_masks
1284 static struct color_fixup_masks
calc_color_correction(struct color_fixup_desc fixup
, DWORD dst_mask
)
1286 struct color_fixup_masks masks
= {0, 0};
1288 if (is_complex_fixup(fixup
))
1290 enum complex_fixup complex_fixup
= get_complex_fixup(fixup
);
1291 FIXME("Complex fixup (%#x) not supported\n", complex_fixup
);
1295 if (fixup
.x_source
!= CHANNEL_SOURCE_X
)
1296 masks
.source
|= WINED3DSP_WRITEMASK_0
;
1297 if (fixup
.y_source
!= CHANNEL_SOURCE_Y
)
1298 masks
.source
|= WINED3DSP_WRITEMASK_1
;
1299 if (fixup
.z_source
!= CHANNEL_SOURCE_Z
)
1300 masks
.source
|= WINED3DSP_WRITEMASK_2
;
1301 if (fixup
.w_source
!= CHANNEL_SOURCE_W
)
1302 masks
.source
|= WINED3DSP_WRITEMASK_3
;
1303 masks
.source
&= dst_mask
;
1305 if (fixup
.x_sign_fixup
)
1306 masks
.sign
|= WINED3DSP_WRITEMASK_0
;
1307 if (fixup
.y_sign_fixup
)
1308 masks
.sign
|= WINED3DSP_WRITEMASK_1
;
1309 if (fixup
.z_sign_fixup
)
1310 masks
.sign
|= WINED3DSP_WRITEMASK_2
;
1311 if (fixup
.w_sign_fixup
)
1312 masks
.sign
|= WINED3DSP_WRITEMASK_3
;
1313 masks
.sign
&= dst_mask
;
1318 static void gen_color_correction(struct wined3d_string_buffer
*buffer
, const char *dst
,
1319 const char *src
, const char *one
, const char *two
,
1320 struct color_fixup_desc fixup
, struct color_fixup_masks masks
)
1322 const char *sign_fixup_src
= dst
;
1327 sign_fixup_src
= "TA";
1329 shader_addline(buffer
, "SWZ %s, %s, %s, %s, %s, %s;\n", sign_fixup_src
, src
,
1330 shader_arb_get_fixup_swizzle(fixup
.x_source
), shader_arb_get_fixup_swizzle(fixup
.y_source
),
1331 shader_arb_get_fixup_swizzle(fixup
.z_source
), shader_arb_get_fixup_swizzle(fixup
.w_source
));
1333 else if (masks
.sign
)
1335 sign_fixup_src
= src
;
1341 char *ptr
= reg_mask
;
1343 if (masks
.sign
!= WINED3DSP_WRITEMASK_ALL
)
1346 if (masks
.sign
& WINED3DSP_WRITEMASK_0
)
1348 if (masks
.sign
& WINED3DSP_WRITEMASK_1
)
1350 if (masks
.sign
& WINED3DSP_WRITEMASK_2
)
1352 if (masks
.sign
& WINED3DSP_WRITEMASK_3
)
1357 shader_addline(buffer
, "MAD %s%s, %s, %s, -%s;\n", dst
, reg_mask
, sign_fixup_src
, two
, one
);
1361 static const char *shader_arb_get_modifier(const struct wined3d_shader_instruction
*ins
)
1364 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1365 if (!ins
->dst_count
) return "";
1367 mod
= ins
->dst
[0].modifiers
;
1369 /* Silently ignore PARTIALPRECISION if its not supported */
1370 if(priv
->target_version
== ARB
) mod
&= ~WINED3DSPDM_PARTIALPRECISION
;
1372 if(mod
& WINED3DSPDM_MSAMPCENTROID
)
1374 FIXME("Unhandled modifier WINED3DSPDM_MSAMPCENTROID\n");
1375 mod
&= ~WINED3DSPDM_MSAMPCENTROID
;
1380 case WINED3DSPDM_SATURATE
| WINED3DSPDM_PARTIALPRECISION
:
1383 case WINED3DSPDM_SATURATE
:
1386 case WINED3DSPDM_PARTIALPRECISION
:
1393 FIXME("Unknown modifiers 0x%08x\n", mod
);
1398 #define TEX_PROJ 0x1
1399 #define TEX_BIAS 0x2
1401 #define TEX_DERIV 0x10
1403 static void shader_hw_sample(const struct wined3d_shader_instruction
*ins
, DWORD sampler_idx
,
1404 const char *dst_str
, const char *coord_reg
, WORD flags
, const char *dsx
, const char *dsy
)
1406 enum wined3d_shader_resource_type resource_type
= ins
->ctx
->reg_maps
->resource_info
[sampler_idx
].type
;
1407 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1408 const char *tex_type
;
1409 BOOL np2_fixup
= FALSE
;
1410 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1412 BOOL pshader
= shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
1413 const struct wined3d_shader
*shader
;
1414 const struct wined3d_device
*device
;
1415 const struct wined3d_gl_info
*gl_info
;
1416 const char *tex_dst
= dst_str
;
1417 struct color_fixup_masks masks
;
1419 /* D3D vertex shader sampler IDs are vertex samplers(0-3), not global d3d samplers */
1420 if(!pshader
) sampler_idx
+= MAX_FRAGMENT_SAMPLERS
;
1422 switch (resource_type
)
1424 case WINED3D_SHADER_RESOURCE_TEXTURE_1D
:
1428 case WINED3D_SHADER_RESOURCE_TEXTURE_2D
:
1429 shader
= ins
->ctx
->shader
;
1430 device
= shader
->device
;
1431 gl_info
= &device
->adapter
->gl_info
;
1433 if (pshader
&& priv
->cur_ps_args
->super
.np2_fixup
& (1u << sampler_idx
)
1434 && gl_info
->supported
[ARB_TEXTURE_RECTANGLE
])
1438 if (shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
))
1440 if (priv
->cur_np2fixup_info
->super
.active
& (1u << sampler_idx
))
1442 if (flags
) FIXME("Only ordinary sampling from NP2 textures is supported.\n");
1443 else np2_fixup
= TRUE
;
1448 case WINED3D_SHADER_RESOURCE_TEXTURE_3D
:
1452 case WINED3D_SHADER_RESOURCE_TEXTURE_CUBE
:
1457 ERR("Unexpected resource type %#x.\n", resource_type
);
1461 /* TEX, TXL, TXD and TXP do not support the "H" modifier,
1462 * so don't use shader_arb_get_modifier
1464 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
) mod
= "_SAT";
1467 /* Fragment samplers always have indentity mapping */
1468 if(sampler_idx
>= MAX_FRAGMENT_SAMPLERS
)
1470 sampler_idx
= priv
->cur_vs_args
->vertex
.samplers
[sampler_idx
- MAX_FRAGMENT_SAMPLERS
];
1475 masks
= calc_color_correction(priv
->cur_ps_args
->super
.color_fixup
[sampler_idx
],
1476 ins
->dst
[0].write_mask
);
1478 if (masks
.source
|| masks
.sign
)
1482 if (flags
& TEX_DERIV
)
1484 if(flags
& TEX_PROJ
) FIXME("Projected texture sampling with custom derivatives\n");
1485 if(flags
& TEX_BIAS
) FIXME("Biased texture sampling with custom derivatives\n");
1486 shader_addline(buffer
, "TXD%s %s, %s, %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
,
1487 dsx
, dsy
, sampler_idx
, tex_type
);
1489 else if(flags
& TEX_LOD
)
1491 if(flags
& TEX_PROJ
) FIXME("Projected texture sampling with explicit lod\n");
1492 if(flags
& TEX_BIAS
) FIXME("Biased texture sampling with explicit lod\n");
1493 shader_addline(buffer
, "TXL%s %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
,
1494 sampler_idx
, tex_type
);
1496 else if (flags
& TEX_BIAS
)
1498 /* Shouldn't be possible, but let's check for it */
1499 if(flags
& TEX_PROJ
) FIXME("Biased and Projected texture sampling\n");
1500 /* TXB takes the 4th component of the source vector automatically, as d3d. Nothing more to do */
1501 shader_addline(buffer
, "TXB%s %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
, sampler_idx
, tex_type
);
1503 else if (flags
& TEX_PROJ
)
1505 shader_addline(buffer
, "TXP%s %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
, sampler_idx
, tex_type
);
1511 const unsigned char idx
= priv
->cur_np2fixup_info
->super
.idx
[sampler_idx
];
1512 shader_addline(buffer
, "MUL TA, np2fixup[%u].%s, %s;\n", idx
>> 1,
1513 (idx
% 2) ? "zwxy" : "xyzw", coord_reg
);
1515 shader_addline(buffer
, "TEX%s %s, TA, texture[%u], %s;\n", mod
, tex_dst
, sampler_idx
, tex_type
);
1518 shader_addline(buffer
, "TEX%s %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
, sampler_idx
, tex_type
);
1523 gen_color_correction(buffer
, dst_str
, tex_dst
,
1524 arb_get_helper_value(WINED3D_SHADER_TYPE_PIXEL
, ARB_ONE
),
1525 arb_get_helper_value(WINED3D_SHADER_TYPE_PIXEL
, ARB_TWO
),
1526 priv
->cur_ps_args
->super
.color_fixup
[sampler_idx
], masks
);
1530 static void shader_arb_get_src_param(const struct wined3d_shader_instruction
*ins
,
1531 const struct wined3d_shader_src_param
*src
, unsigned int tmpreg
, char *outregstr
)
1533 /* Generate a line that does the input modifier computation and return the input register to use */
1534 BOOL is_color
= FALSE
, insert_line
;
1537 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1538 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1539 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
1540 const char *two
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_TWO
);
1542 /* Assume a new line will be added */
1545 /* Get register name */
1546 shader_arb_get_register_name(ins
, &src
->reg
, regstr
, &is_color
);
1547 shader_arb_get_swizzle(src
, is_color
, swzstr
);
1549 switch (src
->modifiers
)
1551 case WINED3DSPSM_NONE
:
1552 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
1553 insert_line
= FALSE
;
1555 case WINED3DSPSM_NEG
:
1556 sprintf(outregstr
, "-%s%s", regstr
, swzstr
);
1557 insert_line
= FALSE
;
1559 case WINED3DSPSM_BIAS
:
1560 shader_addline(buffer
, "ADD T%c, %s, -coefdiv.x;\n", 'A' + tmpreg
, regstr
);
1562 case WINED3DSPSM_BIASNEG
:
1563 shader_addline(buffer
, "ADD T%c, -%s, coefdiv.x;\n", 'A' + tmpreg
, regstr
);
1565 case WINED3DSPSM_SIGN
:
1566 shader_addline(buffer
, "MAD T%c, %s, %s, -%s;\n", 'A' + tmpreg
, regstr
, two
, one
);
1568 case WINED3DSPSM_SIGNNEG
:
1569 shader_addline(buffer
, "MAD T%c, %s, -%s, %s;\n", 'A' + tmpreg
, regstr
, two
, one
);
1571 case WINED3DSPSM_COMP
:
1572 shader_addline(buffer
, "SUB T%c, %s, %s;\n", 'A' + tmpreg
, one
, regstr
);
1574 case WINED3DSPSM_X2
:
1575 shader_addline(buffer
, "ADD T%c, %s, %s;\n", 'A' + tmpreg
, regstr
, regstr
);
1577 case WINED3DSPSM_X2NEG
:
1578 shader_addline(buffer
, "ADD T%c, -%s, -%s;\n", 'A' + tmpreg
, regstr
, regstr
);
1580 case WINED3DSPSM_DZ
:
1581 shader_addline(buffer
, "RCP T%c, %s.z;\n", 'A' + tmpreg
, regstr
);
1582 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
1584 case WINED3DSPSM_DW
:
1585 shader_addline(buffer
, "RCP T%c, %s.w;\n", 'A' + tmpreg
, regstr
);
1586 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
1588 case WINED3DSPSM_ABS
:
1589 if(ctx
->target_version
>= NV2
) {
1590 sprintf(outregstr
, "|%s%s|", regstr
, swzstr
);
1591 insert_line
= FALSE
;
1593 shader_addline(buffer
, "ABS T%c, %s;\n", 'A' + tmpreg
, regstr
);
1596 case WINED3DSPSM_ABSNEG
:
1597 if(ctx
->target_version
>= NV2
) {
1598 sprintf(outregstr
, "-|%s%s|", regstr
, swzstr
);
1600 shader_addline(buffer
, "ABS T%c, %s;\n", 'A' + tmpreg
, regstr
);
1601 sprintf(outregstr
, "-T%c%s", 'A' + tmpreg
, swzstr
);
1603 insert_line
= FALSE
;
1606 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
1607 insert_line
= FALSE
;
1610 /* Return modified or original register, with swizzle */
1612 sprintf(outregstr
, "T%c%s", 'A' + tmpreg
, swzstr
);
1615 static void pshader_hw_bem(const struct wined3d_shader_instruction
*ins
)
1617 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1618 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1619 DWORD sampler_code
= dst
->reg
.idx
[0].offset
;
1621 char src_name
[2][50];
1623 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1625 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed
1627 * Keep in mind that src_name[1] can be "TB" and src_name[0] can be "TA" because modifiers like _x2 are valid
1628 * with bem. So delay loading the first parameter until after the perturbation calculation which needs two
1631 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1632 shader_addline(buffer
, "SWZ TA, bumpenvmat%d, x, z, 0, 0;\n", sampler_code
);
1633 shader_addline(buffer
, "DP3 TC.r, TA, %s;\n", src_name
[1]);
1634 shader_addline(buffer
, "SWZ TA, bumpenvmat%d, y, w, 0, 0;\n", sampler_code
);
1635 shader_addline(buffer
, "DP3 TC.g, TA, %s;\n", src_name
[1]);
1637 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1638 shader_addline(buffer
, "ADD %s, %s, TC;\n", dst_name
, src_name
[0]);
1641 static DWORD
negate_modifiers(DWORD mod
, char *extra_char
)
1646 case WINED3DSPSM_NONE
: return WINED3DSPSM_NEG
;
1647 case WINED3DSPSM_NEG
: return WINED3DSPSM_NONE
;
1648 case WINED3DSPSM_BIAS
: return WINED3DSPSM_BIASNEG
;
1649 case WINED3DSPSM_BIASNEG
: return WINED3DSPSM_BIAS
;
1650 case WINED3DSPSM_SIGN
: return WINED3DSPSM_SIGNNEG
;
1651 case WINED3DSPSM_SIGNNEG
: return WINED3DSPSM_SIGN
;
1652 case WINED3DSPSM_COMP
: *extra_char
= '-'; return WINED3DSPSM_COMP
;
1653 case WINED3DSPSM_X2
: return WINED3DSPSM_X2NEG
;
1654 case WINED3DSPSM_X2NEG
: return WINED3DSPSM_X2
;
1655 case WINED3DSPSM_DZ
: *extra_char
= '-'; return WINED3DSPSM_DZ
;
1656 case WINED3DSPSM_DW
: *extra_char
= '-'; return WINED3DSPSM_DW
;
1657 case WINED3DSPSM_ABS
: return WINED3DSPSM_ABSNEG
;
1658 case WINED3DSPSM_ABSNEG
: return WINED3DSPSM_ABS
;
1660 FIXME("Unknown modifier %u\n", mod
);
1664 static void pshader_hw_cnd(const struct wined3d_shader_instruction
*ins
)
1666 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1667 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1669 char src_name
[3][50];
1670 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1671 ins
->ctx
->reg_maps
->shader_version
.minor
);
1673 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1674 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1676 if (shader_version
<= WINED3D_SHADER_VERSION(1, 3) && ins
->coissue
1677 && ins
->dst
->write_mask
!= WINED3DSP_WRITEMASK_3
)
1679 shader_addline(buffer
, "MOV%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
[1]);
1683 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
1686 /* src0 may have a negate srcmod set, so we can't blindly add "-" to the name */
1687 src0_copy
.modifiers
= negate_modifiers(src0_copy
.modifiers
, &extra_neg
);
1689 shader_arb_get_src_param(ins
, &src0_copy
, 0, src_name
[0]);
1690 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1691 shader_addline(buffer
, "ADD TA, %c%s, coefdiv.x;\n", extra_neg
, src_name
[0]);
1692 shader_addline(buffer
, "CMP%s %s, TA, %s, %s;\n", shader_arb_get_modifier(ins
),
1693 dst_name
, src_name
[1], src_name
[2]);
1697 static void pshader_hw_cmp(const struct wined3d_shader_instruction
*ins
)
1699 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1700 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1702 char src_name
[3][50];
1704 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1706 /* Generate input register names (with modifiers) */
1707 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1708 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1709 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1711 shader_addline(buffer
, "CMP%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1712 dst_name
, src_name
[0], src_name
[2], src_name
[1]);
1715 /** Process the WINED3DSIO_DP2ADD instruction in ARB.
1716 * dst = dot2(src0, src1) + src2 */
1717 static void pshader_hw_dp2add(const struct wined3d_shader_instruction
*ins
)
1719 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1720 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1722 char src_name
[3][50];
1723 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1725 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1726 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1727 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1729 if(ctx
->target_version
>= NV3
)
1731 /* GL_NV_fragment_program2 has a 1:1 matching instruction */
1732 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1733 shader_addline(buffer
, "DP2A%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1734 dst_name
, src_name
[0], src_name
[1], src_name
[2]);
1736 else if(ctx
->target_version
>= NV2
)
1738 /* dst.x = src2.?, src0.x, src1.x + src0.y * src1.y
1739 * dst.y = src2.?, src0.x, src1.z + src0.y * src1.w
1740 * dst.z = src2.?, src0.x, src1.x + src0.y * src1.y
1741 * dst.z = src2.?, src0.x, src1.z + src0.y * src1.w
1743 * Make sure that src1.zw = src1.xy, then we get a classic dp2add
1745 * .xyxy and other swizzles that we could get with this are not valid in
1746 * plain ARBfp, but luckily the NV extension grammar lifts this limitation.
1748 struct wined3d_shader_src_param tmp_param
= ins
->src
[1];
1749 DWORD swizzle
= tmp_param
.swizzle
& 0xf; /* Selects .xy */
1750 tmp_param
.swizzle
= swizzle
| (swizzle
<< 4); /* Creates .xyxy */
1752 shader_arb_get_src_param(ins
, &tmp_param
, 1, src_name
[1]);
1754 shader_addline(buffer
, "X2D%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1755 dst_name
, src_name
[2], src_name
[0], src_name
[1]);
1759 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1760 /* Emulate a DP2 with a DP3 and 0.0. Don't use the dest as temp register, it could be src[1] or src[2]
1761 * src_name[0] can be TA, but TA is a private temp for modifiers, so it is save to overwrite
1763 shader_addline(buffer
, "MOV TA, %s;\n", src_name
[0]);
1764 shader_addline(buffer
, "MOV TA.z, 0.0;\n");
1765 shader_addline(buffer
, "DP3 TA, TA, %s;\n", src_name
[1]);
1766 shader_addline(buffer
, "ADD%s %s, TA, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
[2]);
1770 /* Map the opcode 1-to-1 to the GL code */
1771 static void shader_hw_map2gl(const struct wined3d_shader_instruction
*ins
)
1773 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1774 const char *instruction
;
1775 char arguments
[256], dst_str
[50];
1777 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1779 switch (ins
->handler_idx
)
1781 case WINED3DSIH_ABS
: instruction
= "ABS"; break;
1782 case WINED3DSIH_ADD
: instruction
= "ADD"; break;
1783 case WINED3DSIH_CRS
: instruction
= "XPD"; break;
1784 case WINED3DSIH_DP3
: instruction
= "DP3"; break;
1785 case WINED3DSIH_DP4
: instruction
= "DP4"; break;
1786 case WINED3DSIH_DST
: instruction
= "DST"; break;
1787 case WINED3DSIH_FRC
: instruction
= "FRC"; break;
1788 case WINED3DSIH_LIT
: instruction
= "LIT"; break;
1789 case WINED3DSIH_LRP
: instruction
= "LRP"; break;
1790 case WINED3DSIH_MAD
: instruction
= "MAD"; break;
1791 case WINED3DSIH_MAX
: instruction
= "MAX"; break;
1792 case WINED3DSIH_MIN
: instruction
= "MIN"; break;
1793 case WINED3DSIH_MOV
: instruction
= "MOV"; break;
1794 case WINED3DSIH_MUL
: instruction
= "MUL"; break;
1795 case WINED3DSIH_SGE
: instruction
= "SGE"; break;
1796 case WINED3DSIH_SLT
: instruction
= "SLT"; break;
1797 case WINED3DSIH_SUB
: instruction
= "SUB"; break;
1798 case WINED3DSIH_MOVA
:instruction
= "ARR"; break;
1799 case WINED3DSIH_DSX
: instruction
= "DDX"; break;
1800 default: instruction
= "";
1801 FIXME("Unhandled opcode %s.\n", debug_d3dshaderinstructionhandler(ins
->handler_idx
));
1805 /* Note that shader_arb_add_dst_param() adds spaces. */
1806 arguments
[0] = '\0';
1807 shader_arb_get_dst_param(ins
, dst
, dst_str
);
1808 for (i
= 0; i
< ins
->src_count
; ++i
)
1811 strcat(arguments
, ", ");
1812 shader_arb_get_src_param(ins
, &ins
->src
[i
], i
, operand
);
1813 strcat(arguments
, operand
);
1815 shader_addline(buffer
, "%s%s %s%s;\n", instruction
, shader_arb_get_modifier(ins
), dst_str
, arguments
);
1818 static void shader_hw_nop(const struct wined3d_shader_instruction
*ins
) {}
1820 static DWORD
shader_arb_select_component(DWORD swizzle
, DWORD component
)
1822 return ((swizzle
>> 2 * component
) & 0x3) * 0x55;
1825 static void shader_hw_mov(const struct wined3d_shader_instruction
*ins
)
1827 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
1828 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
1829 BOOL pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
1830 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1831 const char *zero
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ZERO
);
1832 const char *one
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ONE
);
1833 const char *two
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_TWO
);
1835 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1836 char src0_param
[256];
1838 if (ins
->handler_idx
== WINED3DSIH_MOVA
)
1840 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1842 const char *offset
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_VS_REL_OFFSET
);
1844 if(ctx
->target_version
>= NV2
) {
1845 shader_hw_map2gl(ins
);
1848 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_param
);
1849 shader_arb_get_write_mask(ins
, &ins
->dst
[0], write_mask
);
1851 /* This implements the mova formula used in GLSL. The first two instructions
1852 * prepare the sign() part. Note that it is fine to have my_sign(0.0) = 1.0
1856 * A0.x = arl(floor(abs(0.0) + 0.5) * 1.0) = floor(0.5) = 0.0 since arl does a floor
1858 * The ARL is performed when A0 is used - the requested component is read from A0_SHADOW into
1859 * A0.x. We can use the overwritten component of A0_shadow as temporary storage for the sign.
1861 shader_addline(buffer
, "SGE A0_SHADOW%s, %s, %s;\n", write_mask
, src0_param
, zero
);
1862 shader_addline(buffer
, "MAD A0_SHADOW%s, A0_SHADOW, %s, -%s;\n", write_mask
, two
, one
);
1864 shader_addline(buffer
, "ABS TA%s, %s;\n", write_mask
, src0_param
);
1865 shader_addline(buffer
, "ADD TA%s, TA, rel_addr_const.x;\n", write_mask
);
1866 shader_addline(buffer
, "FLR TA%s, TA;\n", write_mask
);
1867 if (shader_data
->rel_offset
)
1869 shader_addline(buffer
, "ADD TA%s, TA, %s;\n", write_mask
, offset
);
1871 shader_addline(buffer
, "MUL A0_SHADOW%s, TA, A0_SHADOW;\n", write_mask
);
1873 ((struct shader_arb_ctx_priv
*)ins
->ctx
->backend_data
)->addr_reg
[0] = '\0';
1875 else if (reg_maps
->shader_version
.major
== 1
1876 && !shader_is_pshader_version(reg_maps
->shader_version
.type
)
1877 && ins
->dst
[0].reg
.type
== WINED3DSPR_ADDR
)
1879 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1880 src0_param
[0] = '\0';
1882 if (shader_data
->rel_offset
&& ctx
->target_version
== ARB
)
1884 const char *offset
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_VS_REL_OFFSET
);
1885 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_param
);
1886 shader_addline(buffer
, "ADD TA.x, %s, %s;\n", src0_param
, offset
);
1887 shader_addline(buffer
, "ARL A0.x, TA.x;\n");
1891 /* Apple's ARB_vertex_program implementation does not accept an ARL source argument
1892 * with more than one component. Thus replicate the first source argument over all
1893 * 4 components. For example, .xyzw -> .x (or better: .xxxx), .zwxy -> .z, etc) */
1894 struct wined3d_shader_src_param tmp_src
= ins
->src
[0];
1895 tmp_src
.swizzle
= shader_arb_select_component(tmp_src
.swizzle
, 0);
1896 shader_arb_get_src_param(ins
, &tmp_src
, 0, src0_param
);
1897 shader_addline(buffer
, "ARL A0.x, %s;\n", src0_param
);
1900 else if (ins
->dst
[0].reg
.type
== WINED3DSPR_COLOROUT
&& !ins
->dst
[0].reg
.idx
[0].offset
&& pshader
)
1902 if (ctx
->ps_post_process
&& shader
->u
.ps
.color0_mov
)
1904 shader_addline(buffer
, "#mov handled in srgb write or fog code\n");
1907 shader_hw_map2gl(ins
);
1911 shader_hw_map2gl(ins
);
1915 static void pshader_hw_texkill(const struct wined3d_shader_instruction
*ins
)
1917 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1918 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1921 /* No swizzles are allowed in d3d's texkill. PS 1.x ignores the 4th component as documented,
1922 * but >= 2.0 honors it (undocumented, but tested by the d3d9 testsuite)
1924 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1926 if (ins
->ctx
->reg_maps
->shader_version
.major
>= 2)
1928 const char *kilsrc
= "TA";
1931 shader_arb_get_register_name(ins
, &dst
->reg
, reg_dest
, &is_color
);
1932 if(dst
->write_mask
== WINED3DSP_WRITEMASK_ALL
)
1938 /* Sigh. KIL doesn't support swizzles/writemasks. KIL passes a writemask, but ".xy" for example
1939 * is not valid as a swizzle in ARB (needs ".xyyy"). Use SWZ to load the register properly, and set
1940 * masked out components to 0(won't kill)
1942 char x
= '0', y
= '0', z
= '0', w
= '0';
1943 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
) x
= 'x';
1944 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
) y
= 'y';
1945 if(dst
->write_mask
& WINED3DSP_WRITEMASK_2
) z
= 'z';
1946 if(dst
->write_mask
& WINED3DSP_WRITEMASK_3
) w
= 'w';
1947 shader_addline(buffer
, "SWZ TA, %s, %c, %c, %c, %c;\n", reg_dest
, x
, y
, z
, w
);
1949 shader_addline(buffer
, "KIL %s;\n", kilsrc
);
1953 /* ARB fp doesn't like swizzles on the parameter of the KIL instruction. To mask the 4th component,
1954 * copy the register into our general purpose TMP variable, overwrite .w and pass TMP to KIL
1956 * ps_1_3 shaders use the texcoord incarnation of the Tx register. ps_1_4 shaders can use the same,
1957 * or pass in any temporary register(in shader phase 2)
1959 if (ins
->ctx
->reg_maps
->shader_version
.minor
<= 3)
1960 sprintf(reg_dest
, "fragment.texcoord[%u]", dst
->reg
.idx
[0].offset
);
1962 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1963 shader_addline(buffer
, "SWZ TA, %s, x, y, z, 1;\n", reg_dest
);
1964 shader_addline(buffer
, "KIL TA;\n");
1968 static void pshader_hw_tex(const struct wined3d_shader_instruction
*ins
)
1970 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1971 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1972 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1973 ins
->ctx
->reg_maps
->shader_version
.minor
);
1974 struct wined3d_shader_src_param src
;
1978 DWORD reg_sampler_code
;
1980 BOOL swizzle_coord
= FALSE
;
1982 /* All versions have a destination register */
1983 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1985 /* 1.0-1.4: Use destination register number as texture code.
1986 2.0+: Use provided sampler number as texture code. */
1987 if (shader_version
< WINED3D_SHADER_VERSION(2,0))
1988 reg_sampler_code
= dst
->reg
.idx
[0].offset
;
1990 reg_sampler_code
= ins
->src
[1].reg
.idx
[0].offset
;
1992 /* 1.0-1.3: Use the texcoord varying.
1993 1.4+: Use provided coordinate source register. */
1994 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
1995 sprintf(reg_coord
, "fragment.texcoord[%u]", reg_sampler_code
);
1997 /* TEX is the only instruction that can handle DW and DZ natively */
1999 if(src
.modifiers
== WINED3DSPSM_DW
) src
.modifiers
= WINED3DSPSM_NONE
;
2000 if(src
.modifiers
== WINED3DSPSM_DZ
) src
.modifiers
= WINED3DSPSM_NONE
;
2001 shader_arb_get_src_param(ins
, &src
, 0, reg_coord
);
2005 * 1.1, 1.2, 1.3: Use WINED3D_TSS_TEXTURETRANSFORMFLAGS
2006 * 1.4: Use WINED3DSPSM_DZ or WINED3DSPSM_DW on src[0]
2007 * 2.0+: Use WINED3DSI_TEXLD_PROJECT on the opcode
2009 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
2012 if (reg_sampler_code
< MAX_TEXTURES
)
2013 flags
= priv
->cur_ps_args
->super
.tex_transform
>> reg_sampler_code
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
;
2014 if (flags
& WINED3D_PSARGS_PROJECTED
)
2016 myflags
|= TEX_PROJ
;
2017 if ((flags
& ~WINED3D_PSARGS_PROJECTED
) == WINED3D_TTFF_COUNT3
)
2018 swizzle_coord
= TRUE
;
2021 else if (shader_version
< WINED3D_SHADER_VERSION(2,0))
2023 enum wined3d_shader_src_modifier src_mod
= ins
->src
[0].modifiers
;
2024 if (src_mod
== WINED3DSPSM_DZ
)
2026 swizzle_coord
= TRUE
;
2027 myflags
|= TEX_PROJ
;
2028 } else if(src_mod
== WINED3DSPSM_DW
) {
2029 myflags
|= TEX_PROJ
;
2032 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) myflags
|= TEX_PROJ
;
2033 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) myflags
|= TEX_BIAS
;
2038 /* TXP cannot handle DZ natively, so move the z coordinate to .w.
2039 * reg_coord is a read-only varying register, so we need a temp reg */
2040 shader_addline(ins
->ctx
->buffer
, "SWZ TA, %s, x, y, z, z;\n", reg_coord
);
2041 strcpy(reg_coord
, "TA");
2044 shader_hw_sample(ins
, reg_sampler_code
, reg_dest
, reg_coord
, myflags
, NULL
, NULL
);
2047 static void pshader_hw_texcoord(const struct wined3d_shader_instruction
*ins
)
2049 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2050 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2051 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
2052 ins
->ctx
->reg_maps
->shader_version
.minor
);
2055 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
2057 DWORD reg
= dst
->reg
.idx
[0].offset
;
2059 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2060 shader_addline(buffer
, "MOV_SAT %s, fragment.texcoord[%u];\n", dst_str
, reg
);
2064 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_src
);
2065 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2066 shader_addline(buffer
, "MOV %s, %s;\n", dst_str
, reg_src
);
2070 static void pshader_hw_texreg2ar(const struct wined3d_shader_instruction
*ins
)
2072 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2075 DWORD reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2079 /* Note that texreg2ar treats Tx as a temporary register, not as a varying */
2080 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2081 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2082 /* Move .x first in case src_str is "TA" */
2083 shader_addline(buffer
, "MOV TA.y, %s.x;\n", src_str
);
2084 shader_addline(buffer
, "MOV TA.x, %s.w;\n", src_str
);
2085 if (reg1
< MAX_TEXTURES
)
2087 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2088 flags
= priv
->cur_ps_args
->super
.tex_transform
>> reg1
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
;
2090 shader_hw_sample(ins
, reg1
, dst_str
, "TA", flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2093 static void pshader_hw_texreg2gb(const struct wined3d_shader_instruction
*ins
)
2095 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2097 DWORD reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2101 /* Note that texreg2gb treats Tx as a temporary register, not as a varying */
2102 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2103 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2104 shader_addline(buffer
, "MOV TA.x, %s.y;\n", src_str
);
2105 shader_addline(buffer
, "MOV TA.y, %s.z;\n", src_str
);
2106 shader_hw_sample(ins
, reg1
, dst_str
, "TA", 0, NULL
, NULL
);
2109 static void pshader_hw_texreg2rgb(const struct wined3d_shader_instruction
*ins
)
2111 DWORD reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2115 /* Note that texreg2rg treats Tx as a temporary register, not as a varying */
2116 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2117 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2118 shader_hw_sample(ins
, reg1
, dst_str
, src_str
, 0, NULL
, NULL
);
2121 static void pshader_hw_texbem(const struct wined3d_shader_instruction
*ins
)
2123 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2124 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2125 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2126 char reg_coord
[40], dst_reg
[50], src_reg
[50];
2127 DWORD reg_dest_code
;
2129 /* All versions have a destination register. The Tx where the texture coordinates come
2130 * from is the varying incarnation of the texture register
2132 reg_dest_code
= dst
->reg
.idx
[0].offset
;
2133 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_reg
);
2134 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_reg
);
2135 sprintf(reg_coord
, "fragment.texcoord[%u]", reg_dest_code
);
2137 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed
2138 * The Tx in which the perturbation map is stored is the tempreg incarnation of the texture register
2140 * GL_NV_fragment_program_option could handle this in one instruction via X2D:
2141 * X2D TA.xy, fragment.texcoord, T%u, bumpenvmat%u.xzyw
2143 * However, the NV extensions are never enabled for <= 2.0 shaders because of the performance penalty that
2144 * comes with it, and texbem is an 1.x only instruction. No 1.x instruction forces us to enable the NV
2147 shader_addline(buffer
, "SWZ TB, bumpenvmat%d, x, z, 0, 0;\n", reg_dest_code
);
2148 shader_addline(buffer
, "DP3 TA.x, TB, %s;\n", src_reg
);
2149 shader_addline(buffer
, "SWZ TB, bumpenvmat%d, y, w, 0, 0;\n", reg_dest_code
);
2150 shader_addline(buffer
, "DP3 TA.y, TB, %s;\n", src_reg
);
2152 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2153 * so we can't let the GL handle this.
2155 if ((priv
->cur_ps_args
->super
.tex_transform
>> reg_dest_code
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
)
2156 & WINED3D_PSARGS_PROJECTED
)
2158 shader_addline(buffer
, "RCP TB.w, %s.w;\n", reg_coord
);
2159 shader_addline(buffer
, "MUL TB.xy, %s, TB.w;\n", reg_coord
);
2160 shader_addline(buffer
, "ADD TA.xy, TA, TB;\n");
2162 shader_addline(buffer
, "ADD TA.xy, TA, %s;\n", reg_coord
);
2165 shader_hw_sample(ins
, reg_dest_code
, dst_reg
, "TA", 0, NULL
, NULL
);
2167 if (ins
->handler_idx
== WINED3DSIH_TEXBEML
)
2169 /* No src swizzles are allowed, so this is ok */
2170 shader_addline(buffer
, "MAD TA, %s.z, luminance%d.x, luminance%d.y;\n",
2171 src_reg
, reg_dest_code
, reg_dest_code
);
2172 shader_addline(buffer
, "MUL %s, %s, TA;\n", dst_reg
, dst_reg
);
2176 static void pshader_hw_texm3x2pad(const struct wined3d_shader_instruction
*ins
)
2178 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2179 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2180 char src0_name
[50], dst_name
[50];
2182 struct wined3d_shader_register tmp_reg
= ins
->dst
[0].reg
;
2184 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2185 /* The next instruction will be a texm3x2tex or texm3x2depth that writes to the uninitialized
2186 * T<reg+1> register. Use this register to store the calculated vector
2188 tmp_reg
.idx
[0].offset
= reg
+ 1;
2189 shader_arb_get_register_name(ins
, &tmp_reg
, dst_name
, &is_color
);
2190 shader_addline(buffer
, "DP3 %s.x, fragment.texcoord[%u], %s;\n", dst_name
, reg
, src0_name
);
2193 static void pshader_hw_texm3x2tex(const struct wined3d_shader_instruction
*ins
)
2195 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2197 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2198 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2204 /* We know that we're writing to the uninitialized T<reg> register, so use it for temporary storage */
2205 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2207 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2208 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2209 shader_addline(buffer
, "DP3 %s.y, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2210 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2211 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2214 static void pshader_hw_texm3x3pad(const struct wined3d_shader_instruction
*ins
)
2216 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2217 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2218 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2219 char src0_name
[50], dst_name
[50];
2220 struct wined3d_shader_register tmp_reg
= ins
->dst
[0].reg
;
2223 /* There are always 2 texm3x3pad instructions followed by one texm3x3[tex,vspec, ...] instruction, with
2224 * incrementing ins->dst[0].register_idx numbers. So the pad instruction already knows the final destination
2225 * register, and this register is uninitialized(otherwise the assembler complains that it is 'redeclared')
2227 tmp_reg
.idx
[0].offset
= reg
+ 2 - tex_mx
->current_row
;
2228 shader_arb_get_register_name(ins
, &tmp_reg
, dst_name
, &is_color
);
2230 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2231 shader_addline(buffer
, "DP3 %s.%c, fragment.texcoord[%u], %s;\n",
2232 dst_name
, 'x' + tex_mx
->current_row
, reg
, src0_name
);
2233 tex_mx
->texcoord_w
[tex_mx
->current_row
++] = reg
;
2236 static void pshader_hw_texm3x3tex(const struct wined3d_shader_instruction
*ins
)
2238 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2239 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2241 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2242 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2244 char src0_name
[50], dst_name
[50];
2247 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2248 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2249 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_name
, reg
, src0_name
);
2251 /* Sample the texture using the calculated coordinates */
2252 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2253 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2254 shader_hw_sample(ins
, reg
, dst_str
, dst_name
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2255 tex_mx
->current_row
= 0;
2258 static void pshader_hw_texm3x3vspec(const struct wined3d_shader_instruction
*ins
)
2260 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2261 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2263 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2264 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2270 /* Get the dst reg without writemask strings. We know this register is uninitialized, so we can use all
2271 * components for temporary data storage
2273 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2274 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2275 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2277 /* Construct the eye-ray vector from w coordinates */
2278 shader_addline(buffer
, "MOV TB.x, fragment.texcoord[%u].w;\n", tex_mx
->texcoord_w
[0]);
2279 shader_addline(buffer
, "MOV TB.y, fragment.texcoord[%u].w;\n", tex_mx
->texcoord_w
[1]);
2280 shader_addline(buffer
, "MOV TB.z, fragment.texcoord[%u].w;\n", reg
);
2282 /* Calculate reflection vector
2284 shader_addline(buffer
, "DP3 %s.w, %s, TB;\n", dst_reg
, dst_reg
);
2285 /* The .w is ignored when sampling, so I can use TB.w to calculate dot(N, N) */
2286 shader_addline(buffer
, "DP3 TB.w, %s, %s;\n", dst_reg
, dst_reg
);
2287 shader_addline(buffer
, "RCP TB.w, TB.w;\n");
2288 shader_addline(buffer
, "MUL %s.w, %s.w, TB.w;\n", dst_reg
, dst_reg
);
2289 shader_addline(buffer
, "MUL %s, %s.w, %s;\n", dst_reg
, dst_reg
, dst_reg
);
2290 shader_addline(buffer
, "MAD %s, coefmul.x, %s, -TB;\n", dst_reg
, dst_reg
);
2292 /* Sample the texture using the calculated coordinates */
2293 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2294 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2295 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2296 tex_mx
->current_row
= 0;
2299 static void pshader_hw_texm3x3spec(const struct wined3d_shader_instruction
*ins
)
2301 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2302 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2304 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2305 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2312 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2313 shader_arb_get_src_param(ins
, &ins
->src
[0], 1, src1_name
);
2314 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2315 /* Note: dst_reg.xy is input here, generated by two texm3x3pad instructions */
2316 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2318 /* Calculate reflection vector.
2321 * dst_reg.xyz = 2 * --------- * N - E
2324 * Which normalizes the normal vector
2326 shader_addline(buffer
, "DP3 %s.w, %s, %s;\n", dst_reg
, dst_reg
, src1_name
);
2327 shader_addline(buffer
, "DP3 TC.w, %s, %s;\n", dst_reg
, dst_reg
);
2328 shader_addline(buffer
, "RCP TC.w, TC.w;\n");
2329 shader_addline(buffer
, "MUL %s.w, %s.w, TC.w;\n", dst_reg
, dst_reg
);
2330 shader_addline(buffer
, "MUL %s, %s.w, %s;\n", dst_reg
, dst_reg
, dst_reg
);
2331 shader_addline(buffer
, "MAD %s, coefmul.x, %s, -%s;\n", dst_reg
, dst_reg
, src1_name
);
2333 /* Sample the texture using the calculated coordinates */
2334 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2335 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2336 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2337 tex_mx
->current_row
= 0;
2340 static void pshader_hw_texdepth(const struct wined3d_shader_instruction
*ins
)
2342 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2343 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2345 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2346 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2348 /* texdepth has an implicit destination, the fragment depth value. It's only parameter,
2349 * which is essentially an input, is the destination register because it is the first
2350 * parameter. According to the msdn, this must be register r5, but let's keep it more flexible
2351 * here(writemasks/swizzles are not valid on texdepth)
2353 shader_arb_get_dst_param(ins
, dst
, dst_name
);
2355 /* According to the msdn, the source register(must be r5) is unusable after
2356 * the texdepth instruction, so we're free to modify it
2358 shader_addline(buffer
, "MIN %s.y, %s.y, %s;\n", dst_name
, dst_name
, one
);
2360 /* How to deal with the special case dst_name.g == 0? if r != 0, then
2361 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
2362 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
2364 shader_addline(buffer
, "RCP %s.y, %s.y;\n", dst_name
, dst_name
);
2365 shader_addline(buffer
, "MUL TA.x, %s.x, %s.y;\n", dst_name
, dst_name
);
2366 shader_addline(buffer
, "MIN TA.x, TA.x, %s;\n", one
);
2367 shader_addline(buffer
, "MAX result.depth, TA.x, %s;\n", zero
);
2370 /** Process the WINED3DSIO_TEXDP3TEX instruction in ARB:
2371 * Take a 3-component dot product of the TexCoord[dstreg] and src,
2372 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
2373 static void pshader_hw_texdp3tex(const struct wined3d_shader_instruction
*ins
)
2375 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2376 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
[0].offset
;
2380 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2381 shader_addline(buffer
, "MOV TB, 0.0;\n");
2382 shader_addline(buffer
, "DP3 TB.x, fragment.texcoord[%u], %s;\n", sampler_idx
, src0
);
2384 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2385 shader_hw_sample(ins
, sampler_idx
, dst_str
, "TB", 0 /* Only one coord, can't be projected */, NULL
, NULL
);
2388 /** Process the WINED3DSIO_TEXDP3 instruction in ARB:
2389 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
2390 static void pshader_hw_texdp3(const struct wined3d_shader_instruction
*ins
)
2392 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2395 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2397 /* Handle output register */
2398 shader_arb_get_dst_param(ins
, dst
, dst_str
);
2399 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2400 shader_addline(buffer
, "DP3 %s, fragment.texcoord[%u], %s;\n", dst_str
, dst
->reg
.idx
[0].offset
, src0
);
2403 /** Process the WINED3DSIO_TEXM3X3 instruction in ARB
2404 * Perform the 3rd row of a 3x3 matrix multiply */
2405 static void pshader_hw_texm3x3(const struct wined3d_shader_instruction
*ins
)
2407 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2408 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2409 char dst_str
[50], dst_name
[50];
2413 shader_arb_get_dst_param(ins
, dst
, dst_str
);
2414 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2415 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2416 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_name
, dst
->reg
.idx
[0].offset
, src0
);
2417 shader_addline(buffer
, "MOV %s, %s;\n", dst_str
, dst_name
);
2420 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in ARB:
2421 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
2422 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
2423 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
2425 static void pshader_hw_texm3x2depth(const struct wined3d_shader_instruction
*ins
)
2427 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2428 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2429 char src0
[50], dst_name
[50];
2431 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2432 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2434 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2435 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2436 shader_addline(buffer
, "DP3 %s.y, fragment.texcoord[%u], %s;\n", dst_name
, dst
->reg
.idx
[0].offset
, src0
);
2438 /* How to deal with the special case dst_name.g == 0? if r != 0, then
2439 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
2440 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
2442 shader_addline(buffer
, "RCP %s.y, %s.y;\n", dst_name
, dst_name
);
2443 shader_addline(buffer
, "MUL %s.x, %s.x, %s.y;\n", dst_name
, dst_name
, dst_name
);
2444 shader_addline(buffer
, "MIN %s.x, %s.x, %s;\n", dst_name
, dst_name
, one
);
2445 shader_addline(buffer
, "MAX result.depth, %s.x, %s;\n", dst_name
, zero
);
2448 /** Handles transforming all WINED3DSIO_M?x? opcodes for
2449 Vertex/Pixel shaders to ARB_vertex_program codes */
2450 static void shader_hw_mnxn(const struct wined3d_shader_instruction
*ins
)
2453 int nComponents
= 0;
2454 struct wined3d_shader_dst_param tmp_dst
= {{0}};
2455 struct wined3d_shader_src_param tmp_src
[2] = {{{0}}};
2456 struct wined3d_shader_instruction tmp_ins
;
2458 memset(&tmp_ins
, 0, sizeof(tmp_ins
));
2460 /* Set constants for the temporary argument */
2461 tmp_ins
.ctx
= ins
->ctx
;
2462 tmp_ins
.dst_count
= 1;
2463 tmp_ins
.dst
= &tmp_dst
;
2464 tmp_ins
.src_count
= 2;
2465 tmp_ins
.src
= tmp_src
;
2467 switch(ins
->handler_idx
)
2469 case WINED3DSIH_M4x4
:
2471 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2473 case WINED3DSIH_M4x3
:
2475 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2477 case WINED3DSIH_M3x4
:
2479 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2481 case WINED3DSIH_M3x3
:
2483 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2485 case WINED3DSIH_M3x2
:
2487 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2490 FIXME("Unhandled opcode %s.\n", debug_d3dshaderinstructionhandler(ins
->handler_idx
));
2494 tmp_dst
= ins
->dst
[0];
2495 tmp_src
[0] = ins
->src
[0];
2496 tmp_src
[1] = ins
->src
[1];
2497 for (i
= 0; i
< nComponents
; ++i
)
2499 tmp_dst
.write_mask
= WINED3DSP_WRITEMASK_0
<< i
;
2500 shader_hw_map2gl(&tmp_ins
);
2501 ++tmp_src
[1].reg
.idx
[0].offset
;
2505 static DWORD
abs_modifier(DWORD mod
, BOOL
*need_abs
)
2511 case WINED3DSPSM_NONE
: return WINED3DSPSM_ABS
;
2512 case WINED3DSPSM_NEG
: return WINED3DSPSM_ABS
;
2513 case WINED3DSPSM_BIAS
: *need_abs
= TRUE
; return WINED3DSPSM_BIAS
;
2514 case WINED3DSPSM_BIASNEG
: *need_abs
= TRUE
; return WINED3DSPSM_BIASNEG
;
2515 case WINED3DSPSM_SIGN
: *need_abs
= TRUE
; return WINED3DSPSM_SIGN
;
2516 case WINED3DSPSM_SIGNNEG
: *need_abs
= TRUE
; return WINED3DSPSM_SIGNNEG
;
2517 case WINED3DSPSM_COMP
: *need_abs
= TRUE
; return WINED3DSPSM_COMP
;
2518 case WINED3DSPSM_X2
: *need_abs
= TRUE
; return WINED3DSPSM_X2
;
2519 case WINED3DSPSM_X2NEG
: *need_abs
= TRUE
; return WINED3DSPSM_X2NEG
;
2520 case WINED3DSPSM_DZ
: *need_abs
= TRUE
; return WINED3DSPSM_DZ
;
2521 case WINED3DSPSM_DW
: *need_abs
= TRUE
; return WINED3DSPSM_DW
;
2522 case WINED3DSPSM_ABS
: return WINED3DSPSM_ABS
;
2523 case WINED3DSPSM_ABSNEG
: return WINED3DSPSM_ABS
;
2525 FIXME("Unknown modifier %u\n", mod
);
2529 static void shader_hw_scalar_op(const struct wined3d_shader_instruction
*ins
)
2531 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2532 const char *instruction
;
2533 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
2534 BOOL need_abs
= FALSE
;
2539 switch(ins
->handler_idx
)
2541 case WINED3DSIH_RSQ
: instruction
= "RSQ"; break;
2542 case WINED3DSIH_RCP
: instruction
= "RCP"; break;
2543 case WINED3DSIH_EXPP
:
2544 if (ins
->ctx
->reg_maps
->shader_version
.major
< 2)
2546 instruction
= "EXP";
2550 case WINED3DSIH_EXP
:
2551 instruction
= "EX2";
2553 case WINED3DSIH_LOG
:
2554 case WINED3DSIH_LOGP
:
2555 /* The precision requirements suggest that LOGP matches ARBvp's LOG
2556 * instruction, but notice that the output of those instructions is
2558 src0_copy
.modifiers
= abs_modifier(src0_copy
.modifiers
, &need_abs
);
2559 instruction
= "LG2";
2561 default: instruction
= "";
2562 FIXME("Unhandled opcode %s.\n", debug_d3dshaderinstructionhandler(ins
->handler_idx
));
2566 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
2568 src0_copy
.swizzle
= shader_arb_select_component(src0_copy
.swizzle
, 3);
2570 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
); /* Destination */
2571 shader_arb_get_src_param(ins
, &src0_copy
, 0, src
);
2575 shader_addline(buffer
, "ABS TA.w, %s;\n", src
);
2576 shader_addline(buffer
, "%s%s %s, TA.w;\n", instruction
, shader_arb_get_modifier(ins
), dst
);
2580 shader_addline(buffer
, "%s%s %s, %s;\n", instruction
, shader_arb_get_modifier(ins
), dst
, src
);
2585 static void shader_hw_nrm(const struct wined3d_shader_instruction
*ins
)
2587 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2590 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2591 BOOL pshader
= shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2592 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2594 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2595 shader_arb_get_src_param(ins
, &ins
->src
[0], 1 /* Use TB */, src_name
);
2597 /* In D3D, NRM of a vector with length zero returns zero. Catch this situation, as
2598 * otherwise NRM or RSQ would return NaN */
2599 if(pshader
&& priv
->target_version
>= NV3
)
2601 /* GL_NV_fragment_program2's NRM needs protection against length zero vectors too
2603 * TODO: Find out if DP3+NRM+MOV is really faster than DP3+RSQ+MUL
2605 shader_addline(buffer
, "DP3C TA, %s, %s;\n", src_name
, src_name
);
2606 shader_addline(buffer
, "NRM%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
);
2607 shader_addline(buffer
, "MOV %s (EQ), %s;\n", dst_name
, zero
);
2609 else if(priv
->target_version
>= NV2
)
2611 shader_addline(buffer
, "DP3C TA.x, %s, %s;\n", src_name
, src_name
);
2612 shader_addline(buffer
, "RSQ TA.x (NE), TA.x;\n");
2613 shader_addline(buffer
, "MUL%s %s, %s, TA.x;\n", shader_arb_get_modifier(ins
), dst_name
,
2618 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2620 shader_addline(buffer
, "DP3 TA.x, %s, %s;\n", src_name
, src_name
);
2621 /* Pass any non-zero value to RSQ if the input vector has a length of zero. The
2622 * RSQ result doesn't matter, as long as multiplying it by 0 returns 0.
2624 shader_addline(buffer
, "SGE TA.y, -TA.x, %s;\n", zero
);
2625 shader_addline(buffer
, "MAD TA.x, %s, TA.y, TA.x;\n", one
);
2627 shader_addline(buffer
, "RSQ TA.x, TA.x;\n");
2628 /* dst.w = src[0].w * 1 / (src.x^2 + src.y^2 + src.z^2)^(1/2) according to msdn*/
2629 shader_addline(buffer
, "MUL%s %s, %s, TA.x;\n", shader_arb_get_modifier(ins
), dst_name
,
2634 static void shader_hw_lrp(const struct wined3d_shader_instruction
*ins
)
2636 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2638 char src_name
[3][50];
2640 /* ARB_fragment_program has a convenient LRP instruction */
2641 if(shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)) {
2642 shader_hw_map2gl(ins
);
2646 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2647 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
2648 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
2649 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
2651 shader_addline(buffer
, "SUB TA, %s, %s;\n", src_name
[1], src_name
[2]);
2652 shader_addline(buffer
, "MAD%s %s, %s, TA, %s;\n", shader_arb_get_modifier(ins
),
2653 dst_name
, src_name
[0], src_name
[2]);
2656 static void shader_hw_sincos(const struct wined3d_shader_instruction
*ins
)
2658 /* This instruction exists in ARB, but the d3d instruction takes two extra parameters which
2659 * must contain fixed constants. So we need a separate function to filter those constants and
2662 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2663 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2664 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2666 char src_name0
[50], src_name1
[50], src_name2
[50];
2669 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
2670 if(shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)) {
2671 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2672 /* No modifiers are supported on SCS */
2673 shader_addline(buffer
, "SCS %s, %s;\n", dst_name
, src_name0
);
2675 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
)
2677 shader_arb_get_register_name(ins
, &dst
->reg
, src_name0
, &is_color
);
2678 shader_addline(buffer
, "MOV_SAT %s, %s;\n", dst_name
, src_name0
);
2680 } else if(priv
->target_version
>= NV2
) {
2681 shader_arb_get_register_name(ins
, &dst
->reg
, dst_name
, &is_color
);
2683 /* Sincos writemask must be .x, .y or .xy */
2684 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
)
2685 shader_addline(buffer
, "COS%s %s.x, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name0
);
2686 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
)
2687 shader_addline(buffer
, "SIN%s %s.y, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name0
);
2689 /* Approximate sine and cosine with a taylor series, as per math textbook. The application passes 8
2690 * helper constants(D3DSINCOSCONST1 and D3DSINCOSCONST2) in src1 and src2.
2692 * sin(x) = x - x^3/3! + x^5/5! - x^7/7! + ...
2693 * cos(x) = 1 - x^2/2! + x^4/4! - x^6/6! + ...
2695 * The constants we get are:
2697 * +1 +1, -1 -1 +1 +1 -1 -1
2698 * ---- , ---- , ---- , ----- , ----- , ----- , ------
2699 * 1!*2 2!*4 3!*8 4!*16 5!*32 6!*64 7!*128
2701 * If used with x^2, x^3, x^4 etc they calculate sin(x/2) and cos(x/2):
2705 * (x/2)^4 = x^4 / 16
2706 * (x/2)^5 = x^5 / 32
2709 * To get the final result:
2710 * sin(x) = 2 * sin(x/2) * cos(x/2)
2711 * cos(x) = cos(x/2)^2 - sin(x/2)^2
2712 * (from sin(x+y) and cos(x+y) rules)
2714 * As per MSDN, dst.z is undefined after the operation, and so is
2715 * dst.x and dst.y if they're masked out by the writemask. Ie
2716 * sincos dst.y, src1, c0, c1
2717 * returns the sine in dst.y. dst.x and dst.z are undefined, dst.w is not touched. The assembler
2718 * vsa.exe also stops with an error if the dest register is the same register as the source
2719 * register. This means we can use dest.xyz as temporary storage. The assembler vsa.exe output also
2720 * indicates that sincos consumes 8 instruction slots in vs_2_0(and, strangely, in vs_3_0).
2722 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
2723 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name2
);
2724 shader_arb_get_register_name(ins
, &dst
->reg
, dst_name
, &is_color
);
2726 shader_addline(buffer
, "MUL %s.x, %s, %s;\n", dst_name
, src_name0
, src_name0
); /* x ^ 2 */
2727 shader_addline(buffer
, "MUL TA.y, %s.x, %s;\n", dst_name
, src_name0
); /* x ^ 3 */
2728 shader_addline(buffer
, "MUL %s.y, TA.y, %s;\n", dst_name
, src_name0
); /* x ^ 4 */
2729 shader_addline(buffer
, "MUL TA.z, %s.y, %s;\n", dst_name
, src_name0
); /* x ^ 5 */
2730 shader_addline(buffer
, "MUL %s.z, TA.z, %s;\n", dst_name
, src_name0
); /* x ^ 6 */
2731 shader_addline(buffer
, "MUL TA.w, %s.z, %s;\n", dst_name
, src_name0
); /* x ^ 7 */
2735 * Unfortunately we don't get the constants in a DP4-capable form. Is there a way to
2736 * properly merge that with MULs in the code above?
2737 * The swizzles .yz and xw however fit into the .yzxw swizzle added to ps_2_0. Maybe
2738 * we can merge the sine and cosine MAD rows to calculate them together.
2740 shader_addline(buffer
, "MUL TA.x, %s, %s.w;\n", src_name0
, src_name2
); /* x^1, +1/(1!*2) */
2741 shader_addline(buffer
, "MAD TA.x, TA.y, %s.x, TA.x;\n", src_name2
); /* -1/(3!*8) */
2742 shader_addline(buffer
, "MAD TA.x, TA.z, %s.w, TA.x;\n", src_name1
); /* +1/(5!*32) */
2743 shader_addline(buffer
, "MAD TA.x, TA.w, %s.x, TA.x;\n", src_name1
); /* -1/(7!*128) */
2746 shader_addline(buffer
, "MAD TA.y, %s.x, %s.y, %s.z;\n", dst_name
, src_name2
, src_name2
); /* -1/(2!*4), +1.0 */
2747 shader_addline(buffer
, "MAD TA.y, %s.y, %s.z, TA.y;\n", dst_name
, src_name1
); /* +1/(4!*16) */
2748 shader_addline(buffer
, "MAD TA.y, %s.z, %s.y, TA.y;\n", dst_name
, src_name1
); /* -1/(6!*64) */
2750 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
) {
2752 shader_addline(buffer
, "MUL TA.z, TA.y, TA.y;\n");
2753 shader_addline(buffer
, "MAD %s.x, -TA.x, TA.x, TA.z;\n", dst_name
);
2755 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
) {
2757 shader_addline(buffer
, "MUL %s.y, TA.x, TA.y;\n", dst_name
);
2758 shader_addline(buffer
, "ADD %s.y, %s.y, %s.y;\n", dst_name
, dst_name
, dst_name
);
2763 static void shader_hw_sgn(const struct wined3d_shader_instruction
*ins
)
2765 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2768 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
2770 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2771 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
);
2773 /* SGN is only valid in vertex shaders */
2774 if(ctx
->target_version
>= NV2
) {
2775 shader_addline(buffer
, "SSG%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
);
2779 /* If SRC > 0.0, -SRC < SRC = TRUE, otherwise false.
2780 * if SRC < 0.0, SRC < -SRC = TRUE. If neither is true, src = 0.0
2782 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
) {
2783 shader_addline(buffer
, "SLT %s, -%s, %s;\n", dst_name
, src_name
, src_name
);
2785 /* src contains TA? Write to the dest first. This won't overwrite our destination.
2786 * Then use TA, and calculate the final result
2788 * Not reading from TA? Store the first result in TA to avoid overwriting the
2789 * destination if src reg = dst reg
2791 if(strstr(src_name
, "TA"))
2793 shader_addline(buffer
, "SLT %s, %s, -%s;\n", dst_name
, src_name
, src_name
);
2794 shader_addline(buffer
, "SLT TA, -%s, %s;\n", src_name
, src_name
);
2795 shader_addline(buffer
, "ADD %s, %s, -TA;\n", dst_name
, dst_name
);
2799 shader_addline(buffer
, "SLT TA, -%s, %s;\n", src_name
, src_name
);
2800 shader_addline(buffer
, "SLT %s, %s, -%s;\n", dst_name
, src_name
, src_name
);
2801 shader_addline(buffer
, "ADD %s, TA, -%s;\n", dst_name
, dst_name
);
2806 static void shader_hw_dsy(const struct wined3d_shader_instruction
*ins
)
2808 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2814 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
);
2815 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src
);
2816 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2818 shader_addline(buffer
, "DDY %s, %s;\n", dst
, src
);
2819 shader_addline(buffer
, "MUL%s %s, %s, ycorrection.y;\n", shader_arb_get_modifier(ins
), dst
, dst_name
);
2822 static void shader_hw_pow(const struct wined3d_shader_instruction
*ins
)
2824 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2825 char src0
[50], src1
[50], dst
[50];
2826 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
2827 BOOL need_abs
= FALSE
;
2828 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2829 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2831 /* POW operates on the absolute value of the input */
2832 src0_copy
.modifiers
= abs_modifier(src0_copy
.modifiers
, &need_abs
);
2834 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
);
2835 shader_arb_get_src_param(ins
, &src0_copy
, 0, src0
);
2836 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src1
);
2839 shader_addline(buffer
, "ABS TA.x, %s;\n", src0
);
2841 shader_addline(buffer
, "MOV TA.x, %s;\n", src0
);
2843 if (priv
->target_version
>= NV2
)
2845 shader_addline(buffer
, "MOVC TA.y, %s;\n", src1
);
2846 shader_addline(buffer
, "POW%s %s, TA.x, TA.y;\n", shader_arb_get_modifier(ins
), dst
);
2847 shader_addline(buffer
, "MOV %s (EQ.y), %s;\n", dst
, one
);
2851 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2852 const char *flt_eps
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_EPS
);
2854 shader_addline(buffer
, "ABS TA.y, %s;\n", src1
);
2855 shader_addline(buffer
, "SGE TA.y, -TA.y, %s;\n", zero
);
2856 /* Possibly add flt_eps to avoid getting float special values */
2857 shader_addline(buffer
, "MAD TA.z, TA.y, %s, %s;\n", flt_eps
, src1
);
2858 shader_addline(buffer
, "POW%s TA.x, TA.x, TA.z;\n", shader_arb_get_modifier(ins
));
2859 shader_addline(buffer
, "MAD TA.x, -TA.x, TA.y, TA.x;\n");
2860 shader_addline(buffer
, "MAD %s, TA.y, %s, TA.x;\n", dst
, one
);
2864 static void shader_hw_loop(const struct wined3d_shader_instruction
*ins
)
2866 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2868 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2871 shader_arb_get_src_param(ins
, &ins
->src
[1], 0, src_name
);
2875 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2876 struct list
*e
= list_head(&priv
->control_frames
);
2877 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2879 if(priv
->loop_depth
> 1) shader_addline(buffer
, "PUSHA aL;\n");
2880 /* The constant loader makes sure to load -1 into iX.w */
2881 shader_addline(buffer
, "ARLC aL, %s.xywz;\n", src_name
);
2882 shader_addline(buffer
, "BRA loop_%u_end (LE.x);\n", control_frame
->no
.loop
);
2883 shader_addline(buffer
, "loop_%u_start:\n", control_frame
->no
.loop
);
2887 shader_addline(buffer
, "LOOP %s;\n", src_name
);
2891 static void shader_hw_rep(const struct wined3d_shader_instruction
*ins
)
2893 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2895 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2897 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
);
2899 /* The constant loader makes sure to load -1 into iX.w */
2902 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2903 struct list
*e
= list_head(&priv
->control_frames
);
2904 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2906 if(priv
->loop_depth
> 1) shader_addline(buffer
, "PUSHA aL;\n");
2908 shader_addline(buffer
, "ARLC aL, %s.xywz;\n", src_name
);
2909 shader_addline(buffer
, "BRA loop_%u_end (LE.x);\n", control_frame
->no
.loop
);
2910 shader_addline(buffer
, "loop_%u_start:\n", control_frame
->no
.loop
);
2914 shader_addline(buffer
, "REP %s;\n", src_name
);
2918 static void shader_hw_endloop(const struct wined3d_shader_instruction
*ins
)
2920 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2921 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2925 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2926 struct list
*e
= list_head(&priv
->control_frames
);
2927 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2929 shader_addline(buffer
, "ARAC aL.xy, aL;\n");
2930 shader_addline(buffer
, "BRA loop_%u_start (GT.x);\n", control_frame
->no
.loop
);
2931 shader_addline(buffer
, "loop_%u_end:\n", control_frame
->no
.loop
);
2933 if(priv
->loop_depth
> 1) shader_addline(buffer
, "POPA aL;\n");
2937 shader_addline(buffer
, "ENDLOOP;\n");
2941 static void shader_hw_endrep(const struct wined3d_shader_instruction
*ins
)
2943 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2944 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2948 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2949 struct list
*e
= list_head(&priv
->control_frames
);
2950 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2952 shader_addline(buffer
, "ARAC aL.xy, aL;\n");
2953 shader_addline(buffer
, "BRA loop_%u_start (GT.x);\n", control_frame
->no
.loop
);
2954 shader_addline(buffer
, "loop_%u_end:\n", control_frame
->no
.loop
);
2956 if(priv
->loop_depth
> 1) shader_addline(buffer
, "POPA aL;\n");
2960 shader_addline(buffer
, "ENDREP;\n");
2964 static const struct control_frame
*find_last_loop(const struct shader_arb_ctx_priv
*priv
)
2966 struct control_frame
*control_frame
;
2968 LIST_FOR_EACH_ENTRY(control_frame
, &priv
->control_frames
, struct control_frame
, entry
)
2970 if(control_frame
->type
== LOOP
|| control_frame
->type
== REP
) return control_frame
;
2972 ERR("Could not find loop for break\n");
2976 static void shader_hw_break(const struct wined3d_shader_instruction
*ins
)
2978 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2979 const struct control_frame
*control_frame
= find_last_loop(ins
->ctx
->backend_data
);
2980 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2984 shader_addline(buffer
, "BRA loop_%u_end;\n", control_frame
->no
.loop
);
2988 shader_addline(buffer
, "BRK;\n");
2992 static const char *get_compare(enum wined3d_shader_rel_op op
)
2996 case WINED3D_SHADER_REL_OP_GT
: return "GT";
2997 case WINED3D_SHADER_REL_OP_EQ
: return "EQ";
2998 case WINED3D_SHADER_REL_OP_GE
: return "GE";
2999 case WINED3D_SHADER_REL_OP_LT
: return "LT";
3000 case WINED3D_SHADER_REL_OP_NE
: return "NE";
3001 case WINED3D_SHADER_REL_OP_LE
: return "LE";
3003 FIXME("Unrecognized operator %#x.\n", op
);
3008 static enum wined3d_shader_rel_op
invert_compare(enum wined3d_shader_rel_op op
)
3012 case WINED3D_SHADER_REL_OP_GT
: return WINED3D_SHADER_REL_OP_LE
;
3013 case WINED3D_SHADER_REL_OP_EQ
: return WINED3D_SHADER_REL_OP_NE
;
3014 case WINED3D_SHADER_REL_OP_GE
: return WINED3D_SHADER_REL_OP_LT
;
3015 case WINED3D_SHADER_REL_OP_LT
: return WINED3D_SHADER_REL_OP_GE
;
3016 case WINED3D_SHADER_REL_OP_NE
: return WINED3D_SHADER_REL_OP_EQ
;
3017 case WINED3D_SHADER_REL_OP_LE
: return WINED3D_SHADER_REL_OP_GT
;
3019 FIXME("Unrecognized operator %#x.\n", op
);
3024 static void shader_hw_breakc(const struct wined3d_shader_instruction
*ins
)
3026 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3027 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3028 const struct control_frame
*control_frame
= find_last_loop(ins
->ctx
->backend_data
);
3031 const char *comp
= get_compare(ins
->flags
);
3033 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
3034 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
3038 /* SUBC CC, src0, src1" works only in pixel shaders, so use TA to throw
3039 * away the subtraction result
3041 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3042 shader_addline(buffer
, "BRA loop_%u_end (%s.x);\n", control_frame
->no
.loop
, comp
);
3046 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3047 shader_addline(buffer
, "BRK (%s.x);\n", comp
);
3051 static void shader_hw_ifc(const struct wined3d_shader_instruction
*ins
)
3053 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3054 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3055 struct list
*e
= list_head(&priv
->control_frames
);
3056 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3060 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3062 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
3063 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
3067 /* Invert the flag. We jump to the else label if the condition is NOT true */
3068 comp
= get_compare(invert_compare(ins
->flags
));
3069 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3070 shader_addline(buffer
, "BRA ifc_%u_else (%s.x);\n", control_frame
->no
.ifc
, comp
);
3074 comp
= get_compare(ins
->flags
);
3075 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3076 shader_addline(buffer
, "IF %s.x;\n", comp
);
3080 static void shader_hw_else(const struct wined3d_shader_instruction
*ins
)
3082 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3083 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3084 struct list
*e
= list_head(&priv
->control_frames
);
3085 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3086 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3090 shader_addline(buffer
, "BRA ifc_%u_endif;\n", control_frame
->no
.ifc
);
3091 shader_addline(buffer
, "ifc_%u_else:\n", control_frame
->no
.ifc
);
3092 control_frame
->had_else
= TRUE
;
3096 shader_addline(buffer
, "ELSE;\n");
3100 static void shader_hw_endif(const struct wined3d_shader_instruction
*ins
)
3102 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3103 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3104 struct list
*e
= list_head(&priv
->control_frames
);
3105 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3106 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3110 if(control_frame
->had_else
)
3112 shader_addline(buffer
, "ifc_%u_endif:\n", control_frame
->no
.ifc
);
3116 shader_addline(buffer
, "#No else branch. else is endif\n");
3117 shader_addline(buffer
, "ifc_%u_else:\n", control_frame
->no
.ifc
);
3122 shader_addline(buffer
, "ENDIF;\n");
3126 static void shader_hw_texldd(const struct wined3d_shader_instruction
*ins
)
3128 DWORD sampler_idx
= ins
->src
[1].reg
.idx
[0].offset
;
3130 char reg_src
[3][40];
3131 WORD flags
= TEX_DERIV
;
3133 shader_arb_get_dst_param(ins
, &ins
->dst
[0], reg_dest
);
3134 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_src
[0]);
3135 shader_arb_get_src_param(ins
, &ins
->src
[2], 1, reg_src
[1]);
3136 shader_arb_get_src_param(ins
, &ins
->src
[3], 2, reg_src
[2]);
3138 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) flags
|= TEX_PROJ
;
3139 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) flags
|= TEX_BIAS
;
3141 shader_hw_sample(ins
, sampler_idx
, reg_dest
, reg_src
[0], flags
, reg_src
[1], reg_src
[2]);
3144 static void shader_hw_texldl(const struct wined3d_shader_instruction
*ins
)
3146 DWORD sampler_idx
= ins
->src
[1].reg
.idx
[0].offset
;
3149 WORD flags
= TEX_LOD
;
3151 shader_arb_get_dst_param(ins
, &ins
->dst
[0], reg_dest
);
3152 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_coord
);
3154 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) flags
|= TEX_PROJ
;
3155 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) flags
|= TEX_BIAS
;
3157 shader_hw_sample(ins
, sampler_idx
, reg_dest
, reg_coord
, flags
, NULL
, NULL
);
3160 static void shader_hw_label(const struct wined3d_shader_instruction
*ins
)
3162 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3163 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3165 priv
->in_main_func
= FALSE
;
3166 /* Call instructions activate the NV extensions, not labels and rets. If there is an uncalled
3167 * subroutine, don't generate a label that will make GL complain
3169 if(priv
->target_version
== ARB
) return;
3171 shader_addline(buffer
, "l%u:\n", ins
->src
[0].reg
.idx
[0].offset
);
3174 static void vshader_add_footer(struct shader_arb_ctx_priv
*priv_ctx
,
3175 const struct arb_vshader_private
*shader_data
, const struct arb_vs_compile_args
*args
,
3176 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
,
3177 struct wined3d_string_buffer
*buffer
)
3181 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
3182 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
3183 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
3184 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
3186 if (args
->super
.fog_src
== VS_FOG_Z
)
3188 shader_addline(buffer
, "MOV result.fogcoord, TMP_OUT.z;\n");
3194 /* posFixup.x is always 1.0, so we can safely use it */
3195 shader_addline(buffer
, "ADD result.fogcoord, posFixup.x, -posFixup.x;\n");
3199 /* Clamp fogcoord */
3200 const char *zero
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ZERO
);
3201 const char *one
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ONE
);
3203 shader_addline(buffer
, "MIN TMP_FOGCOORD.x, TMP_FOGCOORD.x, %s;\n", one
);
3204 shader_addline(buffer
, "MAX result.fogcoord.x, TMP_FOGCOORD.x, %s;\n", zero
);
3208 /* Clipplanes are always stored without y inversion */
3209 if (use_nv_clip(gl_info
) && priv_ctx
->target_version
>= NV2
)
3211 if (args
->super
.clip_enabled
)
3213 for (i
= 0; i
< priv_ctx
->vs_clipplanes
; i
++)
3215 shader_addline(buffer
, "DP4 result.clip[%u].x, TMP_OUT, state.clip[%u].plane;\n", i
, i
);
3219 else if (args
->clip
.boolclip
.clip_texcoord
)
3221 static const char component
[4] = {'x', 'y', 'z', 'w'};
3222 unsigned int cur_clip
= 0;
3223 const char *zero
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_ZERO
);
3225 for (i
= 0; i
< gl_info
->limits
.clipplanes
; ++i
)
3227 if (args
->clip
.boolclip
.clipplane_mask
& (1u << i
))
3229 shader_addline(buffer
, "DP4 TA.%c, TMP_OUT, state.clip[%u].plane;\n",
3230 component
[cur_clip
++], i
);
3236 shader_addline(buffer
, "MOV TA, %s;\n", zero
);
3239 shader_addline(buffer
, "MOV TA.yzw, %s;\n", zero
);
3242 shader_addline(buffer
, "MOV TA.zw, %s;\n", zero
);
3245 shader_addline(buffer
, "MOV TA.w, %s;\n", zero
);
3248 shader_addline(buffer
, "MOV result.texcoord[%u], TA;\n",
3249 args
->clip
.boolclip
.clip_texcoord
- 1);
3252 /* Write the final position.
3254 * OpenGL coordinates specify the center of the pixel while d3d coords specify
3255 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
3256 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
3257 * contains 1.0 to allow a mad, but arb vs swizzles are too restricted for that.
3259 shader_addline(buffer
, "MUL TA, posFixup, TMP_OUT.w;\n");
3260 shader_addline(buffer
, "ADD TMP_OUT.x, TMP_OUT.x, TA.z;\n");
3261 shader_addline(buffer
, "MAD TMP_OUT.y, TMP_OUT.y, posFixup.y, TA.w;\n");
3263 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
3264 * and the glsl equivalent
3266 if (need_helper_const(shader_data
, reg_maps
, gl_info
))
3268 const char *two
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_TWO
);
3269 shader_addline(buffer
, "MAD TMP_OUT.z, TMP_OUT.z, %s, -TMP_OUT.w;\n", two
);
3273 shader_addline(buffer
, "ADD TMP_OUT.z, TMP_OUT.z, TMP_OUT.z;\n");
3274 shader_addline(buffer
, "ADD TMP_OUT.z, TMP_OUT.z, -TMP_OUT.w;\n");
3277 shader_addline(buffer
, "MOV result.position, TMP_OUT;\n");
3279 priv_ctx
->footer_written
= TRUE
;
3282 static void shader_hw_ret(const struct wined3d_shader_instruction
*ins
)
3284 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3285 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3286 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
3287 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3289 if(priv
->target_version
== ARB
) return;
3293 if (priv
->in_main_func
) vshader_add_footer(priv
, shader
->backend_data
,
3294 priv
->cur_vs_args
, ins
->ctx
->reg_maps
, ins
->ctx
->gl_info
, buffer
);
3297 shader_addline(buffer
, "RET;\n");
3300 static void shader_hw_call(const struct wined3d_shader_instruction
*ins
)
3302 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3303 shader_addline(buffer
, "CAL l%u;\n", ins
->src
[0].reg
.idx
[0].offset
);
3306 static BOOL
shader_arb_compile(const struct wined3d_gl_info
*gl_info
, GLenum target
, const char *src
)
3308 const char *ptr
, *line
;
3311 if (TRACE_ON(d3d_shader
))
3314 while ((line
= get_line(&ptr
))) TRACE_(d3d_shader
)(" %.*s", (int)(ptr
- line
), line
);
3317 GL_EXTCALL(glProgramStringARB(target
, GL_PROGRAM_FORMAT_ASCII_ARB
, strlen(src
), src
));
3318 checkGLcall("glProgramStringARB()");
3320 if (FIXME_ON(d3d_shader
))
3322 gl_info
->gl_ops
.gl
.p_glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
3325 FIXME_(d3d_shader
)("Program error at position %d: %s\n\n", pos
,
3326 debugstr_a((const char *)gl_info
->gl_ops
.gl
.p_glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
3328 while ((line
= get_line(&ptr
))) FIXME_(d3d_shader
)(" %.*s", (int)(ptr
- line
), line
);
3329 FIXME_(d3d_shader
)("\n");
3335 if (WARN_ON(d3d_perf
))
3337 GL_EXTCALL(glGetProgramivARB(target
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
3338 checkGLcall("glGetProgramivARB()");
3340 WARN_(d3d_perf
)("Program exceeds native resource limits.\n");
3346 /* Context activation is done by the caller. */
3347 static GLuint
create_arb_blt_vertex_program(const struct wined3d_gl_info
*gl_info
)
3349 GLuint program_id
= 0;
3351 static const char blt_vprogram
[] =
3353 "PARAM c[1] = { { 1, 0.5 } };\n"
3354 "MOV result.position, vertex.position;\n"
3355 "MOV result.color, c[0].x;\n"
3356 "MOV result.texcoord[0], vertex.texcoord[0];\n"
3359 GL_EXTCALL(glGenProgramsARB(1, &program_id
));
3360 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, program_id
));
3361 shader_arb_compile(gl_info
, GL_VERTEX_PROGRAM_ARB
, blt_vprogram
);
3366 /* Context activation is done by the caller. */
3367 static GLuint
create_arb_blt_fragment_program(const struct wined3d_gl_info
*gl_info
,
3368 enum wined3d_gl_resource_type tex_type
, BOOL masked
)
3370 GLuint program_id
= 0;
3371 const char *fprogram
;
3373 static const char * const blt_fprograms_full
[WINED3D_GL_RES_TYPE_COUNT
] =
3375 /* WINED3D_GL_RES_TYPE_TEX_1D */
3377 /* WINED3D_GL_RES_TYPE_TEX_2D */
3380 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
3381 "MOV result.depth.z, R0.x;\n"
3383 /* WINED3D_GL_RES_TYPE_TEX_3D */
3385 /* WINED3D_GL_RES_TYPE_TEX_CUBE */
3388 "TEX R0.x, fragment.texcoord[0], texture[0], CUBE;\n"
3389 "MOV result.depth.z, R0.x;\n"
3391 /* WINED3D_GL_RES_TYPE_TEX_RECT */
3394 "TEX R0.x, fragment.texcoord[0], texture[0], RECT;\n"
3395 "MOV result.depth.z, R0.x;\n"
3397 /* WINED3D_GL_RES_TYPE_BUFFER */
3399 /* WINED3D_GL_RES_TYPE_RB */
3403 static const char * const blt_fprograms_masked
[WINED3D_GL_RES_TYPE_COUNT
] =
3405 /* WINED3D_GL_RES_TYPE_TEX_1D */
3407 /* WINED3D_GL_RES_TYPE_TEX_2D */
3409 "PARAM mask = program.local[0];\n"
3411 "SLT R0.xy, fragment.position, mask.zwzw;\n"
3412 "MUL R0.x, R0.x, R0.y;\n"
3414 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
3415 "MOV result.depth.z, R0.x;\n"
3417 /* WINED3D_GL_RES_TYPE_TEX_3D */
3419 /* WINED3D_GL_RES_TYPE_TEX_CUBE */
3421 "PARAM mask = program.local[0];\n"
3423 "SLT R0.xy, fragment.position, mask.zwzw;\n"
3424 "MUL R0.x, R0.x, R0.y;\n"
3426 "TEX R0.x, fragment.texcoord[0], texture[0], CUBE;\n"
3427 "MOV result.depth.z, R0.x;\n"
3429 /* WINED3D_GL_RES_TYPE_TEX_RECT */
3431 "PARAM mask = program.local[0];\n"
3433 "SLT R0.xy, fragment.position, mask.zwzw;\n"
3434 "MUL R0.x, R0.x, R0.y;\n"
3436 "TEX R0.x, fragment.texcoord[0], texture[0], RECT;\n"
3437 "MOV result.depth.z, R0.x;\n"
3439 /* WINED3D_GL_RES_TYPE_BUFFER */
3441 /* WINED3D_GL_RES_TYPE_RB */
3445 fprogram
= masked
? blt_fprograms_masked
[tex_type
] : blt_fprograms_full
[tex_type
];
3448 FIXME("tex_type %#x not supported, falling back to 2D\n", tex_type
);
3449 tex_type
= WINED3D_GL_RES_TYPE_TEX_2D
;
3450 fprogram
= masked
? blt_fprograms_masked
[tex_type
] : blt_fprograms_full
[tex_type
];
3453 GL_EXTCALL(glGenProgramsARB(1, &program_id
));
3454 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, program_id
));
3455 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, fprogram
);
3460 static void arbfp_add_sRGB_correction(struct wined3d_string_buffer
*buffer
, const char *fragcolor
,
3461 const char *tmp1
, const char *tmp2
, const char *tmp3
, const char *tmp4
, BOOL condcode
)
3463 /* Perform sRGB write correction. See GLX_EXT_framebuffer_sRGB */
3467 /* Sigh. MOVC CC doesn't work, so use one of the temps as dummy dest */
3468 shader_addline(buffer
, "SUBC %s, %s.x, srgb_consts1.x;\n", tmp1
, fragcolor
);
3469 /* Calculate the > 0.0031308 case */
3470 shader_addline(buffer
, "POW %s.x (GE), %s.x, srgb_consts0.x;\n", fragcolor
, fragcolor
);
3471 shader_addline(buffer
, "POW %s.y (GE), %s.y, srgb_consts0.x;\n", fragcolor
, fragcolor
);
3472 shader_addline(buffer
, "POW %s.z (GE), %s.z, srgb_consts0.x;\n", fragcolor
, fragcolor
);
3473 shader_addline(buffer
, "MUL %s.xyz (GE), %s, srgb_consts0.y;\n", fragcolor
, fragcolor
);
3474 shader_addline(buffer
, "SUB %s.xyz (GE), %s, srgb_consts0.z;\n", fragcolor
, fragcolor
);
3475 /* Calculate the < case */
3476 shader_addline(buffer
, "MUL %s.xyz (LT), srgb_consts0.w, %s;\n", fragcolor
, fragcolor
);
3480 /* Calculate the > 0.0031308 case */
3481 shader_addline(buffer
, "POW %s.x, %s.x, srgb_consts0.x;\n", tmp1
, fragcolor
);
3482 shader_addline(buffer
, "POW %s.y, %s.y, srgb_consts0.x;\n", tmp1
, fragcolor
);
3483 shader_addline(buffer
, "POW %s.z, %s.z, srgb_consts0.x;\n", tmp1
, fragcolor
);
3484 shader_addline(buffer
, "MUL %s, %s, srgb_consts0.y;\n", tmp1
, tmp1
);
3485 shader_addline(buffer
, "SUB %s, %s, srgb_consts0.z;\n", tmp1
, tmp1
);
3486 /* Calculate the < case */
3487 shader_addline(buffer
, "MUL %s, srgb_consts0.w, %s;\n", tmp2
, fragcolor
);
3488 /* Get 1.0 / 0.0 masks for > 0.0031308 and < 0.0031308 */
3489 shader_addline(buffer
, "SLT %s, srgb_consts1.x, %s;\n", tmp3
, fragcolor
);
3490 shader_addline(buffer
, "SGE %s, srgb_consts1.x, %s;\n", tmp4
, fragcolor
);
3491 /* Store the components > 0.0031308 in the destination */
3492 shader_addline(buffer
, "MUL %s.xyz, %s, %s;\n", fragcolor
, tmp1
, tmp3
);
3493 /* Add the components that are < 0.0031308 */
3494 shader_addline(buffer
, "MAD %s.xyz, %s, %s, %s;\n", fragcolor
, tmp2
, tmp4
, fragcolor
);
3495 /* Move everything into result.color at once. Nvidia hardware cannot handle partial
3496 * result.color writes(.rgb first, then .a), or handle overwriting already written
3497 * components. The assembler uses a temporary register in this case, which is usually
3498 * not allocated from one of our registers that were used earlier.
3501 /* [0.0;1.0] clamping. Not needed, this is done implicitly */
3504 static const DWORD
*find_loop_control_values(const struct wined3d_shader
*shader
, DWORD idx
)
3506 const struct wined3d_shader_lconst
*constant
;
3508 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsI
, struct wined3d_shader_lconst
, entry
)
3510 if (constant
->idx
== idx
)
3512 return constant
->value
;
3518 static void init_ps_input(const struct wined3d_shader
*shader
,
3519 const struct arb_ps_compile_args
*args
, struct shader_arb_ctx_priv
*priv
)
3521 static const char * const texcoords
[8] =
3523 "fragment.texcoord[0]", "fragment.texcoord[1]", "fragment.texcoord[2]", "fragment.texcoord[3]",
3524 "fragment.texcoord[4]", "fragment.texcoord[5]", "fragment.texcoord[6]", "fragment.texcoord[7]"
3527 const struct wined3d_shader_signature_element
*input
;
3528 const char *semantic_name
;
3531 switch(args
->super
.vp_mode
)
3533 case pretransformed
:
3535 /* The pixelshader has to collect the varyings on its own. In any case properly load
3536 * color0 and color1. In the case of pretransformed vertices also load texcoords. Set
3537 * other attribs to 0.0.
3539 * For fixedfunction this behavior is correct, according to the tests. For pretransformed
3540 * we'd either need a replacement shader that can load other attribs like BINORMAL, or
3541 * load the texcoord attrib pointers to match the pixel shader signature
3543 for (i
= 0; i
< shader
->input_signature
.element_count
; ++i
)
3545 input
= &shader
->input_signature
.elements
[i
];
3546 if (!(semantic_name
= input
->semantic_name
))
3548 semantic_idx
= input
->semantic_idx
;
3550 if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_COLOR
))
3553 priv
->ps_input
[input
->register_idx
] = "fragment.color.primary";
3554 else if (semantic_idx
== 1)
3555 priv
->ps_input
[input
->register_idx
] = "fragment.color.secondary";
3557 priv
->ps_input
[input
->register_idx
] = "0.0";
3559 else if (args
->super
.vp_mode
== fixedfunction
)
3561 priv
->ps_input
[input
->register_idx
] = "0.0";
3563 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
3565 if (semantic_idx
< 8)
3566 priv
->ps_input
[input
->register_idx
] = texcoords
[semantic_idx
];
3568 priv
->ps_input
[input
->register_idx
] = "0.0";
3570 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_FOG
))
3573 priv
->ps_input
[input
->register_idx
] = "fragment.fogcoord";
3575 priv
->ps_input
[input
->register_idx
] = "0.0";
3579 priv
->ps_input
[input
->register_idx
] = "0.0";
3582 TRACE("v%u, semantic %s%u is %s\n", input
->register_idx
,
3583 semantic_name
, semantic_idx
, priv
->ps_input
[input
->register_idx
]);
3588 /* That one is easy. The vertex shaders provide v0-v7 in fragment.texcoord and v8 and v9 in
3591 for(i
= 0; i
< 8; i
++)
3593 priv
->ps_input
[i
] = texcoords
[i
];
3595 priv
->ps_input
[8] = "fragment.color.primary";
3596 priv
->ps_input
[9] = "fragment.color.secondary";
3601 static void arbfp_add_linear_fog(struct wined3d_string_buffer
*buffer
,
3602 const char *fragcolor
, const char *tmp
)
3604 shader_addline(buffer
, "SUB %s.x, state.fog.params.z, fragment.fogcoord.x;\n", tmp
);
3605 shader_addline(buffer
, "MUL_SAT %s.x, %s.x, state.fog.params.w;\n", tmp
, tmp
);
3606 shader_addline(buffer
, "LRP %s.rgb, %s.x, %s, state.fog.color;\n", fragcolor
, tmp
, fragcolor
);
3609 /* Context activation is done by the caller. */
3610 static GLuint
shader_arb_generate_pshader(const struct wined3d_shader
*shader
,
3611 const struct wined3d_gl_info
*gl_info
, struct wined3d_string_buffer
*buffer
,
3612 const struct arb_ps_compile_args
*args
, struct arb_ps_compiled_shader
*compiled
)
3614 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
3615 const DWORD
*function
= shader
->function
;
3618 DWORD next_local
= 0;
3619 struct shader_arb_ctx_priv priv_ctx
;
3620 BOOL dcl_td
= FALSE
;
3621 BOOL want_nv_prog
= FALSE
;
3622 struct arb_pshader_private
*shader_priv
= shader
->backend_data
;
3624 BOOL custom_linear_fog
= FALSE
;
3628 unsigned int i
, found
= 0;
3630 for (i
= 0, map
= reg_maps
->temporary
; map
; map
>>= 1, ++i
)
3633 || (shader
->u
.ps
.color0_mov
&& i
== shader
->u
.ps
.color0_reg
)
3634 || (reg_maps
->shader_version
.major
< 2 && !i
))
3637 sprintf(srgbtmp
[found
], "R%u", i
);
3639 if (found
== 4) break;
3644 sprintf(srgbtmp
[0], "TA");
3645 sprintf(srgbtmp
[1], "TB");
3646 sprintf(srgbtmp
[2], "TC");
3647 sprintf(srgbtmp
[3], "TD");
3651 sprintf(srgbtmp
[1], "TA");
3652 sprintf(srgbtmp
[2], "TB");
3653 sprintf(srgbtmp
[3], "TC");
3656 sprintf(srgbtmp
[2], "TA");
3657 sprintf(srgbtmp
[3], "TB");
3660 sprintf(srgbtmp
[3], "TA");
3666 /* Create the hw ARB shader */
3667 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
3668 priv_ctx
.cur_ps_args
= args
;
3669 priv_ctx
.compiled_fprog
= compiled
;
3670 priv_ctx
.cur_np2fixup_info
= &compiled
->np2fixup_info
;
3671 init_ps_input(shader
, args
, &priv_ctx
);
3672 list_init(&priv_ctx
.control_frames
);
3673 priv_ctx
.ps_post_process
= args
->super
.srgb_correction
;
3675 /* Avoid enabling NV_fragment_program* if we do not need it.
3677 * Enabling GL_NV_fragment_program_option causes the driver to occupy a temporary register,
3678 * and it slows down the shader execution noticeably(about 5%). Usually our instruction emulation
3679 * is faster than what we gain from using higher native instructions. There are some things though
3680 * that cannot be emulated. In that case enable the extensions.
3681 * If the extension is enabled, instruction handlers that support both ways will use it.
3683 * Testing shows no performance difference between OPTION NV_fragment_program2 and NV_fragment_program.
3684 * So enable the best we can get.
3686 if(reg_maps
->usesdsx
|| reg_maps
->usesdsy
|| reg_maps
->loop_depth
> 0 || reg_maps
->usestexldd
||
3687 reg_maps
->usestexldl
|| reg_maps
->usesfacing
|| reg_maps
->usesifc
|| reg_maps
->usescall
)
3689 want_nv_prog
= TRUE
;
3692 shader_addline(buffer
, "!!ARBfp1.0\n");
3693 if (want_nv_prog
&& gl_info
->supported
[NV_FRAGMENT_PROGRAM2
])
3695 shader_addline(buffer
, "OPTION NV_fragment_program2;\n");
3696 priv_ctx
.target_version
= NV3
;
3698 else if (want_nv_prog
&& gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
])
3700 shader_addline(buffer
, "OPTION NV_fragment_program;\n");
3701 priv_ctx
.target_version
= NV2
;
3705 /* This is an error - either we're advertising the wrong shader version, or aren't enforcing some
3708 ERR("The shader requires instructions that are not available in plain GL_ARB_fragment_program\n");
3711 priv_ctx
.target_version
= ARB
;
3714 if (reg_maps
->rt_mask
> 1)
3716 shader_addline(buffer
, "OPTION ARB_draw_buffers;\n");
3719 if (reg_maps
->shader_version
.major
< 3)
3721 switch (args
->super
.fog
)
3723 case WINED3D_FFP_PS_FOG_OFF
:
3725 case WINED3D_FFP_PS_FOG_LINEAR
:
3726 if (gl_info
->quirks
& WINED3D_QUIRK_BROKEN_ARB_FOG
)
3728 custom_linear_fog
= TRUE
;
3729 priv_ctx
.ps_post_process
= TRUE
;
3732 shader_addline(buffer
, "OPTION ARB_fog_linear;\n");
3734 case WINED3D_FFP_PS_FOG_EXP
:
3735 shader_addline(buffer
, "OPTION ARB_fog_exp;\n");
3737 case WINED3D_FFP_PS_FOG_EXP2
:
3738 shader_addline(buffer
, "OPTION ARB_fog_exp2;\n");
3743 /* For now always declare the temps. At least the Nvidia assembler optimizes completely
3744 * unused temps away(but occupies them for the whole shader if they're used once). Always
3745 * declaring them avoids tricky bookkeeping work
3747 shader_addline(buffer
, "TEMP TA;\n"); /* Used for modifiers */
3748 shader_addline(buffer
, "TEMP TB;\n"); /* Used for modifiers */
3749 shader_addline(buffer
, "TEMP TC;\n"); /* Used for modifiers */
3750 if(dcl_td
) shader_addline(buffer
, "TEMP TD;\n"); /* Used for sRGB writing */
3751 shader_addline(buffer
, "PARAM coefdiv = { 0.5, 0.25, 0.125, 0.0625 };\n");
3752 shader_addline(buffer
, "PARAM coefmul = { 2, 4, 8, 16 };\n");
3753 wined3d_ftoa(eps
, ftoa_tmp
);
3754 shader_addline(buffer
, "PARAM ps_helper_const = { 0.0, 1.0, %s, 0.0 };\n", ftoa_tmp
);
3756 if (reg_maps
->shader_version
.major
< 2)
3758 strcpy(fragcolor
, "R0");
3762 if (priv_ctx
.ps_post_process
)
3764 if (shader
->u
.ps
.color0_mov
)
3766 sprintf(fragcolor
, "R%u", shader
->u
.ps
.color0_reg
);
3770 shader_addline(buffer
, "TEMP TMP_COLOR;\n");
3771 strcpy(fragcolor
, "TMP_COLOR");
3774 strcpy(fragcolor
, "result.color");
3778 if (args
->super
.srgb_correction
)
3780 shader_addline(buffer
, "PARAM srgb_consts0 = ");
3781 shader_arb_append_imm_vec4(buffer
, wined3d_srgb_const0
);
3782 shader_addline(buffer
, ";\n");
3783 shader_addline(buffer
, "PARAM srgb_consts1 = ");
3784 shader_arb_append_imm_vec4(buffer
, wined3d_srgb_const1
);
3785 shader_addline(buffer
, ";\n");
3788 /* Base Declarations */
3789 shader_generate_arb_declarations(shader
, reg_maps
, buffer
, gl_info
, NULL
, &priv_ctx
);
3791 for (i
= 0, map
= reg_maps
->bumpmat
; map
; map
>>= 1, ++i
)
3793 unsigned char bump_const
;
3795 if (!(map
& 1)) continue;
3797 bump_const
= compiled
->numbumpenvmatconsts
;
3798 compiled
->bumpenvmatconst
[bump_const
].const_num
= WINED3D_CONST_NUM_UNUSED
;
3799 compiled
->bumpenvmatconst
[bump_const
].texunit
= i
;
3800 compiled
->luminanceconst
[bump_const
].const_num
= WINED3D_CONST_NUM_UNUSED
;
3801 compiled
->luminanceconst
[bump_const
].texunit
= i
;
3803 /* We can fit the constants into the constant limit for sure because texbem, texbeml, bem and beml are only supported
3804 * in 1.x shaders, and GL_ARB_fragment_program has a constant limit of 24 constants. So in the worst case we're loading
3805 * 8 shader constants, 8 bump matrices and 8 luminance parameters and are perfectly fine. (No NP2 fixup on bumpmapped
3806 * textures due to conditional NP2 restrictions)
3808 * Use local constants to load the bump env parameters, not program.env. This avoids collisions with d3d constants of
3809 * shaders in newer shader models. Since the bump env parameters have to share their space with NP2 fixup constants,
3810 * their location is shader dependent anyway and they cannot be loaded globally.
3812 compiled
->bumpenvmatconst
[bump_const
].const_num
= next_local
++;
3813 shader_addline(buffer
, "PARAM bumpenvmat%d = program.local[%d];\n",
3814 i
, compiled
->bumpenvmatconst
[bump_const
].const_num
);
3815 compiled
->numbumpenvmatconsts
= bump_const
+ 1;
3817 if (!(reg_maps
->luminanceparams
& (1u << i
)))
3820 compiled
->luminanceconst
[bump_const
].const_num
= next_local
++;
3821 shader_addline(buffer
, "PARAM luminance%d = program.local[%d];\n",
3822 i
, compiled
->luminanceconst
[bump_const
].const_num
);
3825 for(i
= 0; i
< MAX_CONST_I
; i
++)
3827 compiled
->int_consts
[i
] = WINED3D_CONST_NUM_UNUSED
;
3828 if (reg_maps
->integer_constants
& (1u << i
) && priv_ctx
.target_version
>= NV2
)
3830 const DWORD
*control_values
= find_loop_control_values(shader
, i
);
3834 shader_addline(buffer
, "PARAM I%u = {%u, %u, %u, -1};\n", i
,
3835 control_values
[0], control_values
[1], control_values
[2]);
3839 compiled
->int_consts
[i
] = next_local
;
3840 compiled
->num_int_consts
++;
3841 shader_addline(buffer
, "PARAM I%u = program.local[%u];\n", i
, next_local
++);
3846 if(reg_maps
->vpos
|| reg_maps
->usesdsy
)
3848 compiled
->ycorrection
= next_local
;
3849 shader_addline(buffer
, "PARAM ycorrection = program.local[%u];\n", next_local
++);
3853 shader_addline(buffer
, "TEMP vpos;\n");
3854 /* ycorrection.x: Backbuffer height(onscreen) or 0(offscreen).
3855 * ycorrection.y: -1.0(onscreen), 1.0(offscreen)
3856 * ycorrection.z: 1.0
3857 * ycorrection.w: 0.0
3859 shader_addline(buffer
, "MAD vpos, fragment.position, ycorrection.zyww, ycorrection.wxww;\n");
3860 shader_addline(buffer
, "FLR vpos.xy, vpos;\n");
3865 compiled
->ycorrection
= WINED3D_CONST_NUM_UNUSED
;
3868 /* Load constants to fixup NP2 texcoords if there are still free constants left:
3869 * Constants (texture dimensions) for the NP2 fixup are loaded as local program parameters. This will consume
3870 * at most 8 (MAX_FRAGMENT_SAMPLERS / 2) parameters, which is highly unlikely, since the application had to
3871 * use 16 NP2 textures at the same time. In case that we run out of constants the fixup is simply not
3872 * applied / activated. This will probably result in wrong rendering of the texture, but will save us from
3873 * shader compilation errors and the subsequent errors when drawing with this shader. */
3874 if (priv_ctx
.cur_ps_args
->super
.np2_fixup
) {
3875 unsigned char cur_fixup_sampler
= 0;
3877 struct arb_ps_np2fixup_info
* const fixup
= priv_ctx
.cur_np2fixup_info
;
3878 const WORD map
= priv_ctx
.cur_ps_args
->super
.np2_fixup
;
3879 const UINT max_lconsts
= gl_info
->limits
.arb_ps_local_constants
;
3881 fixup
->offset
= next_local
;
3882 fixup
->super
.active
= 0;
3884 for (i
= 0; i
< MAX_FRAGMENT_SAMPLERS
; ++i
)
3886 if (!(map
& (1u << i
)))
3889 if (fixup
->offset
+ (cur_fixup_sampler
>> 1) < max_lconsts
)
3891 fixup
->super
.active
|= (1u << i
);
3892 fixup
->super
.idx
[i
] = cur_fixup_sampler
++;
3896 FIXME("No free constant found to load NP2 fixup data into shader. "
3897 "Sampling from this texture will probably look wrong.\n");
3902 fixup
->super
.num_consts
= (cur_fixup_sampler
+ 1) >> 1;
3903 if (fixup
->super
.num_consts
) {
3904 shader_addline(buffer
, "PARAM np2fixup[%u] = { program.env[%u..%u] };\n",
3905 fixup
->super
.num_consts
, fixup
->offset
, fixup
->super
.num_consts
+ fixup
->offset
- 1);
3909 if (shader_priv
->clipplane_emulation
!= ~0U && args
->clip
)
3911 shader_addline(buffer
, "KIL fragment.texcoord[%u];\n", shader_priv
->clipplane_emulation
);
3914 /* Base Shader Body */
3915 shader_generate_main(shader
, buffer
, reg_maps
, function
, &priv_ctx
);
3917 if(args
->super
.srgb_correction
) {
3918 arbfp_add_sRGB_correction(buffer
, fragcolor
, srgbtmp
[0], srgbtmp
[1], srgbtmp
[2], srgbtmp
[3],
3919 priv_ctx
.target_version
>= NV2
);
3922 if (custom_linear_fog
)
3923 arbfp_add_linear_fog(buffer
, fragcolor
, "TA");
3925 if(strcmp(fragcolor
, "result.color")) {
3926 shader_addline(buffer
, "MOV result.color, %s;\n", fragcolor
);
3928 shader_addline(buffer
, "END\n");
3930 /* TODO: change to resource.glObjectHandle or something like that */
3931 GL_EXTCALL(glGenProgramsARB(1, &retval
));
3933 TRACE("Creating a hw pixel shader, prg=%d\n", retval
);
3934 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, retval
));
3936 TRACE("Created hw pixel shader, prg=%d\n", retval
);
3937 if (!shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
->buffer
))
3943 static int compare_sig(const struct wined3d_shader_signature
*sig1
, const struct wined3d_shader_signature
*sig2
)
3948 if (sig1
->element_count
!= sig2
->element_count
)
3949 return sig1
->element_count
< sig2
->element_count
? -1 : 1;
3951 for (i
= 0; i
< sig1
->element_count
; ++i
)
3953 const struct wined3d_shader_signature_element
*e1
, *e2
;
3955 e1
= &sig1
->elements
[i
];
3956 e2
= &sig2
->elements
[i
];
3958 if (!e1
->semantic_name
|| !e2
->semantic_name
)
3960 /* Compare pointers, not contents. One string is NULL (element
3961 * does not exist), the other one is not NULL. */
3962 if (e1
->semantic_name
!= e2
->semantic_name
)
3963 return e1
->semantic_name
< e2
->semantic_name
? -1 : 1;
3967 if ((ret
= strcmp(e1
->semantic_name
, e2
->semantic_name
)))
3969 if (e1
->semantic_idx
!= e2
->semantic_idx
)
3970 return e1
->semantic_idx
< e2
->semantic_idx
? -1 : 1;
3971 if (e1
->sysval_semantic
!= e2
->sysval_semantic
)
3972 return e1
->sysval_semantic
< e2
->sysval_semantic
? -1 : 1;
3973 if (e1
->component_type
!= e2
->component_type
)
3974 return e1
->component_type
< e2
->component_type
? -1 : 1;
3975 if (e1
->register_idx
!= e2
->register_idx
)
3976 return e1
->register_idx
< e2
->register_idx
? -1 : 1;
3977 if (e1
->mask
!= e2
->mask
)
3978 return e1
->mask
< e2
->mask
? -1 : 1;
3983 static void clone_sig(struct wined3d_shader_signature
*new, const struct wined3d_shader_signature
*sig
)
3988 new->element_count
= sig
->element_count
;
3989 new->elements
= HeapAlloc(GetProcessHeap(), 0, sizeof(*new->elements
) * new->element_count
);
3990 for (i
= 0; i
< sig
->element_count
; ++i
)
3992 new->elements
[i
] = sig
->elements
[i
];
3994 if (!new->elements
[i
].semantic_name
)
3997 /* Clone the semantic string */
3998 name
= HeapAlloc(GetProcessHeap(), 0, strlen(sig
->elements
[i
].semantic_name
) + 1);
3999 strcpy(name
, sig
->elements
[i
].semantic_name
);
4000 new->elements
[i
].semantic_name
= name
;
4004 static DWORD
find_input_signature(struct shader_arb_priv
*priv
, const struct wined3d_shader_signature
*sig
)
4006 struct wine_rb_entry
*entry
= wine_rb_get(&priv
->signature_tree
, sig
);
4007 struct ps_signature
*found_sig
;
4011 found_sig
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
4012 TRACE("Found existing signature %u\n", found_sig
->idx
);
4013 return found_sig
->idx
;
4015 found_sig
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*found_sig
));
4016 clone_sig(&found_sig
->sig
, sig
);
4017 found_sig
->idx
= priv
->ps_sig_number
++;
4018 TRACE("New signature stored and assigned number %u\n", found_sig
->idx
);
4019 if(wine_rb_put(&priv
->signature_tree
, sig
, &found_sig
->entry
) == -1)
4021 ERR("Failed to insert program entry.\n");
4023 return found_sig
->idx
;
4026 static void init_output_registers(const struct wined3d_shader
*shader
,
4027 const struct wined3d_shader_signature
*ps_input_sig
,
4028 struct shader_arb_ctx_priv
*priv_ctx
, struct arb_vs_compiled_shader
*compiled
)
4031 static const char * const texcoords
[8] =
4033 "result.texcoord[0]", "result.texcoord[1]", "result.texcoord[2]", "result.texcoord[3]",
4034 "result.texcoord[4]", "result.texcoord[5]", "result.texcoord[6]", "result.texcoord[7]"
4036 /* Write generic input varyings 0 to 7 to result.texcoord[], varying 8 to result.color.primary
4037 * and varying 9 to result.color.secondary
4039 static const char * const decl_idx_to_string
[MAX_REG_INPUT
] =
4041 "result.texcoord[0]", "result.texcoord[1]", "result.texcoord[2]", "result.texcoord[3]",
4042 "result.texcoord[4]", "result.texcoord[5]", "result.texcoord[6]", "result.texcoord[7]",
4043 "result.color.primary", "result.color.secondary"
4048 TRACE("Pixel shader uses builtin varyings\n");
4049 /* Map builtins to builtins */
4050 for(i
= 0; i
< 8; i
++)
4052 priv_ctx
->texcrd_output
[i
] = texcoords
[i
];
4054 priv_ctx
->color_output
[0] = "result.color.primary";
4055 priv_ctx
->color_output
[1] = "result.color.secondary";
4056 priv_ctx
->fog_output
= "TMP_FOGCOORD";
4058 /* Map declared regs to builtins. Use "TA" to /dev/null unread output */
4059 for (i
= 0; i
< shader
->output_signature
.element_count
; ++i
)
4061 const struct wined3d_shader_signature_element
*output
= &shader
->output_signature
.elements
[i
];
4063 if (!output
->semantic_name
)
4066 if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_POSITION
))
4068 TRACE("o%u is TMP_OUT\n", output
->register_idx
);
4069 if (!output
->semantic_idx
)
4070 priv_ctx
->vs_output
[output
->register_idx
] = "TMP_OUT";
4072 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4074 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_PSIZE
))
4076 TRACE("o%u is result.pointsize\n", output
->register_idx
);
4077 if (!output
->semantic_idx
)
4078 priv_ctx
->vs_output
[output
->register_idx
] = "result.pointsize";
4080 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4082 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_COLOR
))
4084 TRACE("o%u is result.color.?, idx %u\n", output
->register_idx
, output
->semantic_idx
);
4085 if (!output
->semantic_idx
)
4086 priv_ctx
->vs_output
[output
->register_idx
] = "result.color.primary";
4087 else if (output
->semantic_idx
== 1)
4088 priv_ctx
->vs_output
[output
->register_idx
] = "result.color.secondary";
4089 else priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4091 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
4093 TRACE("o%u is result.texcoord[%u]\n", output
->register_idx
, output
->semantic_idx
);
4094 if (output
->semantic_idx
>= 8)
4095 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4097 priv_ctx
->vs_output
[output
->register_idx
] = texcoords
[output
->semantic_idx
];
4099 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_FOG
))
4101 TRACE("o%u is result.fogcoord\n", output
->register_idx
);
4102 if (output
->semantic_idx
> 0)
4103 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4105 priv_ctx
->vs_output
[output
->register_idx
] = "result.fogcoord";
4109 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4115 TRACE("Pixel shader uses declared varyings\n");
4117 /* Map builtin to declared. /dev/null the results by default to the TA temp reg */
4118 for(i
= 0; i
< 8; i
++)
4120 priv_ctx
->texcrd_output
[i
] = "TA";
4122 priv_ctx
->color_output
[0] = "TA";
4123 priv_ctx
->color_output
[1] = "TA";
4124 priv_ctx
->fog_output
= "TA";
4126 for (i
= 0; i
< ps_input_sig
->element_count
; ++i
)
4128 const struct wined3d_shader_signature_element
*input
= &ps_input_sig
->elements
[i
];
4130 if (!input
->semantic_name
)
4133 /* If a declared input register is not written by builtin arguments, don't write to it.
4134 * GL_NV_vertex_program makes sure the input defaults to 0.0, which is correct with D3D
4136 * Don't care about POSITION and PSIZE here - this is a builtin vertex shader, position goes
4137 * to TMP_OUT in any case
4139 if (shader_match_semantic(input
->semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
4141 if (input
->semantic_idx
< 8)
4142 priv_ctx
->texcrd_output
[input
->semantic_idx
] = decl_idx_to_string
[input
->register_idx
];
4144 else if (shader_match_semantic(input
->semantic_name
, WINED3D_DECL_USAGE_COLOR
))
4146 if (input
->semantic_idx
< 2)
4147 priv_ctx
->color_output
[input
->semantic_idx
] = decl_idx_to_string
[input
->register_idx
];
4149 else if (shader_match_semantic(input
->semantic_name
, WINED3D_DECL_USAGE_FOG
))
4151 if (!input
->semantic_idx
)
4152 priv_ctx
->fog_output
= decl_idx_to_string
[input
->register_idx
];
4159 if (!strcmp(decl_idx_to_string
[input
->register_idx
], "result.color.primary")
4160 || !strcmp(decl_idx_to_string
[input
->register_idx
], "result.color.secondary"))
4162 compiled
->need_color_unclamp
= TRUE
;
4166 /* Map declared to declared */
4167 for (i
= 0; i
< shader
->output_signature
.element_count
; ++i
)
4169 const struct wined3d_shader_signature_element
*output
= &shader
->output_signature
.elements
[i
];
4171 /* Write unread output to TA to throw them away */
4172 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4174 if (!output
->semantic_name
)
4177 if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_POSITION
) && !output
->semantic_idx
)
4179 priv_ctx
->vs_output
[output
->register_idx
] = "TMP_OUT";
4182 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_PSIZE
) && !output
->semantic_idx
)
4184 priv_ctx
->vs_output
[output
->register_idx
] = "result.pointsize";
4188 for (j
= 0; j
< ps_input_sig
->element_count
; ++j
)
4190 const struct wined3d_shader_signature_element
*input
= &ps_input_sig
->elements
[j
];
4192 if (!input
->semantic_name
)
4195 if (!strcmp(input
->semantic_name
, output
->semantic_name
)
4196 && input
->semantic_idx
== output
->semantic_idx
)
4198 priv_ctx
->vs_output
[output
->register_idx
] = decl_idx_to_string
[input
->register_idx
];
4200 if (!strcmp(priv_ctx
->vs_output
[output
->register_idx
], "result.color.primary")
4201 || !strcmp(priv_ctx
->vs_output
[output
->register_idx
], "result.color.secondary"))
4203 compiled
->need_color_unclamp
= TRUE
;
4210 /* Context activation is done by the caller. */
4211 static GLuint
shader_arb_generate_vshader(const struct wined3d_shader
*shader
,
4212 const struct wined3d_gl_info
*gl_info
, struct wined3d_string_buffer
*buffer
,
4213 const struct arb_vs_compile_args
*args
, struct arb_vs_compiled_shader
*compiled
,
4214 const struct wined3d_shader_signature
*ps_input_sig
)
4216 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
4217 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
4218 struct shader_arb_priv
*priv
= shader
->device
->shader_priv
;
4219 const DWORD
*function
= shader
->function
;
4221 DWORD next_local
= 0;
4222 struct shader_arb_ctx_priv priv_ctx
;
4225 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
4226 priv_ctx
.cur_vs_args
= args
;
4227 list_init(&priv_ctx
.control_frames
);
4228 init_output_registers(shader
, ps_input_sig
, &priv_ctx
, compiled
);
4230 /* Create the hw ARB shader */
4231 shader_addline(buffer
, "!!ARBvp1.0\n");
4233 /* Always enable the NV extension if available. Unlike fragment shaders, there is no
4234 * mesurable performance penalty, and we can always make use of it for clipplanes.
4236 if (gl_info
->supported
[NV_VERTEX_PROGRAM3
])
4238 shader_addline(buffer
, "OPTION NV_vertex_program3;\n");
4239 priv_ctx
.target_version
= NV3
;
4240 shader_addline(buffer
, "ADDRESS aL;\n");
4242 else if (gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
])
4244 shader_addline(buffer
, "OPTION NV_vertex_program2;\n");
4245 priv_ctx
.target_version
= NV2
;
4246 shader_addline(buffer
, "ADDRESS aL;\n");
4248 priv_ctx
.target_version
= ARB
;
4251 shader_addline(buffer
, "TEMP TMP_OUT;\n");
4253 shader_addline(buffer
, "TEMP TMP_FOGCOORD;\n");
4254 if (need_helper_const(shader_data
, reg_maps
, gl_info
))
4257 wined3d_ftoa(eps
, ftoa_tmp
);
4258 shader_addline(buffer
, "PARAM helper_const = { 0.0, 1.0, 2.0, %s};\n", ftoa_tmp
);
4260 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
))
4262 shader_addline(buffer
, "PARAM rel_addr_const = { 0.5, %d.0, 0.0, 0.0 };\n", shader_data
->rel_offset
);
4263 shader_addline(buffer
, "TEMP A0_SHADOW;\n");
4266 shader_addline(buffer
, "TEMP TA;\n");
4267 shader_addline(buffer
, "TEMP TB;\n");
4269 /* Base Declarations */
4270 shader_generate_arb_declarations(shader
, reg_maps
, buffer
, gl_info
,
4271 &priv_ctx
.vs_clipplanes
, &priv_ctx
);
4273 for(i
= 0; i
< MAX_CONST_I
; i
++)
4275 compiled
->int_consts
[i
] = WINED3D_CONST_NUM_UNUSED
;
4276 if (reg_maps
->integer_constants
& (1u << i
) && priv_ctx
.target_version
>= NV2
)
4278 const DWORD
*control_values
= find_loop_control_values(shader
, i
);
4282 shader_addline(buffer
, "PARAM I%u = {%u, %u, %u, -1};\n", i
,
4283 control_values
[0], control_values
[1], control_values
[2]);
4287 compiled
->int_consts
[i
] = next_local
;
4288 compiled
->num_int_consts
++;
4289 shader_addline(buffer
, "PARAM I%u = program.local[%u];\n", i
, next_local
++);
4294 /* We need a constant to fixup the final position */
4295 shader_addline(buffer
, "PARAM posFixup = program.local[%u];\n", next_local
);
4296 compiled
->pos_fixup
= next_local
++;
4298 /* Initialize output parameters. GL_ARB_vertex_program does not require special initialization values
4299 * for output parameters. D3D in theory does not do that either, but some applications depend on a
4300 * proper initialization of the secondary color, and programs using the fixed function pipeline without
4301 * a replacement shader depend on the texcoord.w being set properly.
4303 * GL_NV_vertex_program defines that all output values are initialized to {0.0, 0.0, 0.0, 1.0}. This
4304 * assertion is in effect even when using GL_ARB_vertex_program without any NV specific additions. So
4305 * skip this if NV_vertex_program is supported. Otherwise, initialize the secondary color. For the tex-
4306 * coords, we have a flag in the opengl caps. Many cards do not require the texcoord being set, and
4307 * this can eat a number of instructions, so skip it unless this cap is set as well
4309 if (!gl_info
->supported
[NV_VERTEX_PROGRAM
])
4311 const char *color_init
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_0001
);
4312 shader_addline(buffer
, "MOV result.color.secondary, %s;\n", color_init
);
4314 if (gl_info
->quirks
& WINED3D_QUIRK_SET_TEXCOORD_W
&& !priv
->ffp_proj_control
)
4317 const char *one
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_ONE
);
4318 for(i
= 0; i
< MAX_REG_TEXCRD
; i
++)
4320 if (reg_maps
->u
.texcoord_mask
[i
] && reg_maps
->u
.texcoord_mask
[i
] != WINED3DSP_WRITEMASK_ALL
)
4321 shader_addline(buffer
, "MOV result.texcoord[%u].w, %s\n", i
, one
);
4326 /* The shader starts with the main function */
4327 priv_ctx
.in_main_func
= TRUE
;
4328 /* Base Shader Body */
4329 shader_generate_main(shader
, buffer
, reg_maps
, function
, &priv_ctx
);
4331 if (!priv_ctx
.footer_written
) vshader_add_footer(&priv_ctx
,
4332 shader_data
, args
, reg_maps
, gl_info
, buffer
);
4334 shader_addline(buffer
, "END\n");
4336 /* TODO: change to resource.glObjectHandle or something like that */
4337 GL_EXTCALL(glGenProgramsARB(1, &ret
));
4339 TRACE("Creating a hw vertex shader, prg=%d\n", ret
);
4340 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, ret
));
4342 TRACE("Created hw vertex shader, prg=%d\n", ret
);
4343 if (!shader_arb_compile(gl_info
, GL_VERTEX_PROGRAM_ARB
, buffer
->buffer
))
4349 /* Context activation is done by the caller. */
4350 static struct arb_ps_compiled_shader
*find_arb_pshader(struct wined3d_shader
*shader
,
4351 const struct arb_ps_compile_args
*args
)
4353 struct wined3d_device
*device
= shader
->device
;
4354 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
4355 const struct wined3d_d3d_info
*d3d_info
= &device
->adapter
->d3d_info
;
4358 struct arb_ps_compiled_shader
*new_array
;
4359 struct wined3d_string_buffer buffer
;
4360 struct arb_pshader_private
*shader_data
;
4363 if (!shader
->backend_data
)
4365 struct shader_arb_priv
*priv
= device
->shader_priv
;
4367 shader
->backend_data
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
4368 shader_data
= shader
->backend_data
;
4369 shader_data
->clamp_consts
= shader
->reg_maps
.shader_version
.major
== 1;
4371 if (shader
->reg_maps
.shader_version
.major
< 3)
4372 shader_data
->input_signature_idx
= ~0U;
4374 shader_data
->input_signature_idx
= find_input_signature(priv
, &shader
->input_signature
);
4376 TRACE("Shader got assigned input signature index %u\n", shader_data
->input_signature_idx
);
4378 if (!d3d_info
->vs_clipping
)
4379 shader_data
->clipplane_emulation
= shader_find_free_input_register(&shader
->reg_maps
,
4380 d3d_info
->limits
.ffp_blend_stages
- 1);
4382 shader_data
->clipplane_emulation
= ~0U;
4384 shader_data
= shader
->backend_data
;
4386 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4387 * so a linear search is more performant than a hashmap or a binary search
4388 * (cache coherency etc)
4390 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4392 if (!memcmp(&shader_data
->gl_shaders
[i
].args
, args
, sizeof(*args
)))
4393 return &shader_data
->gl_shaders
[i
];
4396 TRACE("No matching GL shader found, compiling a new shader\n");
4397 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4398 if (shader_data
->num_gl_shaders
)
4400 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4401 new_array
= HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, shader_data
->gl_shaders
,
4402 new_size
* sizeof(*shader_data
->gl_shaders
));
4404 new_array
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
->gl_shaders
));
4409 ERR("Out of memory\n");
4412 shader_data
->gl_shaders
= new_array
;
4413 shader_data
->shader_array_size
= new_size
;
4416 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4418 pixelshader_update_resource_types(shader
, args
->super
.tex_types
);
4420 if (!string_buffer_init(&buffer
))
4422 ERR("Failed to initialize shader buffer.\n");
4426 ret
= shader_arb_generate_pshader(shader
, gl_info
, &buffer
, args
,
4427 &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
]);
4428 string_buffer_free(&buffer
);
4429 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].prgId
= ret
;
4431 return &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++];
4434 static inline BOOL
vs_args_equal(const struct arb_vs_compile_args
*stored
, const struct arb_vs_compile_args
*new,
4435 const DWORD use_map
, BOOL skip_int
) {
4436 if((stored
->super
.swizzle_map
& use_map
) != new->super
.swizzle_map
) return FALSE
;
4437 if(stored
->super
.clip_enabled
!= new->super
.clip_enabled
) return FALSE
;
4438 if(stored
->super
.fog_src
!= new->super
.fog_src
) return FALSE
;
4439 if(stored
->clip
.boolclip_compare
!= new->clip
.boolclip_compare
) return FALSE
;
4440 if(stored
->ps_signature
!= new->ps_signature
) return FALSE
;
4441 if(stored
->vertex
.samplers_compare
!= new->vertex
.samplers_compare
) return FALSE
;
4442 if(skip_int
) return TRUE
;
4444 return !memcmp(stored
->loop_ctrl
, new->loop_ctrl
, sizeof(stored
->loop_ctrl
));
4447 static struct arb_vs_compiled_shader
*find_arb_vshader(struct wined3d_shader
*shader
,
4448 const struct wined3d_gl_info
*gl_info
, DWORD use_map
, const struct arb_vs_compile_args
*args
,
4449 const struct wined3d_shader_signature
*ps_input_sig
)
4453 struct arb_vs_compiled_shader
*new_array
;
4454 struct wined3d_string_buffer buffer
;
4455 struct arb_vshader_private
*shader_data
;
4458 if (!shader
->backend_data
)
4460 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
4462 shader
->backend_data
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
4463 shader_data
= shader
->backend_data
;
4465 if ((gl_info
->quirks
& WINED3D_QUIRK_ARB_VS_OFFSET_LIMIT
)
4466 && reg_maps
->min_rel_offset
<= reg_maps
->max_rel_offset
)
4468 if (reg_maps
->max_rel_offset
- reg_maps
->min_rel_offset
> 127)
4470 FIXME("The difference between the minimum and maximum relative offset is > 127.\n");
4471 FIXME("Which this OpenGL implementation does not support. Try using GLSL.\n");
4472 FIXME("Min: %u, Max: %u.\n", reg_maps
->min_rel_offset
, reg_maps
->max_rel_offset
);
4474 else if (reg_maps
->max_rel_offset
- reg_maps
->min_rel_offset
> 63)
4475 shader_data
->rel_offset
= reg_maps
->min_rel_offset
+ 63;
4476 else if (reg_maps
->max_rel_offset
> 63)
4477 shader_data
->rel_offset
= reg_maps
->min_rel_offset
;
4480 shader_data
= shader
->backend_data
;
4482 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4483 * so a linear search is more performant than a hashmap or a binary search
4484 * (cache coherency etc)
4486 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
4487 if (vs_args_equal(&shader_data
->gl_shaders
[i
].args
, args
,
4488 use_map
, gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]))
4490 return &shader_data
->gl_shaders
[i
];
4494 TRACE("No matching GL shader found, compiling a new shader\n");
4496 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4497 if (shader_data
->num_gl_shaders
)
4499 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4500 new_array
= HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, shader_data
->gl_shaders
,
4501 new_size
* sizeof(*shader_data
->gl_shaders
));
4503 new_array
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
->gl_shaders
));
4508 ERR("Out of memory\n");
4511 shader_data
->gl_shaders
= new_array
;
4512 shader_data
->shader_array_size
= new_size
;
4515 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4517 if (!string_buffer_init(&buffer
))
4519 ERR("Failed to initialize shader buffer.\n");
4523 ret
= shader_arb_generate_vshader(shader
, gl_info
, &buffer
, args
,
4524 &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
],
4526 string_buffer_free(&buffer
);
4527 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].prgId
= ret
;
4529 return &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++];
4532 static void find_arb_ps_compile_args(const struct wined3d_state
*state
,
4533 const struct wined3d_context
*context
, const struct wined3d_shader
*shader
,
4534 struct arb_ps_compile_args
*args
)
4536 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4537 const struct wined3d_d3d_info
*d3d_info
= context
->d3d_info
;
4541 find_ps_compile_args(state
, shader
, context
->stream_info
.position_transformed
, &args
->super
, context
);
4543 /* This forces all local boolean constants to 1 to make them stateblock independent */
4544 args
->bools
= shader
->reg_maps
.local_bool_consts
;
4546 for(i
= 0; i
< MAX_CONST_B
; i
++)
4548 if (state
->ps_consts_b
[i
])
4549 args
->bools
|= ( 1u << i
);
4552 /* Only enable the clip plane emulation KIL if at least one clipplane is enabled. The KIL instruction
4553 * is quite expensive because it forces the driver to disable early Z discards. It is cheaper to
4554 * duplicate the shader than have a no-op KIL instruction in every shader
4556 if (!d3d_info
->vs_clipping
&& use_vs(state
)
4557 && state
->render_states
[WINED3D_RS_CLIPPING
]
4558 && state
->render_states
[WINED3D_RS_CLIPPLANEENABLE
])
4563 /* Skip if unused or local, or supported natively */
4564 int_skip
= ~shader
->reg_maps
.integer_constants
| shader
->reg_maps
.local_int_consts
;
4565 if (int_skip
== 0xffff || gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
])
4567 memset(args
->loop_ctrl
, 0, sizeof(args
->loop_ctrl
));
4571 for (i
= 0; i
< MAX_CONST_I
; ++i
)
4573 if (int_skip
& (1u << i
))
4575 args
->loop_ctrl
[i
][0] = 0;
4576 args
->loop_ctrl
[i
][1] = 0;
4577 args
->loop_ctrl
[i
][2] = 0;
4581 args
->loop_ctrl
[i
][0] = state
->ps_consts_i
[i
* 4];
4582 args
->loop_ctrl
[i
][1] = state
->ps_consts_i
[i
* 4 + 1];
4583 args
->loop_ctrl
[i
][2] = state
->ps_consts_i
[i
* 4 + 2];
4588 static void find_arb_vs_compile_args(const struct wined3d_state
*state
,
4589 const struct wined3d_context
*context
, const struct wined3d_shader
*shader
,
4590 struct arb_vs_compile_args
*args
)
4592 const struct wined3d_device
*device
= shader
->device
;
4593 const struct wined3d_adapter
*adapter
= device
->adapter
;
4594 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4595 const struct wined3d_d3d_info
*d3d_info
= context
->d3d_info
;
4599 find_vs_compile_args(state
, shader
, context
->stream_info
.swizzle_map
, &args
->super
, d3d_info
);
4601 args
->clip
.boolclip_compare
= 0;
4604 const struct wined3d_shader
*ps
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
4605 const struct arb_pshader_private
*shader_priv
= ps
->backend_data
;
4606 args
->ps_signature
= shader_priv
->input_signature_idx
;
4608 args
->clip
.boolclip
.clip_texcoord
= shader_priv
->clipplane_emulation
+ 1;
4612 args
->ps_signature
= ~0;
4613 if (!d3d_info
->vs_clipping
&& adapter
->fragment_pipe
== &arbfp_fragment_pipeline
)
4614 args
->clip
.boolclip
.clip_texcoord
= ffp_clip_emul(context
) ? d3d_info
->limits
.ffp_blend_stages
: 0;
4615 /* Otherwise: Setting boolclip_compare set clip_texcoord to 0 */
4618 if (args
->clip
.boolclip
.clip_texcoord
)
4620 if (state
->render_states
[WINED3D_RS_CLIPPING
])
4621 args
->clip
.boolclip
.clipplane_mask
= (unsigned char)state
->render_states
[WINED3D_RS_CLIPPLANEENABLE
];
4622 /* clipplane_mask was set to 0 by setting boolclip_compare to 0 */
4625 /* This forces all local boolean constants to 1 to make them stateblock independent */
4626 args
->clip
.boolclip
.bools
= shader
->reg_maps
.local_bool_consts
;
4627 /* TODO: Figure out if it would be better to store bool constants as bitmasks in the stateblock */
4628 for(i
= 0; i
< MAX_CONST_B
; i
++)
4630 if (state
->vs_consts_b
[i
])
4631 args
->clip
.boolclip
.bools
|= (1u << i
);
4634 args
->vertex
.samplers
[0] = context
->tex_unit_map
[MAX_FRAGMENT_SAMPLERS
+ 0];
4635 args
->vertex
.samplers
[1] = context
->tex_unit_map
[MAX_FRAGMENT_SAMPLERS
+ 1];
4636 args
->vertex
.samplers
[2] = context
->tex_unit_map
[MAX_FRAGMENT_SAMPLERS
+ 2];
4637 args
->vertex
.samplers
[3] = 0;
4639 /* Skip if unused or local */
4640 int_skip
= ~shader
->reg_maps
.integer_constants
| shader
->reg_maps
.local_int_consts
;
4641 /* This is about flow control, not clipping. */
4642 if (int_skip
== 0xffff || gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
])
4644 memset(args
->loop_ctrl
, 0, sizeof(args
->loop_ctrl
));
4648 for(i
= 0; i
< MAX_CONST_I
; i
++)
4650 if (int_skip
& (1u << i
))
4652 args
->loop_ctrl
[i
][0] = 0;
4653 args
->loop_ctrl
[i
][1] = 0;
4654 args
->loop_ctrl
[i
][2] = 0;
4658 args
->loop_ctrl
[i
][0] = state
->vs_consts_i
[i
* 4];
4659 args
->loop_ctrl
[i
][1] = state
->vs_consts_i
[i
* 4 + 1];
4660 args
->loop_ctrl
[i
][2] = state
->vs_consts_i
[i
* 4 + 2];
4665 /* Context activation is done by the caller. */
4666 static void shader_arb_select(void *shader_priv
, struct wined3d_context
*context
,
4667 const struct wined3d_state
*state
)
4669 struct shader_arb_priv
*priv
= shader_priv
;
4670 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4673 /* Deal with pixel shaders first so the vertex shader arg function has the input signature ready */
4676 struct wined3d_shader
*ps
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
4677 struct arb_ps_compile_args compile_args
;
4678 struct arb_ps_compiled_shader
*compiled
;
4680 TRACE("Using pixel shader %p.\n", ps
);
4681 find_arb_ps_compile_args(state
, context
, ps
, &compile_args
);
4682 compiled
= find_arb_pshader(ps
, &compile_args
);
4683 priv
->current_fprogram_id
= compiled
->prgId
;
4684 priv
->compiled_fprog
= compiled
;
4686 /* Bind the fragment program */
4687 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, priv
->current_fprogram_id
));
4688 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->current_fprogram_id);");
4690 if (!priv
->use_arbfp_fixed_func
)
4691 priv
->fragment_pipe
->enable_extension(gl_info
, FALSE
);
4693 /* Enable OpenGL fragment programs. */
4694 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
4695 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
4697 TRACE("Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", priv
->current_fprogram_id
);
4699 /* Pixel Shader 1.x constants are clamped to [-1;1], Pixel Shader 2.0 constants are not. If switching between
4700 * a 1.x and newer shader, reload the first 8 constants
4702 if (priv
->last_ps_const_clamped
!= ((struct arb_pshader_private
*)ps
->backend_data
)->clamp_consts
)
4704 priv
->last_ps_const_clamped
= ((struct arb_pshader_private
*)ps
->backend_data
)->clamp_consts
;
4705 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, 8);
4706 for(i
= 0; i
< 8; i
++)
4708 priv
->pshader_const_dirty
[i
] = 1;
4710 /* Also takes care of loading local constants */
4711 shader_arb_load_constants_internal(shader_priv
, context
, state
, TRUE
, FALSE
, TRUE
);
4715 UINT rt_height
= state
->fb
->render_targets
[0]->height
;
4716 shader_arb_ps_local_constants(compiled
, context
, state
, rt_height
);
4719 /* Force constant reloading for the NP2 fixup (see comment in shader_glsl_select for more info) */
4720 if (compiled
->np2fixup_info
.super
.active
)
4721 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_NP2_FIXUP
;
4723 if (ps
->load_local_constsF
)
4724 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_F
;
4728 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
] && !priv
->use_arbfp_fixed_func
)
4730 /* Disable only if we're not using arbfp fixed function fragment
4731 * processing. If this is used, keep GL_FRAGMENT_PROGRAM_ARB
4732 * enabled, and the fixed function pipeline will bind the fixed
4733 * function replacement shader. */
4734 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4735 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4736 priv
->current_fprogram_id
= 0;
4738 priv
->fragment_pipe
->enable_extension(gl_info
, TRUE
);
4743 struct wined3d_shader
*vs
= state
->shader
[WINED3D_SHADER_TYPE_VERTEX
];
4744 struct arb_vs_compile_args compile_args
;
4745 struct arb_vs_compiled_shader
*compiled
;
4746 const struct wined3d_shader_signature
*ps_input_sig
;
4748 TRACE("Using vertex shader %p\n", vs
);
4749 find_arb_vs_compile_args(state
, context
, vs
, &compile_args
);
4751 /* Instead of searching for the signature in the signature list, read the one from the
4752 * current pixel shader. It's maybe not the shader where the signature came from, but it
4753 * is the same signature and faster to find. */
4754 if (compile_args
.ps_signature
== ~0U)
4755 ps_input_sig
= NULL
;
4757 ps_input_sig
= &state
->shader
[WINED3D_SHADER_TYPE_PIXEL
]->input_signature
;
4759 compiled
= find_arb_vshader(vs
, context
->gl_info
, context
->stream_info
.use_map
,
4760 &compile_args
, ps_input_sig
);
4761 priv
->current_vprogram_id
= compiled
->prgId
;
4762 priv
->compiled_vprog
= compiled
;
4764 /* Bind the vertex program */
4765 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->current_vprogram_id
));
4766 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->current_vprogram_id);");
4768 priv
->vertex_pipe
->vp_enable(gl_info
, FALSE
);
4770 /* Enable OpenGL vertex programs */
4771 gl_info
->gl_ops
.gl
.p_glEnable(GL_VERTEX_PROGRAM_ARB
);
4772 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
4773 TRACE("Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", priv
->current_vprogram_id
);
4774 shader_arb_vs_local_constants(compiled
, context
, state
);
4776 if(priv
->last_vs_color_unclamp
!= compiled
->need_color_unclamp
) {
4777 priv
->last_vs_color_unclamp
= compiled
->need_color_unclamp
;
4779 if (gl_info
->supported
[ARB_COLOR_BUFFER_FLOAT
])
4781 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, !compiled
->need_color_unclamp
));
4782 checkGLcall("glClampColorARB");
4784 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
4788 if (vs
->load_local_constsF
)
4789 context
->constant_update_mask
|= WINED3D_SHADER_CONST_VS_F
;
4793 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
4795 priv
->current_vprogram_id
= 0;
4796 gl_info
->gl_ops
.gl
.p_glDisable(GL_VERTEX_PROGRAM_ARB
);
4797 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4799 priv
->vertex_pipe
->vp_enable(gl_info
, TRUE
);
4804 /* Context activation is done by the caller. */
4805 static void shader_arb_disable(void *shader_priv
, struct wined3d_context
*context
)
4807 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4808 struct shader_arb_priv
*priv
= shader_priv
;
4810 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
4812 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4813 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4814 priv
->current_fprogram_id
= 0;
4816 priv
->fragment_pipe
->enable_extension(gl_info
, FALSE
);
4818 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
4820 priv
->current_vprogram_id
= 0;
4821 gl_info
->gl_ops
.gl
.p_glDisable(GL_VERTEX_PROGRAM_ARB
);
4822 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4824 priv
->vertex_pipe
->vp_enable(gl_info
, FALSE
);
4826 if (gl_info
->supported
[ARB_COLOR_BUFFER_FLOAT
] && priv
->last_vs_color_unclamp
)
4828 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, GL_FIXED_ONLY_ARB
));
4829 checkGLcall("glClampColorARB");
4830 priv
->last_vs_color_unclamp
= FALSE
;
4833 context
->shader_update_mask
= (1u << WINED3D_SHADER_TYPE_PIXEL
)
4834 | (1u << WINED3D_SHADER_TYPE_VERTEX
)
4835 | (1u << WINED3D_SHADER_TYPE_GEOMETRY
)
4836 | (1u << WINED3D_SHADER_TYPE_HULL
);
4839 /* Context activation is done by the caller. */
4840 static void shader_arb_select_depth_blt(void *shader_priv
, const struct wined3d_gl_info
*gl_info
,
4841 enum wined3d_gl_resource_type tex_type
, const SIZE
*ds_mask_size
)
4843 const float mask
[] = {0.0f
, 0.0f
, (float)ds_mask_size
->cx
, (float)ds_mask_size
->cy
};
4844 BOOL masked
= ds_mask_size
->cx
&& ds_mask_size
->cy
;
4845 struct shader_arb_priv
*priv
= shader_priv
;
4846 GLuint
*blt_fprogram
;
4848 if (!priv
->depth_blt_vprogram_id
) priv
->depth_blt_vprogram_id
= create_arb_blt_vertex_program(gl_info
);
4849 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->depth_blt_vprogram_id
));
4850 gl_info
->gl_ops
.gl
.p_glEnable(GL_VERTEX_PROGRAM_ARB
);
4852 blt_fprogram
= masked
? &priv
->depth_blt_fprogram_id_masked
[tex_type
] : &priv
->depth_blt_fprogram_id_full
[tex_type
];
4853 if (!*blt_fprogram
) *blt_fprogram
= create_arb_blt_fragment_program(gl_info
, tex_type
, masked
);
4854 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, *blt_fprogram
));
4855 if (masked
) GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, 0, mask
));
4856 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
4859 /* Context activation is done by the caller. */
4860 static void shader_arb_deselect_depth_blt(void *shader_priv
, const struct wined3d_gl_info
*gl_info
)
4862 struct shader_arb_priv
*priv
= shader_priv
;
4864 if (priv
->current_vprogram_id
) {
4865 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->current_vprogram_id
));
4866 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
4868 TRACE("Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB.\n", priv
->current_vprogram_id
);
4872 gl_info
->gl_ops
.gl
.p_glDisable(GL_VERTEX_PROGRAM_ARB
);
4873 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4876 if (priv
->current_fprogram_id
) {
4877 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, priv
->current_fprogram_id
));
4878 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
4880 TRACE("Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB.\n", priv
->current_fprogram_id
);
4882 else if(!priv
->use_arbfp_fixed_func
)
4884 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4885 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4889 static void shader_arb_destroy(struct wined3d_shader
*shader
)
4891 struct wined3d_device
*device
= shader
->device
;
4892 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
4894 if (shader_is_pshader_version(shader
->reg_maps
.shader_version
.type
))
4896 struct arb_pshader_private
*shader_data
= shader
->backend_data
;
4899 if(!shader_data
) return; /* This can happen if a shader was never compiled */
4901 if (shader_data
->num_gl_shaders
)
4903 struct wined3d_context
*context
= context_acquire(device
, NULL
);
4905 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4907 GL_EXTCALL(glDeleteProgramsARB(1, &shader_data
->gl_shaders
[i
].prgId
));
4908 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader_data->gl_shaders[i].prgId))");
4911 context_release(context
);
4914 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
);
4915 HeapFree(GetProcessHeap(), 0, shader_data
);
4916 shader
->backend_data
= NULL
;
4920 struct arb_vshader_private
*shader_data
= shader
->backend_data
;
4923 if(!shader_data
) return; /* This can happen if a shader was never compiled */
4925 if (shader_data
->num_gl_shaders
)
4927 struct wined3d_context
*context
= context_acquire(device
, NULL
);
4929 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4931 GL_EXTCALL(glDeleteProgramsARB(1, &shader_data
->gl_shaders
[i
].prgId
));
4932 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader_data->gl_shaders[i].prgId))");
4935 context_release(context
);
4938 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
);
4939 HeapFree(GetProcessHeap(), 0, shader_data
);
4940 shader
->backend_data
= NULL
;
4944 static int sig_tree_compare(const void *key
, const struct wine_rb_entry
*entry
)
4946 struct ps_signature
*e
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
4947 return compare_sig(key
, &e
->sig
);
4950 static const struct wine_rb_functions sig_tree_functions
=
4958 static HRESULT
shader_arb_alloc(struct wined3d_device
*device
, const struct wined3d_vertex_pipe_ops
*vertex_pipe
,
4959 const struct fragment_pipeline
*fragment_pipe
)
4961 struct shader_arb_priv
*priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*priv
));
4962 struct fragment_caps fragment_caps
;
4963 void *vertex_priv
, *fragment_priv
;
4964 const struct wined3d_d3d_info
*d3d_info
= &device
->adapter
->d3d_info
;
4966 if (!(vertex_priv
= vertex_pipe
->vp_alloc(&arb_program_shader_backend
, priv
)))
4968 ERR("Failed to initialize vertex pipe.\n");
4969 HeapFree(GetProcessHeap(), 0, priv
);
4973 if (!(fragment_priv
= fragment_pipe
->alloc_private(&arb_program_shader_backend
, priv
)))
4975 ERR("Failed to initialize fragment pipe.\n");
4976 vertex_pipe
->vp_free(device
);
4977 HeapFree(GetProcessHeap(), 0, priv
);
4981 priv
->vshader_const_dirty
= HeapAlloc(GetProcessHeap(), 0,
4982 sizeof(*priv
->vshader_const_dirty
) * d3d_info
->limits
.vs_uniform_count
);
4983 if (!priv
->vshader_const_dirty
)
4985 memset(priv
->vshader_const_dirty
, 1,
4986 sizeof(*priv
->vshader_const_dirty
) * d3d_info
->limits
.vs_uniform_count
);
4988 priv
->pshader_const_dirty
= HeapAlloc(GetProcessHeap(), 0,
4989 sizeof(*priv
->pshader_const_dirty
) * d3d_info
->limits
.ps_uniform_count
);
4990 if (!priv
->pshader_const_dirty
)
4992 memset(priv
->pshader_const_dirty
, 1,
4993 sizeof(*priv
->pshader_const_dirty
) * d3d_info
->limits
.ps_uniform_count
);
4995 if(wine_rb_init(&priv
->signature_tree
, &sig_tree_functions
) == -1)
4997 ERR("RB tree init failed\n");
5001 priv
->vertex_pipe
= vertex_pipe
;
5002 priv
->fragment_pipe
= fragment_pipe
;
5003 fragment_pipe
->get_caps(&device
->adapter
->gl_info
, &fragment_caps
);
5004 priv
->ffp_proj_control
= fragment_caps
.wined3d_caps
& WINED3D_FRAGMENT_CAP_PROJ_CONTROL
;
5006 device
->vertex_priv
= vertex_priv
;
5007 device
->fragment_priv
= fragment_priv
;
5008 device
->shader_priv
= priv
;
5013 HeapFree(GetProcessHeap(), 0, priv
->pshader_const_dirty
);
5014 HeapFree(GetProcessHeap(), 0, priv
->vshader_const_dirty
);
5015 fragment_pipe
->free_private(device
);
5016 vertex_pipe
->vp_free(device
);
5017 HeapFree(GetProcessHeap(), 0, priv
);
5018 return E_OUTOFMEMORY
;
5021 static void release_signature(struct wine_rb_entry
*entry
, void *context
)
5023 struct ps_signature
*sig
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
5026 for (i
= 0; i
< sig
->sig
.element_count
; ++i
)
5028 HeapFree(GetProcessHeap(), 0, (char *)sig
->sig
.elements
[i
].semantic_name
);
5030 HeapFree(GetProcessHeap(), 0, sig
->sig
.elements
);
5031 HeapFree(GetProcessHeap(), 0, sig
);
5034 /* Context activation is done by the caller. */
5035 static void shader_arb_free(struct wined3d_device
*device
)
5037 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
5038 struct shader_arb_priv
*priv
= device
->shader_priv
;
5041 if (priv
->depth_blt_vprogram_id
)
5042 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_vprogram_id
));
5044 for (i
= 0; i
< WINED3D_GL_RES_TYPE_COUNT
; ++i
)
5046 if (priv
->depth_blt_fprogram_id_full
[i
])
5048 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_fprogram_id_full
[i
]));
5050 if (priv
->depth_blt_fprogram_id_masked
[i
])
5052 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_fprogram_id_masked
[i
]));
5056 wine_rb_destroy(&priv
->signature_tree
, release_signature
, NULL
);
5057 HeapFree(GetProcessHeap(), 0, priv
->pshader_const_dirty
);
5058 HeapFree(GetProcessHeap(), 0, priv
->vshader_const_dirty
);
5059 priv
->fragment_pipe
->free_private(device
);
5060 priv
->vertex_pipe
->vp_free(device
);
5061 HeapFree(GetProcessHeap(), 0, device
->shader_priv
);
5064 static BOOL
shader_arb_allocate_context_data(struct wined3d_context
*context
)
5069 static void shader_arb_free_context_data(struct wined3d_context
*context
)
5071 struct shader_arb_priv
*priv
;
5073 if (!context
->swapchain
)
5076 priv
= context
->swapchain
->device
->shader_priv
;
5077 if (priv
->last_context
== context
)
5078 priv
->last_context
= NULL
;
5081 static void shader_arb_init_context_state(struct wined3d_context
*context
) {}
5083 static void shader_arb_get_caps(const struct wined3d_gl_info
*gl_info
, struct shader_caps
*caps
)
5085 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
5090 /* 96 is the minimum allowed value of MAX_PROGRAM_ENV_PARAMETERS_ARB
5091 * for vertex programs. If the native limit is less than that it's
5092 * not very useful, and e.g. Mesa swrast returns 0, probably to
5093 * indicate it's a software implementation. */
5094 if (gl_info
->limits
.arb_vs_native_constants
< 96)
5095 vs_consts
= gl_info
->limits
.arb_vs_float_constants
;
5097 vs_consts
= min(gl_info
->limits
.arb_vs_float_constants
, gl_info
->limits
.arb_vs_native_constants
);
5099 if (gl_info
->supported
[NV_VERTEX_PROGRAM3
])
5102 TRACE("Hardware vertex shader version 3.0 enabled (NV_VERTEX_PROGRAM3)\n");
5104 else if (vs_consts
>= 256)
5106 /* Shader Model 2.0 requires at least 256 vertex shader constants */
5108 TRACE("Hardware vertex shader version 2.0 enabled (ARB_PROGRAM)\n");
5113 TRACE("Hardware vertex shader version 1.1 enabled (ARB_PROGRAM)\n");
5115 caps
->vs_version
= min(wined3d_settings
.max_sm_vs
, vs_version
);
5116 caps
->vs_uniform_count
= vs_consts
;
5120 caps
->vs_version
= 0;
5121 caps
->vs_uniform_count
= 0;
5124 caps
->hs_version
= 0;
5125 caps
->gs_version
= 0;
5127 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
5132 /* Similar as above for vertex programs, but the minimum for fragment
5133 * programs is 24. */
5134 if (gl_info
->limits
.arb_ps_native_constants
< 24)
5135 ps_consts
= gl_info
->limits
.arb_ps_float_constants
;
5137 ps_consts
= min(gl_info
->limits
.arb_ps_float_constants
, gl_info
->limits
.arb_ps_native_constants
);
5139 if (gl_info
->supported
[NV_FRAGMENT_PROGRAM2
])
5142 TRACE("Hardware pixel shader version 3.0 enabled (NV_FRAGMENT_PROGRAM2)\n");
5144 else if (ps_consts
>= 32)
5146 /* Shader Model 2.0 requires at least 32 pixel shader constants */
5148 TRACE("Hardware pixel shader version 2.0 enabled (ARB_PROGRAM)\n");
5153 TRACE("Hardware pixel shader version 1.4 enabled (ARB_PROGRAM)\n");
5155 caps
->ps_version
= min(wined3d_settings
.max_sm_ps
, ps_version
);
5156 caps
->ps_uniform_count
= ps_consts
;
5157 caps
->ps_1x_max_value
= 8.0f
;
5161 caps
->ps_version
= 0;
5162 caps
->ps_uniform_count
= 0;
5163 caps
->ps_1x_max_value
= 0.0f
;
5166 caps
->varying_count
= 0;
5167 caps
->wined3d_caps
= WINED3D_SHADER_CAP_SRGB_WRITE
;
5168 if (use_nv_clip(gl_info
))
5169 caps
->wined3d_caps
|= WINED3D_SHADER_CAP_VS_CLIPPING
;
5172 static BOOL
shader_arb_color_fixup_supported(struct color_fixup_desc fixup
)
5174 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
5176 TRACE("Checking support for color_fixup:\n");
5177 dump_color_fixup_desc(fixup
);
5180 /* We support everything except complex conversions. */
5181 if (!is_complex_fixup(fixup
))
5187 TRACE("[FAILED]\n");
5191 static void shader_arb_add_instruction_modifiers(const struct wined3d_shader_instruction
*ins
) {
5193 char write_mask
[20], regstr
[50];
5194 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
5195 BOOL is_color
= FALSE
;
5196 const struct wined3d_shader_dst_param
*dst
;
5198 if (!ins
->dst_count
) return;
5202 if (!shift
) return; /* Saturate alone is handled by the instructions */
5204 shader_arb_get_write_mask(ins
, dst
, write_mask
);
5205 shader_arb_get_register_name(ins
, &dst
->reg
, regstr
, &is_color
);
5207 /* Generate a line that does the output modifier computation
5208 * FIXME: _SAT vs shift? _SAT alone is already handled in the instructions, if this
5209 * maps problems in e.g. _d4_sat modify shader_arb_get_modifier
5211 shader_addline(buffer
, "MUL%s %s%s, %s, %s;\n", shader_arb_get_modifier(ins
),
5212 regstr
, write_mask
, regstr
, shift_tab
[shift
]);
5215 static const SHADER_HANDLER shader_arb_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
5217 /* WINED3DSIH_ABS */ shader_hw_map2gl
,
5218 /* WINED3DSIH_ADD */ shader_hw_map2gl
,
5219 /* WINED3DSIH_AND */ NULL
,
5220 /* WINED3DSIH_BEM */ pshader_hw_bem
,
5221 /* WINED3DSIH_BREAK */ shader_hw_break
,
5222 /* WINED3DSIH_BREAKC */ shader_hw_breakc
,
5223 /* WINED3DSIH_BREAKP */ NULL
,
5224 /* WINED3DSIH_CALL */ shader_hw_call
,
5225 /* WINED3DSIH_CALLNZ */ NULL
,
5226 /* WINED3DSIH_CMP */ pshader_hw_cmp
,
5227 /* WINED3DSIH_CND */ pshader_hw_cnd
,
5228 /* WINED3DSIH_CRS */ shader_hw_map2gl
,
5229 /* WINED3DSIH_CUT */ NULL
,
5230 /* WINED3DSIH_DCL */ shader_hw_nop
,
5231 /* WINED3DSIH_DCL_CONSTANT_BUFFER */ shader_hw_nop
,
5232 /* WINED3DSIH_DCL_GLOBAL_FLAGS */ NULL
,
5233 /* WINED3DSIH_DCL_IMMEDIATE_CONSTANT_BUFFER */ NULL
,
5234 /* WINED3DSIH_DCL_INPUT */ NULL
,
5235 /* WINED3DSIH_DCL_INPUT_CONTROL_POINT_COUNT */ NULL
,
5236 /* WINED3DSIH_DCL_INPUT_PRIMITIVE */ shader_hw_nop
,
5237 /* WINED3DSIH_DCL_INPUT_PS */ NULL
,
5238 /* WINED3DSIH_DCL_INPUT_PS_SGV */ NULL
,
5239 /* WINED3DSIH_DCL_INPUT_PS_SIV */ NULL
,
5240 /* WINED3DSIH_DCL_INPUT_SGV */ NULL
,
5241 /* WINED3DSIH_DCL_INPUT_SIV */ NULL
,
5242 /* WINED3DSIH_DCL_OUTPUT */ NULL
,
5243 /* WINED3DSIH_DCL_OUTPUT_CONTROL_POINT_COUNT */ NULL
,
5244 /* WINED3DSIH_DCL_OUTPUT_SIV */ NULL
,
5245 /* WINED3DSIH_DCL_OUTPUT_TOPOLOGY */ shader_hw_nop
,
5246 /* WINED3DSIH_DCL_RESOURCE_STRUCTURED */ NULL
,
5247 /* WINED3DSIH_DCL_SAMPLER */ NULL
,
5248 /* WINED3DSIH_DCL_TEMPS */ NULL
,
5249 /* WINED3DSIH_DCL_UAV_TYPED */ NULL
,
5250 /* WINED3DSIH_DCL_VERTICES_OUT */ shader_hw_nop
,
5251 /* WINED3DSIH_DEF */ shader_hw_nop
,
5252 /* WINED3DSIH_DEFB */ shader_hw_nop
,
5253 /* WINED3DSIH_DEFI */ shader_hw_nop
,
5254 /* WINED3DSIH_DIV */ NULL
,
5255 /* WINED3DSIH_DP2 */ NULL
,
5256 /* WINED3DSIH_DP2ADD */ pshader_hw_dp2add
,
5257 /* WINED3DSIH_DP3 */ shader_hw_map2gl
,
5258 /* WINED3DSIH_DP4 */ shader_hw_map2gl
,
5259 /* WINED3DSIH_DST */ shader_hw_map2gl
,
5260 /* WINED3DSIH_DSX */ shader_hw_map2gl
,
5261 /* WINED3DSIH_DSX_COARSE */ NULL
,
5262 /* WINED3DSIH_DSX_FINE */ NULL
,
5263 /* WINED3DSIH_DSY */ shader_hw_dsy
,
5264 /* WINED3DSIH_DSY_COARSE */ NULL
,
5265 /* WINED3DSIH_DSY_FINE */ NULL
,
5266 /* WINED3DSIH_ELSE */ shader_hw_else
,
5267 /* WINED3DSIH_EMIT */ NULL
,
5268 /* WINED3DSIH_ENDIF */ shader_hw_endif
,
5269 /* WINED3DSIH_ENDLOOP */ shader_hw_endloop
,
5270 /* WINED3DSIH_ENDREP */ shader_hw_endrep
,
5271 /* WINED3DSIH_EQ */ NULL
,
5272 /* WINED3DSIH_EXP */ shader_hw_scalar_op
,
5273 /* WINED3DSIH_EXPP */ shader_hw_scalar_op
,
5274 /* WINED3DSIH_FRC */ shader_hw_map2gl
,
5275 /* WINED3DSIH_FTOI */ NULL
,
5276 /* WINED3DSIH_FTOU */ NULL
,
5277 /* WINED3DSIH_GE */ NULL
,
5278 /* WINED3DSIH_HS_DECLS */ NULL
,
5279 /* WINED3DSIH_HS_FORK_PHASE */ NULL
,
5280 /* WINED3DSIH_IADD */ NULL
,
5281 /* WINED3DSIH_IEQ */ NULL
,
5282 /* WINED3DSIH_IF */ NULL
/* Hardcoded into the shader */,
5283 /* WINED3DSIH_IFC */ shader_hw_ifc
,
5284 /* WINED3DSIH_IGE */ NULL
,
5285 /* WINED3DSIH_ILT */ NULL
,
5286 /* WINED3DSIH_IMAD */ NULL
,
5287 /* WINED3DSIH_IMAX */ NULL
,
5288 /* WINED3DSIH_IMIN */ NULL
,
5289 /* WINED3DSIH_IMUL */ NULL
,
5290 /* WINED3DSIH_INE */ NULL
,
5291 /* WINED3DSIH_INEG */ NULL
,
5292 /* WINED3DSIH_ISHL */ NULL
,
5293 /* WINED3DSIH_ITOF */ NULL
,
5294 /* WINED3DSIH_LABEL */ shader_hw_label
,
5295 /* WINED3DSIH_LD */ NULL
,
5296 /* WINED3DSIH_LD2DMS */ NULL
,
5297 /* WINED3DSIH_LD_STRUCTURED */ NULL
,
5298 /* WINED3DSIH_LIT */ shader_hw_map2gl
,
5299 /* WINED3DSIH_LOG */ shader_hw_scalar_op
,
5300 /* WINED3DSIH_LOGP */ shader_hw_scalar_op
,
5301 /* WINED3DSIH_LOOP */ shader_hw_loop
,
5302 /* WINED3DSIH_LRP */ shader_hw_lrp
,
5303 /* WINED3DSIH_LT */ NULL
,
5304 /* WINED3DSIH_M3x2 */ shader_hw_mnxn
,
5305 /* WINED3DSIH_M3x3 */ shader_hw_mnxn
,
5306 /* WINED3DSIH_M3x4 */ shader_hw_mnxn
,
5307 /* WINED3DSIH_M4x3 */ shader_hw_mnxn
,
5308 /* WINED3DSIH_M4x4 */ shader_hw_mnxn
,
5309 /* WINED3DSIH_MAD */ shader_hw_map2gl
,
5310 /* WINED3DSIH_MAX */ shader_hw_map2gl
,
5311 /* WINED3DSIH_MIN */ shader_hw_map2gl
,
5312 /* WINED3DSIH_MOV */ shader_hw_mov
,
5313 /* WINED3DSIH_MOVA */ shader_hw_mov
,
5314 /* WINED3DSIH_MOVC */ NULL
,
5315 /* WINED3DSIH_MUL */ shader_hw_map2gl
,
5316 /* WINED3DSIH_NE */ NULL
,
5317 /* WINED3DSIH_NOP */ shader_hw_nop
,
5318 /* WINED3DSIH_NOT */ NULL
,
5319 /* WINED3DSIH_NRM */ shader_hw_nrm
,
5320 /* WINED3DSIH_OR */ NULL
,
5321 /* WINED3DSIH_PHASE */ shader_hw_nop
,
5322 /* WINED3DSIH_POW */ shader_hw_pow
,
5323 /* WINED3DSIH_RCP */ shader_hw_scalar_op
,
5324 /* WINED3DSIH_REP */ shader_hw_rep
,
5325 /* WINED3DSIH_RESINFO */ NULL
,
5326 /* WINED3DSIH_RET */ shader_hw_ret
,
5327 /* WINED3DSIH_ROUND_NI */ NULL
,
5328 /* WINED3DSIH_ROUND_PI */ NULL
,
5329 /* WINED3DSIH_ROUND_Z */ NULL
,
5330 /* WINED3DSIH_RSQ */ shader_hw_scalar_op
,
5331 /* WINED3DSIH_SAMPLE */ NULL
,
5332 /* WINED3DSIH_SAMPLE_B */ NULL
,
5333 /* WINED3DSIH_SAMPLE_C */ NULL
,
5334 /* WINED3DSIH_SAMPLE_C_LZ */ NULL
,
5335 /* WINED3DSIH_SAMPLE_GRAD */ NULL
,
5336 /* WINED3DSIH_SAMPLE_LOD */ NULL
,
5337 /* WINED3DSIH_SETP */ NULL
,
5338 /* WINED3DSIH_SGE */ shader_hw_map2gl
,
5339 /* WINED3DSIH_SGN */ shader_hw_sgn
,
5340 /* WINED3DSIH_SINCOS */ shader_hw_sincos
,
5341 /* WINED3DSIH_SLT */ shader_hw_map2gl
,
5342 /* WINED3DSIH_SQRT */ NULL
,
5343 /* WINED3DSIH_STORE_UAV_TYPED */ NULL
,
5344 /* WINED3DSIH_SUB */ shader_hw_map2gl
,
5345 /* WINED3DSIH_TEX */ pshader_hw_tex
,
5346 /* WINED3DSIH_TEXBEM */ pshader_hw_texbem
,
5347 /* WINED3DSIH_TEXBEML */ pshader_hw_texbem
,
5348 /* WINED3DSIH_TEXCOORD */ pshader_hw_texcoord
,
5349 /* WINED3DSIH_TEXDEPTH */ pshader_hw_texdepth
,
5350 /* WINED3DSIH_TEXDP3 */ pshader_hw_texdp3
,
5351 /* WINED3DSIH_TEXDP3TEX */ pshader_hw_texdp3tex
,
5352 /* WINED3DSIH_TEXKILL */ pshader_hw_texkill
,
5353 /* WINED3DSIH_TEXLDD */ shader_hw_texldd
,
5354 /* WINED3DSIH_TEXLDL */ shader_hw_texldl
,
5355 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_hw_texm3x2depth
,
5356 /* WINED3DSIH_TEXM3x2PAD */ pshader_hw_texm3x2pad
,
5357 /* WINED3DSIH_TEXM3x2TEX */ pshader_hw_texm3x2tex
,
5358 /* WINED3DSIH_TEXM3x3 */ pshader_hw_texm3x3
,
5359 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
5360 /* WINED3DSIH_TEXM3x3PAD */ pshader_hw_texm3x3pad
,
5361 /* WINED3DSIH_TEXM3x3SPEC */ pshader_hw_texm3x3spec
,
5362 /* WINED3DSIH_TEXM3x3TEX */ pshader_hw_texm3x3tex
,
5363 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_hw_texm3x3vspec
,
5364 /* WINED3DSIH_TEXREG2AR */ pshader_hw_texreg2ar
,
5365 /* WINED3DSIH_TEXREG2GB */ pshader_hw_texreg2gb
,
5366 /* WINED3DSIH_TEXREG2RGB */ pshader_hw_texreg2rgb
,
5367 /* WINED3DSIH_UDIV */ NULL
,
5368 /* WINED3DSIH_UGE */ NULL
,
5369 /* WINED3DSIH_USHR */ NULL
,
5370 /* WINED3DSIH_UTOF */ NULL
,
5371 /* WINED3DSIH_XOR */ NULL
,
5374 static BOOL
get_bool_const(const struct wined3d_shader_instruction
*ins
,
5375 const struct wined3d_shader
*shader
, DWORD idx
)
5377 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
5378 BOOL vshader
= shader_is_vshader_version(reg_maps
->shader_version
.type
);
5379 const struct wined3d_shader_lconst
*constant
;
5381 WORD flag
= (1u << idx
);
5382 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5384 if (reg_maps
->local_bool_consts
& flag
)
5386 /* What good is an if(bool) with a hardcoded local constant? I don't know, but handle it */
5387 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsB
, struct wined3d_shader_lconst
, entry
)
5389 if (constant
->idx
== idx
)
5391 return constant
->value
[0];
5394 ERR("Local constant not found\n");
5399 if(vshader
) bools
= priv
->cur_vs_args
->clip
.boolclip
.bools
;
5400 else bools
= priv
->cur_ps_args
->bools
;
5401 return bools
& flag
;
5405 static void get_loop_control_const(const struct wined3d_shader_instruction
*ins
,
5406 const struct wined3d_shader
*shader
, UINT idx
, struct wined3d_shader_loop_control
*loop_control
)
5408 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
5409 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5411 /* Integer constants can either be a local constant, or they can be stored in the shader
5412 * type specific compile args. */
5413 if (reg_maps
->local_int_consts
& (1u << idx
))
5415 const struct wined3d_shader_lconst
*constant
;
5417 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsI
, struct wined3d_shader_lconst
, entry
)
5419 if (constant
->idx
== idx
)
5421 loop_control
->count
= constant
->value
[0];
5422 loop_control
->start
= constant
->value
[1];
5423 /* Step is signed. */
5424 loop_control
->step
= (int)constant
->value
[2];
5428 /* If this happens the flag was set incorrectly */
5429 ERR("Local constant not found\n");
5430 loop_control
->count
= 0;
5431 loop_control
->start
= 0;
5432 loop_control
->step
= 0;
5436 switch (reg_maps
->shader_version
.type
)
5438 case WINED3D_SHADER_TYPE_VERTEX
:
5439 /* Count and aL start value are unsigned */
5440 loop_control
->count
= priv
->cur_vs_args
->loop_ctrl
[idx
][0];
5441 loop_control
->start
= priv
->cur_vs_args
->loop_ctrl
[idx
][1];
5442 /* Step is signed. */
5443 loop_control
->step
= ((char)priv
->cur_vs_args
->loop_ctrl
[idx
][2]);
5446 case WINED3D_SHADER_TYPE_PIXEL
:
5447 loop_control
->count
= priv
->cur_ps_args
->loop_ctrl
[idx
][0];
5448 loop_control
->start
= priv
->cur_ps_args
->loop_ctrl
[idx
][1];
5449 loop_control
->step
= ((char)priv
->cur_ps_args
->loop_ctrl
[idx
][2]);
5453 FIXME("Unhandled shader type %#x.\n", reg_maps
->shader_version
.type
);
5458 static void record_instruction(struct list
*list
, const struct wined3d_shader_instruction
*ins
)
5461 struct wined3d_shader_dst_param
*dst_param
;
5462 struct wined3d_shader_src_param
*src_param
= NULL
, *rel_addr
;
5463 struct recorded_instruction
*rec
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*rec
));
5466 ERR("Out of memory\n");
5471 dst_param
= HeapAlloc(GetProcessHeap(), 0, sizeof(*dst_param
));
5472 if(!dst_param
) goto free
;
5473 *dst_param
= *ins
->dst
;
5474 if (ins
->dst
->reg
.idx
[0].rel_addr
)
5476 rel_addr
= HeapAlloc(GetProcessHeap(), 0, sizeof(*rel_addr
));
5479 *rel_addr
= *ins
->dst
->reg
.idx
[0].rel_addr
;
5480 dst_param
->reg
.idx
[0].rel_addr
= rel_addr
;
5482 rec
->ins
.dst
= dst_param
;
5484 src_param
= HeapAlloc(GetProcessHeap(), 0, sizeof(*src_param
) * ins
->src_count
);
5487 for (i
= 0; i
< ins
->src_count
; ++i
)
5489 src_param
[i
] = ins
->src
[i
];
5490 if (ins
->src
[i
].reg
.idx
[0].rel_addr
)
5492 rel_addr
= HeapAlloc(GetProcessHeap(), 0, sizeof(*rel_addr
));
5495 *rel_addr
= *ins
->src
[i
].reg
.idx
[0].rel_addr
;
5496 src_param
[i
].reg
.idx
[0].rel_addr
= rel_addr
;
5499 rec
->ins
.src
= src_param
;
5500 list_add_tail(list
, &rec
->entry
);
5504 ERR("Out of memory\n");
5507 HeapFree(GetProcessHeap(), 0, (void *)dst_param
->reg
.idx
[0].rel_addr
);
5508 HeapFree(GetProcessHeap(), 0, dst_param
);
5512 for(i
= 0; i
< ins
->src_count
; i
++)
5514 HeapFree(GetProcessHeap(), 0, (void *)src_param
[i
].reg
.idx
[0].rel_addr
);
5516 HeapFree(GetProcessHeap(), 0, src_param
);
5518 HeapFree(GetProcessHeap(), 0, rec
);
5521 static void free_recorded_instruction(struct list
*list
)
5523 struct recorded_instruction
*rec_ins
, *entry2
;
5526 LIST_FOR_EACH_ENTRY_SAFE(rec_ins
, entry2
, list
, struct recorded_instruction
, entry
)
5528 list_remove(&rec_ins
->entry
);
5529 if (rec_ins
->ins
.dst
)
5531 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.dst
->reg
.idx
[0].rel_addr
);
5532 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.dst
);
5534 if (rec_ins
->ins
.src
)
5536 for (i
= 0; i
< rec_ins
->ins
.src_count
; ++i
)
5538 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.src
[i
].reg
.idx
[0].rel_addr
);
5540 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.src
);
5542 HeapFree(GetProcessHeap(), 0, rec_ins
);
5546 static void pop_control_frame(const struct wined3d_shader_instruction
*ins
)
5548 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5549 struct control_frame
*control_frame
;
5551 if (ins
->handler_idx
== WINED3DSIH_ENDLOOP
|| ins
->handler_idx
== WINED3DSIH_ENDREP
)
5553 struct list
*e
= list_head(&priv
->control_frames
);
5554 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5555 list_remove(&control_frame
->entry
);
5556 HeapFree(GetProcessHeap(), 0, control_frame
);
5559 else if (ins
->handler_idx
== WINED3DSIH_ENDIF
)
5561 /* Non-ifc ENDIFs were already handled previously. */
5562 struct list
*e
= list_head(&priv
->control_frames
);
5563 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5564 list_remove(&control_frame
->entry
);
5565 HeapFree(GetProcessHeap(), 0, control_frame
);
5569 static void shader_arb_handle_instruction(const struct wined3d_shader_instruction
*ins
) {
5570 SHADER_HANDLER hw_fct
;
5571 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5572 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
5573 struct control_frame
*control_frame
;
5574 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
5577 if(ins
->handler_idx
== WINED3DSIH_LOOP
|| ins
->handler_idx
== WINED3DSIH_REP
)
5579 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
5580 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5582 if(ins
->handler_idx
== WINED3DSIH_LOOP
) control_frame
->type
= LOOP
;
5583 if(ins
->handler_idx
== WINED3DSIH_REP
) control_frame
->type
= REP
;
5585 if(priv
->target_version
>= NV2
)
5587 control_frame
->no
.loop
= priv
->num_loops
++;
5592 /* Don't bother recording when we're in a not used if branch */
5598 if(!priv
->recording
)
5600 list_init(&priv
->record
);
5601 priv
->recording
= TRUE
;
5602 control_frame
->outer_loop
= TRUE
;
5603 get_loop_control_const(ins
, shader
, ins
->src
[0].reg
.idx
[0].offset
, &control_frame
->loop_control
);
5604 return; /* Instruction is handled */
5606 /* Record this loop in the outer loop's recording */
5609 else if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
|| ins
->handler_idx
== WINED3DSIH_ENDREP
)
5611 if(priv
->target_version
>= NV2
)
5613 /* Nothing to do. The control frame is popped after the HW instr handler */
5617 struct list
*e
= list_head(&priv
->control_frames
);
5618 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5619 list_remove(&control_frame
->entry
);
5621 if(control_frame
->outer_loop
)
5623 unsigned int iteration
;
5627 /* Turn off recording before playback */
5628 priv
->recording
= FALSE
;
5630 /* Move the recorded instructions to a separate list and get them out of the private data
5631 * structure. If there are nested loops, the shader_arb_handle_instruction below will
5632 * be recorded again, thus priv->record might be overwritten
5635 list_move_tail(©
, &priv
->record
);
5636 list_init(&priv
->record
);
5638 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5640 shader_addline(buffer
, "#unrolling loop: %u iterations, aL=%u, inc %d\n",
5641 control_frame
->loop_control
.count
, control_frame
->loop_control
.start
,
5642 control_frame
->loop_control
.step
);
5643 aL
= control_frame
->loop_control
.start
;
5647 shader_addline(buffer
, "#unrolling rep: %u iterations\n", control_frame
->loop_control
.count
);
5650 for (iteration
= 0; iteration
< control_frame
->loop_control
.count
; ++iteration
)
5652 struct recorded_instruction
*rec_ins
;
5653 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5656 shader_addline(buffer
, "#Iteration %u, aL=%d\n", iteration
, aL
);
5660 shader_addline(buffer
, "#Iteration %u\n", iteration
);
5663 LIST_FOR_EACH_ENTRY(rec_ins
, ©
, struct recorded_instruction
, entry
)
5665 shader_arb_handle_instruction(&rec_ins
->ins
);
5668 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5670 aL
+= control_frame
->loop_control
.step
;
5673 shader_addline(buffer
, "#end loop/rep\n");
5675 free_recorded_instruction(©
);
5676 HeapFree(GetProcessHeap(), 0, control_frame
);
5677 return; /* Instruction is handled */
5681 /* This is a nested loop. Proceed to the normal recording function */
5682 HeapFree(GetProcessHeap(), 0, control_frame
);
5689 record_instruction(&priv
->record
, ins
);
5694 if(ins
->handler_idx
== WINED3DSIH_IF
)
5696 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
5697 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5698 control_frame
->type
= IF
;
5700 bool_const
= get_bool_const(ins
, shader
, ins
->src
[0].reg
.idx
[0].offset
);
5701 if (ins
->src
[0].modifiers
== WINED3DSPSM_NOT
)
5702 bool_const
= !bool_const
;
5703 if (!priv
->muted
&& !bool_const
)
5705 shader_addline(buffer
, "#if(FALSE){\n");
5707 control_frame
->muting
= TRUE
;
5709 else shader_addline(buffer
, "#if(TRUE) {\n");
5711 return; /* Instruction is handled */
5713 else if(ins
->handler_idx
== WINED3DSIH_IFC
)
5715 /* IF(bool) and if_cond(a, b) use the same ELSE and ENDIF tokens */
5716 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
5717 control_frame
->type
= IFC
;
5718 control_frame
->no
.ifc
= priv
->num_ifcs
++;
5719 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5721 else if(ins
->handler_idx
== WINED3DSIH_ELSE
)
5723 struct list
*e
= list_head(&priv
->control_frames
);
5724 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5726 if(control_frame
->type
== IF
)
5728 shader_addline(buffer
, "#} else {\n");
5729 if(!priv
->muted
&& !control_frame
->muting
)
5732 control_frame
->muting
= TRUE
;
5734 else if(control_frame
->muting
) priv
->muted
= FALSE
;
5735 return; /* Instruction is handled. */
5737 /* In case of an ifc, generate a HW shader instruction */
5738 if (control_frame
->type
!= IFC
)
5739 ERR("Control frame does not match.\n");
5741 else if(ins
->handler_idx
== WINED3DSIH_ENDIF
)
5743 struct list
*e
= list_head(&priv
->control_frames
);
5744 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5746 if(control_frame
->type
== IF
)
5748 shader_addline(buffer
, "#} endif\n");
5749 if(control_frame
->muting
) priv
->muted
= FALSE
;
5750 list_remove(&control_frame
->entry
);
5751 HeapFree(GetProcessHeap(), 0, control_frame
);
5752 return; /* Instruction is handled */
5754 /* In case of an ifc, generate a HW shader instruction */
5755 if (control_frame
->type
!= IFC
)
5756 ERR("Control frame does not match.\n");
5761 pop_control_frame(ins
);
5765 /* Select handler */
5766 hw_fct
= shader_arb_instruction_handler_table
[ins
->handler_idx
];
5768 /* Unhandled opcode */
5771 FIXME("Backend can't handle opcode %s.\n", debug_d3dshaderinstructionhandler(ins
->handler_idx
));
5776 pop_control_frame(ins
);
5778 shader_arb_add_instruction_modifiers(ins
);
5781 static BOOL
shader_arb_has_ffp_proj_control(void *shader_priv
)
5783 struct shader_arb_priv
*priv
= shader_priv
;
5785 return priv
->ffp_proj_control
;
5788 const struct wined3d_shader_backend_ops arb_program_shader_backend
=
5790 shader_arb_handle_instruction
,
5793 shader_arb_select_depth_blt
,
5794 shader_arb_deselect_depth_blt
,
5795 shader_arb_update_float_vertex_constants
,
5796 shader_arb_update_float_pixel_constants
,
5797 shader_arb_load_constants
,
5801 shader_arb_allocate_context_data
,
5802 shader_arb_free_context_data
,
5803 shader_arb_init_context_state
,
5804 shader_arb_get_caps
,
5805 shader_arb_color_fixup_supported
,
5806 shader_arb_has_ffp_proj_control
,
5809 /* ARB_fragment_program fixed function pipeline replacement definitions */
5810 #define ARB_FFP_CONST_TFACTOR 0
5811 #define ARB_FFP_CONST_COLOR_KEY_LOW ((ARB_FFP_CONST_TFACTOR) + 1)
5812 #define ARB_FFP_CONST_COLOR_KEY_HIGH ((ARB_FFP_CONST_COLOR_KEY_LOW) + 1)
5813 #define ARB_FFP_CONST_SPECULAR_ENABLE ((ARB_FFP_CONST_COLOR_KEY_HIGH) + 1)
5814 #define ARB_FFP_CONST_CONSTANT(i) ((ARB_FFP_CONST_SPECULAR_ENABLE) + 1 + i)
5815 #define ARB_FFP_CONST_BUMPMAT(i) ((ARB_FFP_CONST_CONSTANT(7)) + 1 + i)
5816 #define ARB_FFP_CONST_LUMINANCE(i) ((ARB_FFP_CONST_BUMPMAT(7)) + 1 + i)
5818 struct arbfp_ffp_desc
5820 struct ffp_frag_desc parent
;
5824 /* Context activation is done by the caller. */
5825 static void arbfp_enable(const struct wined3d_gl_info
*gl_info
, BOOL enable
)
5829 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
5830 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
5834 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
5835 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
5839 static void *arbfp_alloc(const struct wined3d_shader_backend_ops
*shader_backend
, void *shader_priv
)
5841 struct shader_arb_priv
*priv
;
5843 /* Share private data between the shader backend and the pipeline
5844 * replacement, if both are the arb implementation. This is needed to
5845 * figure out whether ARBfp should be disabled if no pixel shader is bound
5847 if (shader_backend
== &arb_program_shader_backend
)
5849 else if (!(priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*priv
))))
5852 if (wine_rb_init(&priv
->fragment_shaders
, &wined3d_ffp_frag_program_rb_functions
) == -1)
5854 ERR("Failed to initialize rbtree.\n");
5855 if (priv
!= shader_priv
)
5856 HeapFree(GetProcessHeap(), 0, priv
);
5859 priv
->use_arbfp_fixed_func
= TRUE
;
5864 /* Context activation is done by the caller. */
5865 static void arbfp_free_ffpshader(struct wine_rb_entry
*entry
, void *context
)
5867 const struct wined3d_gl_info
*gl_info
= context
;
5868 struct arbfp_ffp_desc
*entry_arb
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_ffp_desc
, parent
.entry
);
5870 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb
->shader
));
5871 checkGLcall("glDeleteProgramsARB(1, &entry_arb->shader)");
5872 HeapFree(GetProcessHeap(), 0, entry_arb
);
5875 /* Context activation is done by the caller. */
5876 static void arbfp_free(struct wined3d_device
*device
)
5878 struct shader_arb_priv
*priv
= device
->fragment_priv
;
5880 wine_rb_destroy(&priv
->fragment_shaders
, arbfp_free_ffpshader
, &device
->adapter
->gl_info
);
5881 priv
->use_arbfp_fixed_func
= FALSE
;
5883 if (device
->shader_backend
!= &arb_program_shader_backend
)
5885 HeapFree(GetProcessHeap(), 0, device
->fragment_priv
);
5889 static void arbfp_get_caps(const struct wined3d_gl_info
*gl_info
, struct fragment_caps
*caps
)
5891 caps
->wined3d_caps
= WINED3D_FRAGMENT_CAP_PROJ_CONTROL
5892 | WINED3D_FRAGMENT_CAP_SRGB_WRITE
5893 | WINED3D_FRAGMENT_CAP_COLOR_KEY
;
5894 caps
->PrimitiveMiscCaps
= WINED3DPMISCCAPS_TSSARGTEMP
;
5895 caps
->TextureOpCaps
= WINED3DTEXOPCAPS_DISABLE
|
5896 WINED3DTEXOPCAPS_SELECTARG1
|
5897 WINED3DTEXOPCAPS_SELECTARG2
|
5898 WINED3DTEXOPCAPS_MODULATE4X
|
5899 WINED3DTEXOPCAPS_MODULATE2X
|
5900 WINED3DTEXOPCAPS_MODULATE
|
5901 WINED3DTEXOPCAPS_ADDSIGNED2X
|
5902 WINED3DTEXOPCAPS_ADDSIGNED
|
5903 WINED3DTEXOPCAPS_ADD
|
5904 WINED3DTEXOPCAPS_SUBTRACT
|
5905 WINED3DTEXOPCAPS_ADDSMOOTH
|
5906 WINED3DTEXOPCAPS_BLENDCURRENTALPHA
|
5907 WINED3DTEXOPCAPS_BLENDFACTORALPHA
|
5908 WINED3DTEXOPCAPS_BLENDTEXTUREALPHA
|
5909 WINED3DTEXOPCAPS_BLENDDIFFUSEALPHA
|
5910 WINED3DTEXOPCAPS_BLENDTEXTUREALPHAPM
|
5911 WINED3DTEXOPCAPS_MODULATEALPHA_ADDCOLOR
|
5912 WINED3DTEXOPCAPS_MODULATECOLOR_ADDALPHA
|
5913 WINED3DTEXOPCAPS_MODULATEINVCOLOR_ADDALPHA
|
5914 WINED3DTEXOPCAPS_MODULATEINVALPHA_ADDCOLOR
|
5915 WINED3DTEXOPCAPS_DOTPRODUCT3
|
5916 WINED3DTEXOPCAPS_MULTIPLYADD
|
5917 WINED3DTEXOPCAPS_LERP
|
5918 WINED3DTEXOPCAPS_BUMPENVMAP
|
5919 WINED3DTEXOPCAPS_BUMPENVMAPLUMINANCE
;
5921 /* TODO: Implement WINED3DTEXOPCAPS_PREMODULATE */
5923 caps
->MaxTextureBlendStages
= 8;
5924 caps
->MaxSimultaneousTextures
= min(gl_info
->limits
.fragment_samplers
, 8);
5927 static DWORD
arbfp_get_emul_mask(const struct wined3d_gl_info
*gl_info
)
5929 return GL_EXT_EMUL_ARB_MULTITEXTURE
| GL_EXT_EMUL_EXT_FOG_COORD
;
5932 static void state_texfactor_arbfp(struct wined3d_context
*context
,
5933 const struct wined3d_state
*state
, DWORD state_id
)
5935 struct wined3d_device
*device
= context
->swapchain
->device
;
5936 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5937 struct wined3d_color color
;
5939 /* Don't load the parameter if we're using an arbfp pixel shader,
5940 * otherwise we'll overwrite application provided constants. */
5941 if (device
->shader_backend
== &arb_program_shader_backend
)
5943 struct shader_arb_priv
*priv
;
5945 if (use_ps(state
)) return;
5947 priv
= device
->shader_priv
;
5948 priv
->pshader_const_dirty
[ARB_FFP_CONST_TFACTOR
] = 1;
5949 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_TFACTOR
+ 1);
5952 wined3d_color_from_d3dcolor(&color
, state
->render_states
[WINED3D_RS_TEXTUREFACTOR
]);
5953 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_TFACTOR
, &color
.r
));
5954 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_TFACTOR, &color.r)");
5957 static void state_arb_specularenable(struct wined3d_context
*context
,
5958 const struct wined3d_state
*state
, DWORD state_id
)
5960 struct wined3d_device
*device
= context
->swapchain
->device
;
5961 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5964 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
5965 * application provided constants
5967 if (device
->shader_backend
== &arb_program_shader_backend
)
5969 struct shader_arb_priv
*priv
;
5971 if (use_ps(state
)) return;
5973 priv
= device
->shader_priv
;
5974 priv
->pshader_const_dirty
[ARB_FFP_CONST_SPECULAR_ENABLE
] = 1;
5975 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_SPECULAR_ENABLE
+ 1);
5978 if (state
->render_states
[WINED3D_RS_SPECULARENABLE
])
5980 /* The specular color has no alpha */
5981 col
[0] = 1.0f
; col
[1] = 1.0f
;
5982 col
[2] = 1.0f
; col
[3] = 0.0f
;
5984 col
[0] = 0.0f
; col
[1] = 0.0f
;
5985 col
[2] = 0.0f
; col
[3] = 0.0f
;
5987 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_SPECULAR_ENABLE
, col
));
5988 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_SPECULAR_ENABLE, col)");
5991 static void set_bumpmat_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
5993 DWORD stage
= (state_id
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
5994 struct wined3d_device
*device
= context
->swapchain
->device
;
5995 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5998 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_BUMP_ENV
;
6000 if (device
->shader_backend
== &arb_program_shader_backend
)
6002 struct shader_arb_priv
*priv
= device
->shader_priv
;
6004 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants. */
6008 priv
->pshader_const_dirty
[ARB_FFP_CONST_BUMPMAT(stage
)] = 1;
6009 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_BUMPMAT(stage
) + 1);
6012 mat
[0][0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT00
]);
6013 mat
[0][1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT01
]);
6014 mat
[1][0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT10
]);
6015 mat
[1][1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT11
]);
6017 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_BUMPMAT(stage
), &mat
[0][0]));
6018 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_BUMPMAT(stage), &mat[0][0])");
6021 static void tex_bumpenvlum_arbfp(struct wined3d_context
*context
,
6022 const struct wined3d_state
*state
, DWORD state_id
)
6024 DWORD stage
= (state_id
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
6025 struct wined3d_device
*device
= context
->swapchain
->device
;
6026 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
6029 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_BUMP_ENV
;
6031 if (device
->shader_backend
== &arb_program_shader_backend
)
6033 struct shader_arb_priv
*priv
= device
->shader_priv
;
6035 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants. */
6039 priv
->pshader_const_dirty
[ARB_FFP_CONST_LUMINANCE(stage
)] = 1;
6040 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_LUMINANCE(stage
) + 1);
6043 param
[0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_LSCALE
]);
6044 param
[1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_LOFFSET
]);
6048 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_LUMINANCE(stage
), param
));
6049 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_LUMINANCE(stage), param)");
6052 static void alpha_test_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6054 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
6058 TRACE("context %p, state %p, state_id %#x.\n", context
, state
, state_id
);
6060 if (state
->render_states
[WINED3D_RS_ALPHATESTENABLE
])
6062 gl_info
->gl_ops
.gl
.p_glEnable(GL_ALPHA_TEST
);
6063 checkGLcall("glEnable GL_ALPHA_TEST");
6067 gl_info
->gl_ops
.gl
.p_glDisable(GL_ALPHA_TEST
);
6068 checkGLcall("glDisable GL_ALPHA_TEST");
6072 ref
= ((float)state
->render_states
[WINED3D_RS_ALPHAREF
]) / 255.0f
;
6073 glParm
= wined3d_gl_compare_func(state
->render_states
[WINED3D_RS_ALPHAFUNC
]);
6077 gl_info
->gl_ops
.gl
.p_glAlphaFunc(glParm
, ref
);
6078 checkGLcall("glAlphaFunc");
6082 static void color_key_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6084 struct wined3d_device
*device
= context
->swapchain
->device
;
6085 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
6086 struct wined3d_color float_key
[2];
6087 const struct wined3d_texture
*texture
= state
->textures
[0];
6092 /* Don't load the parameter if we're using an arbfp pixel shader,
6093 * otherwise we'll overwrite application provided constants. */
6094 if (device
->shader_backend
== &arb_program_shader_backend
)
6096 struct shader_arb_priv
*priv
;
6101 priv
= device
->shader_priv
;
6102 priv
->pshader_const_dirty
[ARB_FFP_CONST_COLOR_KEY_LOW
] = 1;
6103 priv
->pshader_const_dirty
[ARB_FFP_CONST_COLOR_KEY_HIGH
] = 1;
6104 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_COLOR_KEY_HIGH
+ 1);
6107 wined3d_format_get_float_color_key(texture
->resource
.format
, &texture
->async
.src_blt_color_key
, float_key
);
6109 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_COLOR_KEY_LOW
, &float_key
[0].r
));
6110 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_COLOR_KEY_LOW, &float_key[0].r)");
6111 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_COLOR_KEY_HIGH
, &float_key
[1].r
));
6112 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_COLOR_KEY_HIGH, &float_key[1].r)");
6115 static const char *get_argreg(struct wined3d_string_buffer
*buffer
, DWORD argnum
, unsigned int stage
, DWORD arg
)
6119 if(arg
== ARG_UNUSED
) return "unused"; /* This is the marker for unused registers */
6121 switch(arg
& WINED3DTA_SELECTMASK
) {
6122 case WINED3DTA_DIFFUSE
:
6123 ret
= "fragment.color.primary"; break;
6125 case WINED3DTA_CURRENT
:
6129 case WINED3DTA_TEXTURE
:
6131 case 0: ret
= "tex0"; break;
6132 case 1: ret
= "tex1"; break;
6133 case 2: ret
= "tex2"; break;
6134 case 3: ret
= "tex3"; break;
6135 case 4: ret
= "tex4"; break;
6136 case 5: ret
= "tex5"; break;
6137 case 6: ret
= "tex6"; break;
6138 case 7: ret
= "tex7"; break;
6139 default: ret
= "unknown texture";
6143 case WINED3DTA_TFACTOR
:
6144 ret
= "tfactor"; break;
6146 case WINED3DTA_SPECULAR
:
6147 ret
= "fragment.color.secondary"; break;
6149 case WINED3DTA_TEMP
:
6150 ret
= "tempreg"; break;
6152 case WINED3DTA_CONSTANT
:
6153 FIXME("Implement perstage constants\n");
6155 case 0: ret
= "const0"; break;
6156 case 1: ret
= "const1"; break;
6157 case 2: ret
= "const2"; break;
6158 case 3: ret
= "const3"; break;
6159 case 4: ret
= "const4"; break;
6160 case 5: ret
= "const5"; break;
6161 case 6: ret
= "const6"; break;
6162 case 7: ret
= "const7"; break;
6163 default: ret
= "unknown constant";
6171 if(arg
& WINED3DTA_COMPLEMENT
) {
6172 shader_addline(buffer
, "SUB arg%u, const.x, %s;\n", argnum
, ret
);
6173 if(argnum
== 0) ret
= "arg0";
6174 if(argnum
== 1) ret
= "arg1";
6175 if(argnum
== 2) ret
= "arg2";
6177 if(arg
& WINED3DTA_ALPHAREPLICATE
) {
6178 shader_addline(buffer
, "MOV arg%u, %s.w;\n", argnum
, ret
);
6179 if(argnum
== 0) ret
= "arg0";
6180 if(argnum
== 1) ret
= "arg1";
6181 if(argnum
== 2) ret
= "arg2";
6186 static void gen_ffp_instr(struct wined3d_string_buffer
*buffer
, unsigned int stage
, BOOL color
,
6187 BOOL alpha
, DWORD dst
, DWORD op
, DWORD dw_arg0
, DWORD dw_arg1
, DWORD dw_arg2
)
6189 const char *dstmask
, *dstreg
, *arg0
, *arg1
, *arg2
;
6190 unsigned int mul
= 1;
6192 if(color
&& alpha
) dstmask
= "";
6193 else if(color
) dstmask
= ".xyz";
6194 else dstmask
= ".w";
6196 if(dst
== tempreg
) dstreg
= "tempreg";
6197 else dstreg
= "ret";
6199 arg0
= get_argreg(buffer
, 0, stage
, dw_arg0
);
6200 arg1
= get_argreg(buffer
, 1, stage
, dw_arg1
);
6201 arg2
= get_argreg(buffer
, 2, stage
, dw_arg2
);
6205 case WINED3D_TOP_DISABLE
:
6208 case WINED3D_TOP_SELECT_ARG2
:
6211 case WINED3D_TOP_SELECT_ARG1
:
6212 shader_addline(buffer
, "MOV %s%s, %s;\n", dstreg
, dstmask
, arg1
);
6215 case WINED3D_TOP_MODULATE_4X
:
6218 case WINED3D_TOP_MODULATE_2X
:
6221 case WINED3D_TOP_MODULATE
:
6222 shader_addline(buffer
, "MUL %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6225 case WINED3D_TOP_ADD_SIGNED_2X
:
6228 case WINED3D_TOP_ADD_SIGNED
:
6229 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
6232 case WINED3D_TOP_ADD
:
6233 shader_addline(buffer
, "ADD_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6236 case WINED3D_TOP_SUBTRACT
:
6237 shader_addline(buffer
, "SUB_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6240 case WINED3D_TOP_ADD_SMOOTH
:
6241 shader_addline(buffer
, "SUB arg1, const.x, %s;\n", arg1
);
6242 shader_addline(buffer
, "MAD_SAT %s%s, arg1, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6245 case WINED3D_TOP_BLEND_CURRENT_ALPHA
:
6246 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_CURRENT
);
6247 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6249 case WINED3D_TOP_BLEND_FACTOR_ALPHA
:
6250 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TFACTOR
);
6251 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6253 case WINED3D_TOP_BLEND_TEXTURE_ALPHA
:
6254 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
6255 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6257 case WINED3D_TOP_BLEND_DIFFUSE_ALPHA
:
6258 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_DIFFUSE
);
6259 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6262 case WINED3D_TOP_BLEND_TEXTURE_ALPHA_PM
:
6263 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
6264 shader_addline(buffer
, "SUB arg0.w, const.x, %s.w;\n", arg0
);
6265 shader_addline(buffer
, "MAD_SAT %s%s, %s, arg0.w, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6268 /* D3DTOP_PREMODULATE ???? */
6270 case WINED3D_TOP_MODULATE_INVALPHA_ADD_COLOR
:
6271 shader_addline(buffer
, "SUB arg0.w, const.x, %s;\n", arg1
);
6272 shader_addline(buffer
, "MAD_SAT %s%s, arg0.w, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6274 case WINED3D_TOP_MODULATE_ALPHA_ADD_COLOR
:
6275 shader_addline(buffer
, "MAD_SAT %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
6277 case WINED3D_TOP_MODULATE_INVCOLOR_ADD_ALPHA
:
6278 shader_addline(buffer
, "SUB arg0, const.x, %s;\n", arg1
);
6279 shader_addline(buffer
, "MAD_SAT %s%s, arg0, %s, %s.w;\n", dstreg
, dstmask
, arg2
, arg1
);
6281 case WINED3D_TOP_MODULATE_COLOR_ADD_ALPHA
:
6282 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s.w;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
6285 case WINED3D_TOP_DOTPRODUCT3
:
6287 shader_addline(buffer
, "SUB arg1, %s, const.w;\n", arg1
);
6288 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
6289 shader_addline(buffer
, "DP3_SAT %s%s, arg1, arg2;\n", dstreg
, dstmask
);
6292 case WINED3D_TOP_MULTIPLY_ADD
:
6293 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg0
);
6296 case WINED3D_TOP_LERP
:
6297 /* The msdn is not quite right here */
6298 shader_addline(buffer
, "LRP %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6301 case WINED3D_TOP_BUMPENVMAP
:
6302 case WINED3D_TOP_BUMPENVMAP_LUMINANCE
:
6303 /* Those are handled in the first pass of the shader(generation pass 1 and 2) already */
6307 FIXME("Unhandled texture op %08x\n", op
);
6311 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.y;\n", dstreg
, dstmask
, dstreg
);
6313 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.z;\n", dstreg
, dstmask
, dstreg
);
6316 static const char *arbfp_texture_target(enum wined3d_gl_resource_type type
)
6320 case WINED3D_GL_RES_TYPE_TEX_1D
:
6322 case WINED3D_GL_RES_TYPE_TEX_2D
:
6324 case WINED3D_GL_RES_TYPE_TEX_3D
:
6326 case WINED3D_GL_RES_TYPE_TEX_CUBE
:
6328 case WINED3D_GL_RES_TYPE_TEX_RECT
:
6331 return "unexpected_resource_type";
6335 static GLuint
gen_arbfp_ffp_shader(const struct ffp_frag_settings
*settings
, const struct wined3d_gl_info
*gl_info
)
6338 struct wined3d_string_buffer buffer
;
6339 BOOL tex_read
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
6340 BOOL bump_used
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
6341 BOOL luminance_used
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
6342 UINT lowest_disabled_stage
;
6343 const char *textype
;
6345 char colorcor_dst
[8];
6347 DWORD arg0
, arg1
, arg2
;
6348 BOOL tempreg_used
= FALSE
, tfactor_used
= FALSE
;
6350 BOOL custom_linear_fog
= FALSE
;
6351 struct color_fixup_masks masks
;
6353 if (!string_buffer_init(&buffer
))
6355 ERR("Failed to initialize shader buffer.\n");
6359 shader_addline(&buffer
, "!!ARBfp1.0\n");
6361 if (settings
->color_key_enabled
)
6363 shader_addline(&buffer
, "PARAM color_key_low = program.env[%u];\n", ARB_FFP_CONST_COLOR_KEY_LOW
);
6364 shader_addline(&buffer
, "PARAM color_key_high = program.env[%u];\n", ARB_FFP_CONST_COLOR_KEY_HIGH
);
6368 /* Find out which textures are read */
6369 for (stage
= 0; stage
< MAX_TEXTURES
; ++stage
)
6371 if (settings
->op
[stage
].cop
== WINED3D_TOP_DISABLE
)
6373 arg0
= settings
->op
[stage
].carg0
& WINED3DTA_SELECTMASK
;
6374 arg1
= settings
->op
[stage
].carg1
& WINED3DTA_SELECTMASK
;
6375 arg2
= settings
->op
[stage
].carg2
& WINED3DTA_SELECTMASK
;
6376 if(arg0
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6377 if(arg1
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6378 if(arg2
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6380 if (settings
->op
[stage
].cop
== WINED3D_TOP_BLEND_TEXTURE_ALPHA
)
6381 tex_read
[stage
] = TRUE
;
6382 if (settings
->op
[stage
].cop
== WINED3D_TOP_BLEND_TEXTURE_ALPHA_PM
)
6383 tex_read
[stage
] = TRUE
;
6384 if (settings
->op
[stage
].cop
== WINED3D_TOP_BUMPENVMAP
)
6386 bump_used
[stage
] = TRUE
;
6387 tex_read
[stage
] = TRUE
;
6389 if (settings
->op
[stage
].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
)
6391 bump_used
[stage
] = TRUE
;
6392 tex_read
[stage
] = TRUE
;
6393 luminance_used
[stage
] = TRUE
;
6395 else if (settings
->op
[stage
].cop
== WINED3D_TOP_BLEND_FACTOR_ALPHA
)
6397 tfactor_used
= TRUE
;
6400 if(arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
) {
6401 tfactor_used
= TRUE
;
6404 if(settings
->op
[stage
].dst
== tempreg
) tempreg_used
= TRUE
;
6405 if(arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
) {
6406 tempreg_used
= TRUE
;
6409 if (settings
->op
[stage
].aop
== WINED3D_TOP_DISABLE
)
6411 arg0
= settings
->op
[stage
].aarg0
& WINED3DTA_SELECTMASK
;
6412 arg1
= settings
->op
[stage
].aarg1
& WINED3DTA_SELECTMASK
;
6413 arg2
= settings
->op
[stage
].aarg2
& WINED3DTA_SELECTMASK
;
6414 if(arg0
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6415 if(arg1
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6416 if(arg2
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6418 if(arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
) {
6419 tempreg_used
= TRUE
;
6421 if(arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
) {
6422 tfactor_used
= TRUE
;
6425 lowest_disabled_stage
= stage
;
6427 switch (settings
->fog
)
6429 case WINED3D_FFP_PS_FOG_OFF
: break;
6430 case WINED3D_FFP_PS_FOG_LINEAR
:
6431 if (gl_info
->quirks
& WINED3D_QUIRK_BROKEN_ARB_FOG
)
6433 custom_linear_fog
= TRUE
;
6436 shader_addline(&buffer
, "OPTION ARB_fog_linear;\n");
6439 case WINED3D_FFP_PS_FOG_EXP
: shader_addline(&buffer
, "OPTION ARB_fog_exp;\n"); break;
6440 case WINED3D_FFP_PS_FOG_EXP2
: shader_addline(&buffer
, "OPTION ARB_fog_exp2;\n"); break;
6441 default: FIXME("Unexpected fog setting %d\n", settings
->fog
);
6444 shader_addline(&buffer
, "PARAM const = {1, 2, 4, 0.5};\n");
6445 shader_addline(&buffer
, "TEMP TMP;\n");
6446 shader_addline(&buffer
, "TEMP ret;\n");
6447 if(tempreg_used
|| settings
->sRGB_write
) shader_addline(&buffer
, "TEMP tempreg;\n");
6448 shader_addline(&buffer
, "TEMP arg0;\n");
6449 shader_addline(&buffer
, "TEMP arg1;\n");
6450 shader_addline(&buffer
, "TEMP arg2;\n");
6451 for(stage
= 0; stage
< MAX_TEXTURES
; stage
++) {
6452 if(!tex_read
[stage
]) continue;
6453 shader_addline(&buffer
, "TEMP tex%u;\n", stage
);
6454 if(!bump_used
[stage
]) continue;
6455 shader_addline(&buffer
, "PARAM bumpmat%u = program.env[%u];\n", stage
, ARB_FFP_CONST_BUMPMAT(stage
));
6456 if(!luminance_used
[stage
]) continue;
6457 shader_addline(&buffer
, "PARAM luminance%u = program.env[%u];\n", stage
, ARB_FFP_CONST_LUMINANCE(stage
));
6460 shader_addline(&buffer
, "PARAM tfactor = program.env[%u];\n", ARB_FFP_CONST_TFACTOR
);
6462 shader_addline(&buffer
, "PARAM specular_enable = program.env[%u];\n", ARB_FFP_CONST_SPECULAR_ENABLE
);
6464 if (settings
->sRGB_write
)
6466 shader_addline(&buffer
, "PARAM srgb_consts0 = ");
6467 shader_arb_append_imm_vec4(&buffer
, wined3d_srgb_const0
);
6468 shader_addline(&buffer
, ";\n");
6469 shader_addline(&buffer
, "PARAM srgb_consts1 = ");
6470 shader_arb_append_imm_vec4(&buffer
, wined3d_srgb_const1
);
6471 shader_addline(&buffer
, ";\n");
6474 if (lowest_disabled_stage
< 7 && settings
->emul_clipplanes
)
6475 shader_addline(&buffer
, "KIL fragment.texcoord[7];\n");
6477 if (tempreg_used
|| settings
->sRGB_write
)
6478 shader_addline(&buffer
, "MOV tempreg, 0.0;\n");
6480 /* Generate texture sampling instructions */
6481 for (stage
= 0; stage
< MAX_TEXTURES
&& settings
->op
[stage
].cop
!= WINED3D_TOP_DISABLE
; ++stage
)
6483 if (!tex_read
[stage
])
6486 textype
= arbfp_texture_target(settings
->op
[stage
].tex_type
);
6488 if(settings
->op
[stage
].projected
== proj_none
) {
6490 } else if(settings
->op
[stage
].projected
== proj_count4
||
6491 settings
->op
[stage
].projected
== proj_count3
) {
6494 FIXME("Unexpected projection mode %d\n", settings
->op
[stage
].projected
);
6499 && (settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP
6500 || settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
))
6502 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, x, z, 0, 0;\n", stage
- 1);
6503 shader_addline(&buffer
, "DP3 ret.x, arg1, tex%u;\n", stage
- 1);
6504 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, y, w, 0, 0;\n", stage
- 1);
6505 shader_addline(&buffer
, "DP3 ret.y, arg1, tex%u;\n", stage
- 1);
6507 /* with projective textures, texbem only divides the static texture coord, not the displacement,
6508 * so multiply the displacement with the dividing parameter before passing it to TXP
6510 if (settings
->op
[stage
].projected
!= proj_none
) {
6511 if(settings
->op
[stage
].projected
== proj_count4
) {
6512 shader_addline(&buffer
, "MOV ret.w, fragment.texcoord[%u].w;\n", stage
);
6513 shader_addline(&buffer
, "MUL ret.xyz, ret, fragment.texcoord[%u].w, fragment.texcoord[%u];\n", stage
, stage
);
6515 shader_addline(&buffer
, "MOV ret.w, fragment.texcoord[%u].z;\n", stage
);
6516 shader_addline(&buffer
, "MAD ret.xyz, ret, fragment.texcoord[%u].z, fragment.texcoord[%u];\n", stage
, stage
);
6519 shader_addline(&buffer
, "ADD ret, ret, fragment.texcoord[%u];\n", stage
);
6522 shader_addline(&buffer
, "%s tex%u, ret, texture[%u], %s;\n",
6523 instr
, stage
, stage
, textype
);
6524 if (settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
)
6526 shader_addline(&buffer
, "MAD_SAT ret.x, tex%u.z, luminance%u.x, luminance%u.y;\n",
6527 stage
- 1, stage
- 1, stage
- 1);
6528 shader_addline(&buffer
, "MUL tex%u, tex%u, ret.x;\n", stage
, stage
);
6530 } else if(settings
->op
[stage
].projected
== proj_count3
) {
6531 shader_addline(&buffer
, "MOV ret, fragment.texcoord[%u];\n", stage
);
6532 shader_addline(&buffer
, "MOV ret.w, ret.z;\n");
6533 shader_addline(&buffer
, "%s tex%u, ret, texture[%u], %s;\n",
6534 instr
, stage
, stage
, textype
);
6536 shader_addline(&buffer
, "%s tex%u, fragment.texcoord[%u], texture[%u], %s;\n",
6537 instr
, stage
, stage
, stage
, textype
);
6540 sprintf(colorcor_dst
, "tex%u", stage
);
6541 masks
= calc_color_correction(settings
->op
[stage
].color_fixup
, WINED3DSP_WRITEMASK_ALL
);
6542 gen_color_correction(&buffer
, colorcor_dst
, colorcor_dst
, "const.x", "const.y",
6543 settings
->op
[stage
].color_fixup
, masks
);
6546 if (settings
->color_key_enabled
)
6548 shader_addline(&buffer
, "SLT TMP, tex0, color_key_low;\n"); /* below low key */
6549 shader_addline(&buffer
, "SGE ret, tex0, color_key_high;\n"); /* above high key */
6550 shader_addline(&buffer
, "ADD TMP, TMP, ret;\n"); /* or */
6551 shader_addline(&buffer
, "DP4 TMP.b, TMP, TMP;\n"); /* on any channel */
6552 shader_addline(&buffer
, "SGE TMP, -TMP.b, 0.0;\n"); /* logical not */
6553 shader_addline(&buffer
, "KIL -TMP;\n"); /* discard if true */
6556 shader_addline(&buffer
, "MOV ret, fragment.color.primary;\n");
6558 /* Generate the main shader */
6559 for (stage
= 0; stage
< MAX_TEXTURES
; ++stage
)
6561 if (settings
->op
[stage
].cop
== WINED3D_TOP_DISABLE
)
6564 if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG1
6565 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG1
)
6566 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
;
6567 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG1
6568 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG2
)
6569 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg2
;
6570 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG2
6571 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG1
)
6572 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg1
;
6573 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG2
6574 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG2
)
6575 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
6577 op_equal
= settings
->op
[stage
].aop
== settings
->op
[stage
].cop
6578 && settings
->op
[stage
].carg0
== settings
->op
[stage
].aarg0
6579 && settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
6580 && settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
6582 if (settings
->op
[stage
].aop
== WINED3D_TOP_DISABLE
)
6584 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].dst
,
6585 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6586 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6590 gen_ffp_instr(&buffer
, stage
, TRUE
, TRUE
, settings
->op
[stage
].dst
,
6591 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6592 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6594 else if (settings
->op
[stage
].cop
!= WINED3D_TOP_BUMPENVMAP
6595 && settings
->op
[stage
].cop
!= WINED3D_TOP_BUMPENVMAP_LUMINANCE
)
6597 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].dst
,
6598 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6599 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6600 gen_ffp_instr(&buffer
, stage
, FALSE
, TRUE
, settings
->op
[stage
].dst
,
6601 settings
->op
[stage
].aop
, settings
->op
[stage
].aarg0
,
6602 settings
->op
[stage
].aarg1
, settings
->op
[stage
].aarg2
);
6606 if (settings
->sRGB_write
|| custom_linear_fog
)
6608 shader_addline(&buffer
, "MAD ret, fragment.color.secondary, specular_enable, ret;\n");
6609 if (settings
->sRGB_write
)
6610 arbfp_add_sRGB_correction(&buffer
, "ret", "arg0", "arg1", "arg2", "tempreg", FALSE
);
6611 if (custom_linear_fog
)
6612 arbfp_add_linear_fog(&buffer
, "ret", "arg0");
6613 shader_addline(&buffer
, "MOV result.color, ret;\n");
6617 shader_addline(&buffer
, "MAD result.color, fragment.color.secondary, specular_enable, ret;\n");
6621 shader_addline(&buffer
, "END\n");
6623 /* Generate the shader */
6624 GL_EXTCALL(glGenProgramsARB(1, &ret
));
6625 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, ret
));
6626 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
6628 string_buffer_free(&buffer
);
6632 static void fragment_prog_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6634 const struct wined3d_device
*device
= context
->swapchain
->device
;
6635 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
6636 struct shader_arb_priv
*priv
= device
->fragment_priv
;
6637 BOOL use_pshader
= use_ps(state
);
6638 struct ffp_frag_settings settings
;
6639 const struct arbfp_ffp_desc
*desc
;
6642 TRACE("context %p, state %p, state_id %#x.\n", context
, state
, state_id
);
6644 if (isStateDirty(context
, STATE_RENDER(WINED3D_RS_FOGENABLE
)))
6646 if (!use_pshader
&& device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
)
6648 /* Reload fixed function constants since they collide with the
6649 * pixel shader constants. */
6650 for (i
= 0; i
< MAX_TEXTURES
; ++i
)
6652 set_bumpmat_arbfp(context
, state
, STATE_TEXTURESTAGE(i
, WINED3D_TSS_BUMPENV_MAT00
));
6654 state_texfactor_arbfp(context
, state
, STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
));
6655 state_arb_specularenable(context
, state
, STATE_RENDER(WINED3D_RS_SPECULARENABLE
));
6656 color_key_arbfp(context
, state
, STATE_COLOR_KEY
);
6658 else if (use_pshader
)
6660 context
->shader_update_mask
|= 1u << WINED3D_SHADER_TYPE_PIXEL
;
6667 /* Find or create a shader implementing the fixed function pipeline
6668 * settings, then activate it. */
6669 gen_ffp_frag_op(context
, state
, &settings
, FALSE
);
6670 desc
= (const struct arbfp_ffp_desc
*)find_ffp_frag_shader(&priv
->fragment_shaders
, &settings
);
6672 struct arbfp_ffp_desc
*new_desc
= HeapAlloc(GetProcessHeap(), 0, sizeof(*new_desc
));
6675 ERR("Out of memory\n");
6679 new_desc
->parent
.settings
= settings
;
6680 new_desc
->shader
= gen_arbfp_ffp_shader(&settings
, gl_info
);
6681 add_ffp_frag_shader(&priv
->fragment_shaders
, &new_desc
->parent
);
6682 TRACE("Allocated fixed function replacement shader descriptor %p\n", new_desc
);
6686 /* Now activate the replacement program. GL_FRAGMENT_PROGRAM_ARB is already active (however, note the
6687 * comment above the shader_select call below). If e.g. GLSL is active, the shader_select call will
6690 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, desc
->shader
));
6691 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, desc->shader)");
6692 priv
->current_fprogram_id
= desc
->shader
;
6694 if (device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
)
6696 /* Reload fixed function constants since they collide with the
6697 * pixel shader constants. */
6698 for (i
= 0; i
< MAX_TEXTURES
; ++i
)
6700 set_bumpmat_arbfp(context
, state
, STATE_TEXTURESTAGE(i
, WINED3D_TSS_BUMPENV_MAT00
));
6702 state_texfactor_arbfp(context
, state
, STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
));
6703 state_arb_specularenable(context
, state
, STATE_RENDER(WINED3D_RS_SPECULARENABLE
));
6705 context
->last_was_pshader
= FALSE
;
6707 else if (!context
->last_was_pshader
)
6709 if (device
->shader_backend
== &arb_program_shader_backend
)
6710 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_F
;
6711 context
->last_was_pshader
= TRUE
;
6714 context
->shader_update_mask
|= 1u << WINED3D_SHADER_TYPE_PIXEL
;
6717 /* We can't link the fog states to the fragment state directly since the
6718 * vertex pipeline links them to FOGENABLE. A different linking in different
6719 * pipeline parts can't be expressed in the combined state table, so we need
6720 * to handle that with a forwarding function. The other invisible side effect
6721 * is that changing the fog start and fog end (which links to FOGENABLE in
6722 * vertex) results in the fragment_prog_arbfp function being called because
6723 * FOGENABLE is dirty, which calls this function here. */
6724 static void state_arbfp_fog(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6726 enum fogsource new_source
;
6727 DWORD fogstart
= state
->render_states
[WINED3D_RS_FOGSTART
];
6728 DWORD fogend
= state
->render_states
[WINED3D_RS_FOGEND
];
6730 TRACE("context %p, state %p, state_id %#x.\n", context
, state
, state_id
);
6732 if (!isStateDirty(context
, STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
)))
6733 fragment_prog_arbfp(context
, state
, state_id
);
6735 if (!state
->render_states
[WINED3D_RS_FOGENABLE
])
6738 if (state
->render_states
[WINED3D_RS_FOGTABLEMODE
] == WINED3D_FOG_NONE
)
6742 new_source
= FOGSOURCE_VS
;
6746 if (state
->render_states
[WINED3D_RS_FOGVERTEXMODE
] == WINED3D_FOG_NONE
|| context
->last_was_rhw
)
6747 new_source
= FOGSOURCE_COORD
;
6749 new_source
= FOGSOURCE_FFP
;
6754 new_source
= FOGSOURCE_FFP
;
6757 if (new_source
!= context
->fog_source
|| fogstart
== fogend
)
6759 context
->fog_source
= new_source
;
6760 state_fogstartend(context
, state
, STATE_RENDER(WINED3D_RS_FOGSTART
));
6764 static void textransform(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6766 if (!isStateDirty(context
, STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
)))
6767 fragment_prog_arbfp(context
, state
, state_id
);
6770 static const struct StateEntryTemplate arbfp_fragmentstate_template
[] =
6772 {STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
), { STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
), state_texfactor_arbfp
}, WINED3D_GL_EXT_NONE
},
6773 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6774 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6775 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6776 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6777 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6778 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6779 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6780 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6781 {STATE_TEXTURESTAGE(0, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6782 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6783 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6784 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6785 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6786 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6787 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6788 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6789 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6790 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6791 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6792 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6793 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6794 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6795 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6796 {STATE_TEXTURESTAGE(1, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6797 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6798 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6799 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6800 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6801 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6802 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6803 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6804 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6805 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6806 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6807 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6808 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6809 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6810 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6811 {STATE_TEXTURESTAGE(2, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6812 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6813 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6814 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6815 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6816 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6817 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6818 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6819 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6820 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6821 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6822 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6823 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6824 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6825 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6826 {STATE_TEXTURESTAGE(3, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6827 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6828 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6829 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6830 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6831 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6832 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6833 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6834 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6835 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6836 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6837 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6838 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6839 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6840 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6841 {STATE_TEXTURESTAGE(4, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6842 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6843 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6844 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6845 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6846 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6847 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6848 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6849 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6850 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6851 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6852 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6853 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6854 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6855 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6856 {STATE_TEXTURESTAGE(5, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6857 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6858 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6859 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6860 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6861 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6862 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6863 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6864 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6865 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6866 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6867 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6868 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6869 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6870 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6871 {STATE_TEXTURESTAGE(6, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6872 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6873 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6874 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6875 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6876 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6877 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6878 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6879 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6880 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6881 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6882 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6883 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6884 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6885 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6886 {STATE_TEXTURESTAGE(7, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6887 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6888 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6889 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6890 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6891 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6892 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6893 {STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
6894 {STATE_RENDER(WINED3D_RS_ALPHAFUNC
), { STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6895 {STATE_RENDER(WINED3D_RS_ALPHAREF
), { STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6896 {STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), { STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), alpha_test_arbfp
}, WINED3D_GL_EXT_NONE
},
6897 {STATE_RENDER(WINED3D_RS_COLORKEYENABLE
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6898 {STATE_COLOR_KEY
, { STATE_COLOR_KEY
, color_key_arbfp
}, WINED3D_GL_EXT_NONE
},
6899 {STATE_RENDER(WINED3D_RS_FOGENABLE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), state_arbfp_fog
}, WINED3D_GL_EXT_NONE
},
6900 {STATE_RENDER(WINED3D_RS_FOGTABLEMODE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6901 {STATE_RENDER(WINED3D_RS_FOGVERTEXMODE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6902 {STATE_RENDER(WINED3D_RS_FOGSTART
), { STATE_RENDER(WINED3D_RS_FOGSTART
), state_fogstartend
}, WINED3D_GL_EXT_NONE
},
6903 {STATE_RENDER(WINED3D_RS_FOGEND
), { STATE_RENDER(WINED3D_RS_FOGSTART
), NULL
}, WINED3D_GL_EXT_NONE
},
6904 {STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), { STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), state_srgbwrite
}, ARB_FRAMEBUFFER_SRGB
},
6905 {STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6906 {STATE_RENDER(WINED3D_RS_FOGCOLOR
), { STATE_RENDER(WINED3D_RS_FOGCOLOR
), state_fogcolor
}, WINED3D_GL_EXT_NONE
},
6907 {STATE_RENDER(WINED3D_RS_FOGDENSITY
), { STATE_RENDER(WINED3D_RS_FOGDENSITY
), state_fogdensity
}, WINED3D_GL_EXT_NONE
},
6908 {STATE_TEXTURESTAGE(0,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(0, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6909 {STATE_TEXTURESTAGE(1,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(1, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6910 {STATE_TEXTURESTAGE(2,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(2, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6911 {STATE_TEXTURESTAGE(3,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(3, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6912 {STATE_TEXTURESTAGE(4,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(4, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6913 {STATE_TEXTURESTAGE(5,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(5, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6914 {STATE_TEXTURESTAGE(6,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(6, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6915 {STATE_TEXTURESTAGE(7,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(7, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6916 {STATE_RENDER(WINED3D_RS_SPECULARENABLE
), { STATE_RENDER(WINED3D_RS_SPECULARENABLE
), state_arb_specularenable
}, WINED3D_GL_EXT_NONE
},
6917 {STATE_RENDER(WINED3D_RS_SHADEMODE
), { STATE_RENDER(WINED3D_RS_SHADEMODE
), state_shademode
}, WINED3D_GL_EXT_NONE
},
6918 {0 /* Terminate */, { 0, 0 }, WINED3D_GL_EXT_NONE
},
6921 static BOOL
arbfp_alloc_context_data(struct wined3d_context
*context
)
6926 static void arbfp_free_context_data(struct wined3d_context
*context
)
6930 const struct fragment_pipeline arbfp_fragment_pipeline
= {
6933 arbfp_get_emul_mask
,
6936 arbfp_alloc_context_data
,
6937 arbfp_free_context_data
,
6938 shader_arb_color_fixup_supported
,
6939 arbfp_fragmentstate_template
,
6942 struct arbfp_blit_type
6944 enum complex_fixup fixup
: 4;
6945 enum wined3d_gl_resource_type res_type
: 3;
6946 DWORD use_color_key
: 1;
6950 struct arbfp_blit_desc
6953 struct arbfp_blit_type type
;
6954 struct wine_rb_entry entry
;
6957 #define ARBFP_BLIT_PARAM_SIZE 0
6958 #define ARBFP_BLIT_PARAM_COLOR_KEY_LOW 1
6959 #define ARBFP_BLIT_PARAM_COLOR_KEY_HIGH 2
6961 struct arbfp_blit_priv
6963 struct wine_rb_tree shaders
;
6964 GLuint palette_texture
;
6967 static int arbfp_blit_type_compare(const void *key
, const struct wine_rb_entry
*entry
)
6969 const struct arbfp_blit_type
*ka
= key
;
6970 const struct arbfp_blit_type
*kb
= &WINE_RB_ENTRY_VALUE(entry
, const struct arbfp_blit_desc
, entry
)->type
;
6972 return memcmp(ka
, kb
, sizeof(*ka
));
6975 /* Context activation is done by the caller. */
6976 static void arbfp_free_blit_shader(struct wine_rb_entry
*entry
, void *context
)
6978 const struct wined3d_gl_info
*gl_info
= context
;
6979 struct arbfp_blit_desc
*entry_arb
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_blit_desc
, entry
);
6981 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb
->shader
));
6982 checkGLcall("glDeleteProgramsARB(1, &entry_arb->shader)");
6983 HeapFree(GetProcessHeap(), 0, entry_arb
);
6986 static const struct wine_rb_functions wined3d_arbfp_blit_rb_functions
=
6991 arbfp_blit_type_compare
,
6994 static HRESULT
arbfp_blit_alloc(struct wined3d_device
*device
)
6996 struct arbfp_blit_priv
*priv
;
6998 if (!(priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*priv
))))
6999 return E_OUTOFMEMORY
;
7001 if (wine_rb_init(&priv
->shaders
, &wined3d_arbfp_blit_rb_functions
) == -1)
7003 ERR("Failed to initialize rbtree.\n");
7004 HeapFree(GetProcessHeap(), 0, priv
);
7005 return E_OUTOFMEMORY
;
7008 device
->blit_priv
= priv
;
7013 /* Context activation is done by the caller. */
7014 static void arbfp_blit_free(struct wined3d_device
*device
)
7016 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
7017 struct arbfp_blit_priv
*priv
= device
->blit_priv
;
7019 wine_rb_destroy(&priv
->shaders
, arbfp_free_blit_shader
, &device
->adapter
->gl_info
);
7020 checkGLcall("Delete blit programs");
7022 if (priv
->palette_texture
)
7023 gl_info
->gl_ops
.gl
.p_glDeleteTextures(1, &priv
->palette_texture
);
7025 HeapFree(GetProcessHeap(), 0, device
->blit_priv
);
7026 device
->blit_priv
= NULL
;
7029 static BOOL
gen_planar_yuv_read(struct wined3d_string_buffer
*buffer
, const struct arbfp_blit_type
*type
,
7033 const char *tex
, *texinstr
= "TXP";
7035 if (type
->fixup
== COMPLEX_FIXUP_UYVY
)
7046 tex
= arbfp_texture_target(type
->res_type
);
7047 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_RECT
)
7050 /* First we have to read the chroma values. This means we need at least two pixels(no filtering),
7051 * or 4 pixels(with filtering). To get the unmodified chromas, we have to rid ourselves of the
7052 * filtering when we sample the texture.
7054 * These are the rules for reading the chroma:
7060 * So we have to get the sampling x position in non-normalized coordinates in integers
7062 if (type
->res_type
!= WINED3D_GL_RES_TYPE_TEX_RECT
)
7064 shader_addline(buffer
, "MUL texcrd.xy, fragment.texcoord[0], size.x;\n");
7065 shader_addline(buffer
, "MOV texcrd.w, size.x;\n");
7069 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7071 /* We must not allow filtering between pixel x and x+1, this would mix U and V
7072 * Vertical filtering is ok. However, bear in mind that the pixel center is at
7075 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
7076 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, coef.y;\n");
7078 /* Divide the x coordinate by 0.5 and get the fraction. This gives 0.25 and 0.75 for the
7079 * even and odd pixels respectively
7081 shader_addline(buffer
, "MUL texcrd2, texcrd, coef.y;\n");
7082 shader_addline(buffer
, "FRC texcrd2, texcrd2;\n");
7084 /* Sample Pixel 1 */
7085 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
7087 /* Put the value into either of the chroma values */
7088 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
7089 shader_addline(buffer
, "MUL chroma.x, luminance.%c, temp.x;\n", chroma
);
7090 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
7091 shader_addline(buffer
, "MUL chroma.y, luminance.%c, temp.x;\n", chroma
);
7093 /* Sample pixel 2. If we read an even pixel(SLT above returned 1), sample
7094 * the pixel right to the current one. Otherwise, sample the left pixel.
7095 * Bias and scale the SLT result to -1;1 and add it to the texcrd.x.
7097 shader_addline(buffer
, "MAD temp.x, temp.x, coef.z, -coef.x;\n");
7098 shader_addline(buffer
, "ADD texcrd.x, texcrd, temp.x;\n");
7099 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
7101 /* Put the value into the other chroma */
7102 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
7103 shader_addline(buffer
, "MAD chroma.y, luminance.%c, temp.x, chroma.y;\n", chroma
);
7104 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
7105 shader_addline(buffer
, "MAD chroma.x, luminance.%c, temp.x, chroma.x;\n", chroma
);
7107 /* TODO: If filtering is enabled, sample a 2nd pair of pixels left or right of
7108 * the current one and lerp the two U and V values
7111 /* This gives the correctly filtered luminance value */
7112 shader_addline(buffer
, "TEX luminance, fragment.texcoord[0], texture[0], %s;\n", tex
);
7117 static BOOL
gen_yv12_read(struct wined3d_string_buffer
*buffer
, const struct arbfp_blit_type
*type
,
7121 static const float yv12_coef
[]
7122 = {2.0f
/ 3.0f
, 1.0f
/ 6.0f
, (2.0f
/ 3.0f
) + (1.0f
/ 6.0f
), 1.0f
/ 3.0f
};
7124 tex
= arbfp_texture_target(type
->res_type
);
7126 /* YV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
7127 * V and a (W/2)x(H/2) U plane, each with 8 bit per pixel. So the effective
7128 * bitdepth is 12 bits per pixel. Since the U and V planes have only half the
7129 * pitch of the luminance plane, the packing into the gl texture is a bit
7130 * unfortunate. If the whole texture is interpreted as luminance data it looks
7131 * approximately like this:
7133 * +----------------------------------+----
7145 * +----------------+-----------------+----
7147 * | V even rows | V odd rows |
7149 * +----------------+------------------ -
7151 * | U even rows | U odd rows |
7153 * +----------------+-----------------+----
7157 * So it appears as if there are 4 chroma images, but in fact the odd rows
7158 * in the chroma images are in the same row as the even ones. So its is
7159 * kinda tricky to read
7161 * When reading from rectangle textures, keep in mind that the input y coordinates
7162 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height
7164 shader_addline(buffer
, "PARAM yv12_coef = ");
7165 shader_arb_append_imm_vec4(buffer
, yv12_coef
);
7166 shader_addline(buffer
, ";\n");
7168 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7169 /* the chroma planes have only half the width */
7170 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, coef.y;\n");
7172 /* The first value is between 2/3 and 5/6th of the texture's height, so scale+bias
7173 * the coordinate. Also read the right side of the image when reading odd lines
7175 * Don't forget to clamp the y values in into the range, otherwise we'll get filtering
7178 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7181 shader_addline(buffer
, "RCP chroma.w, size.y;\n");
7183 shader_addline(buffer
, "MUL texcrd2.y, texcrd.y, size.y;\n");
7185 shader_addline(buffer
, "FLR texcrd2.y, texcrd2.y;\n");
7186 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, yv12_coef.y, yv12_coef.x;\n");
7188 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
7189 shader_addline(buffer
, "ADD texcrd2.x, texcrd2.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
7190 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
7191 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
7192 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
7194 /* clamp, keep the half pixel origin in mind */
7195 shader_addline(buffer
, "MAD temp.y, coef.y, chroma.w, yv12_coef.x;\n");
7196 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7197 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.z;\n");
7198 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7202 /* Read from [size - size+size/4] */
7203 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
7204 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, coef.w, size.y;\n");
7206 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
7207 shader_addline(buffer
, "ADD texcrd2.x, texcrd.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
7208 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
7209 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
7210 shader_addline(buffer
, "MUL texcrd2.x, texcrd2.x, size.x;\n");
7211 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
7213 /* Make sure to read exactly from the pixel center */
7214 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
7215 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, coef.y;\n");
7218 shader_addline(buffer
, "MAD temp.y, size.y, coef.w, size.y;\n");
7219 shader_addline(buffer
, "ADD temp.y, temp.y, -coef.y;\n");
7220 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7221 shader_addline(buffer
, "ADD temp.y, size.y, coef.y;\n");
7222 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7224 /* Read the texture, put the result into the output register */
7225 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7226 shader_addline(buffer
, "MOV chroma.x, temp.w;\n");
7228 /* The other chroma value is 1/6th of the texture lower, from 5/6th to 6/6th
7229 * No need to clamp because we're just reusing the already clamped value from above
7231 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7232 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, yv12_coef.y;\n");
7234 shader_addline(buffer
, "MAD texcrd.y, size.y, coef.w, texcrd.y;\n");
7235 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7236 shader_addline(buffer
, "MOV chroma.y, temp.w;\n");
7238 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
7239 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
7240 * values due to filtering
7242 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7243 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7245 /* Multiply the y coordinate by 2/3 and clamp it */
7246 shader_addline(buffer
, "MUL texcrd.y, texcrd.y, yv12_coef.x;\n");
7247 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.x;\n");
7248 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7249 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7253 /* Reading from texture_rectangles is pretty straightforward, just use the unmodified
7254 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
7257 shader_addline(buffer
, "ADD temp.x, size.y, -coef.y;\n");
7258 shader_addline(buffer
, "MIN texcrd.y, texcrd.y, size.x;\n");
7259 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7266 static BOOL
gen_nv12_read(struct wined3d_string_buffer
*buffer
, const struct arbfp_blit_type
*type
,
7270 static const float nv12_coef
[]
7271 = {2.0f
/ 3.0f
, 1.0f
/ 3.0f
, 1.0f
, 1.0f
};
7273 tex
= arbfp_texture_target(type
->res_type
);
7275 /* NV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
7276 * sized plane where each component is an UV pair. So the effective
7277 * bitdepth is 12 bits per pixel If the whole texture is interpreted as luminance
7278 * data it looks approximately like this:
7280 * +----------------------------------+----
7292 * +----------------------------------+----
7293 * |UVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUV|
7294 * |UVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUV|
7300 * +----------------------------------+----
7302 * When reading from rectangle textures, keep in mind that the input y coordinates
7303 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height. */
7305 shader_addline(buffer
, "PARAM nv12_coef = ");
7306 shader_arb_append_imm_vec4(buffer
, nv12_coef
);
7307 shader_addline(buffer
, ";\n");
7309 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7310 /* We only have half the number of chroma pixels. */
7311 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, coef.y;\n");
7313 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7315 shader_addline(buffer
, "RCP chroma.w, size.x;\n");
7316 shader_addline(buffer
, "RCP chroma.z, size.y;\n");
7318 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, nv12_coef.y, nv12_coef.x;\n");
7320 /* We must not allow filtering horizontally, this would mix U and V.
7321 * Vertical filtering is ok. However, bear in mind that the pixel center is at
7322 * 0.5, so add 0.5. */
7324 /* Convert to non-normalized coordinates so we can find the
7325 * individual pixel. */
7326 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, size.x;\n");
7327 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
7328 /* Multiply by 2 since chroma components are stored in UV pixel pairs,
7329 * add 0.5 to hit the center of the pixel. */
7330 shader_addline(buffer
, "MAD texcrd.x, texcrd.x, coef.z, coef.y;\n");
7332 /* Convert back to normalized coordinates. */
7333 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, chroma.w;\n");
7335 /* Clamp, keep the half pixel origin in mind. */
7336 shader_addline(buffer
, "MAD temp.y, coef.y, chroma.z, nv12_coef.x;\n");
7337 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7338 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.z, nv12_coef.z;\n");
7339 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7343 /* Read from [size - size+size/2] */
7344 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, coef.y, size.y;\n");
7346 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
7347 /* Multiply by 2 since chroma components are stored in UV pixel pairs,
7348 * add 0.5 to hit the center of the pixel. */
7349 shader_addline(buffer
, "MAD texcrd.x, texcrd.x, coef.z, coef.y;\n");
7352 shader_addline(buffer
, "MAD temp.y, size.y, coef.y, size.y;\n");
7353 shader_addline(buffer
, "ADD temp.y, temp.y, -coef.y;\n");
7354 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7355 shader_addline(buffer
, "ADD temp.y, size.y, coef.y;\n");
7356 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7358 /* Read the texture, put the result into the output register. */
7359 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7360 shader_addline(buffer
, "MOV chroma.y, temp.w;\n");
7362 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7365 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, chroma.w;\n");
7370 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, coef.x;\n");
7372 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7373 shader_addline(buffer
, "MOV chroma.x, temp.w;\n");
7375 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
7376 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
7377 * values due to filtering. */
7378 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7379 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7381 /* Multiply the y coordinate by 2/3 and clamp it */
7382 shader_addline(buffer
, "MUL texcrd.y, texcrd.y, nv12_coef.x;\n");
7383 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, nv12_coef.x;\n");
7384 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7385 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7389 /* Reading from texture_rectangles is pretty straightforward, just use the unmodified
7390 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
7393 shader_addline(buffer
, "ADD temp.x, size.y, -coef.y;\n");
7394 shader_addline(buffer
, "MIN texcrd.y, texcrd.y, size.x;\n");
7395 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7402 /* Context activation is done by the caller. */
7403 static GLuint
gen_p8_shader(struct arbfp_blit_priv
*priv
,
7404 const struct wined3d_gl_info
*gl_info
, const struct arbfp_blit_type
*type
)
7407 struct wined3d_string_buffer buffer
;
7408 const char *tex_target
= arbfp_texture_target(type
->res_type
);
7410 /* This should not happen because we only use this conversion for
7411 * present blits which don't use color keying. */
7412 if (type
->use_color_key
)
7413 FIXME("Implement P8 color keying.\n");
7416 if (!string_buffer_init(&buffer
))
7418 ERR("Failed to initialize shader buffer.\n");
7422 GL_EXTCALL(glGenProgramsARB(1, &shader
));
7423 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7426 string_buffer_free(&buffer
);
7430 shader_addline(&buffer
, "!!ARBfp1.0\n");
7431 shader_addline(&buffer
, "TEMP index;\n");
7433 /* { 255/256, 0.5/255*255/256, 0, 0 } */
7434 shader_addline(&buffer
, "PARAM constants = { 0.996, 0.00195, 0, 0 };\n");
7436 /* The alpha-component contains the palette index */
7437 shader_addline(&buffer
, "TEX index, fragment.texcoord[0], texture[0], %s;\n", tex_target
);
7439 /* Scale the index by 255/256 and add a bias of '0.5' in order to sample in the middle */
7440 shader_addline(&buffer
, "MAD index.a, index.a, constants.x, constants.y;\n");
7442 /* Use the alpha-component as an index in the palette to get the final color */
7443 shader_addline(&buffer
, "TEX result.color, index.a, texture[1], 1D;\n");
7444 shader_addline(&buffer
, "END\n");
7446 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
7448 string_buffer_free(&buffer
);
7453 /* Context activation is done by the caller. */
7454 static void upload_palette(const struct wined3d_texture
*texture
, struct wined3d_context
*context
)
7456 const struct wined3d_palette
*palette
= texture
->swapchain
? texture
->swapchain
->palette
: NULL
;
7457 struct wined3d_device
*device
= texture
->resource
.device
;
7458 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
7459 struct arbfp_blit_priv
*priv
= device
->blit_priv
;
7461 if (!priv
->palette_texture
)
7462 gl_info
->gl_ops
.gl
.p_glGenTextures(1, &priv
->palette_texture
);
7464 GL_EXTCALL(glActiveTexture(GL_TEXTURE1
));
7465 gl_info
->gl_ops
.gl
.p_glBindTexture(GL_TEXTURE_1D
, priv
->palette_texture
);
7467 gl_info
->gl_ops
.gl
.p_glTexEnvi(GL_TEXTURE_ENV
, GL_TEXTURE_ENV_MODE
, GL_REPLACE
);
7469 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_MIN_FILTER
, GL_NEAREST
);
7470 /* Make sure we have discrete color levels. */
7471 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_MAG_FILTER
, GL_NEAREST
);
7472 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_WRAP_S
, GL_CLAMP_TO_EDGE
);
7473 /* TODO: avoid unneeded uploads in the future by adding some SFLAG_PALETTE_DIRTY mechanism */
7476 gl_info
->gl_ops
.gl
.p_glTexImage1D(GL_TEXTURE_1D
, 0, GL_RGB
, 256, 0, GL_BGRA
,
7477 GL_UNSIGNED_INT_8_8_8_8_REV
, palette
->colors
);
7481 static const DWORD black
;
7482 FIXME("P8 surface loaded without a palette.\n");
7483 gl_info
->gl_ops
.gl
.p_glTexImage1D(GL_TEXTURE_1D
, 0, GL_RGB
, 1, 0, GL_BGRA
,
7484 GL_UNSIGNED_INT_8_8_8_8_REV
, &black
);
7487 /* Switch back to unit 0 in which the 2D texture will be stored. */
7488 context_active_texture(context
, gl_info
, 0);
7491 /* Context activation is done by the caller. */
7492 static GLuint
gen_yuv_shader(struct arbfp_blit_priv
*priv
, const struct wined3d_gl_info
*gl_info
,
7493 const struct arbfp_blit_type
*type
)
7496 struct wined3d_string_buffer buffer
;
7497 char luminance_component
;
7499 if (type
->use_color_key
)
7500 FIXME("Implement YUV color keying.\n");
7503 if (!string_buffer_init(&buffer
))
7505 ERR("Failed to initialize shader buffer.\n");
7509 GL_EXTCALL(glGenProgramsARB(1, &shader
));
7510 checkGLcall("GL_EXTCALL(glGenProgramsARB(1, &shader))");
7511 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7512 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
7515 string_buffer_free(&buffer
);
7519 /* The YUY2 and UYVY formats contain two pixels packed into a 32 bit macropixel,
7520 * giving effectively 16 bit per pixel. The color consists of a luminance(Y) and
7521 * two chroma(U and V) values. Each macropixel has two luminance values, one for
7522 * each single pixel it contains, and one U and one V value shared between both
7525 * The data is loaded into an A8L8 texture. With YUY2, the luminance component
7526 * contains the luminance and alpha the chroma. With UYVY it is vice versa. Thus
7527 * take the format into account when generating the read swizzles
7529 * Reading the Y value is straightforward - just sample the texture. The hardware
7530 * takes care of filtering in the horizontal and vertical direction.
7532 * Reading the U and V values is harder. We have to avoid filtering horizontally,
7533 * because that would mix the U and V values of one pixel or two adjacent pixels.
7534 * Thus floor the texture coordinate and add 0.5 to get an unfiltered read,
7535 * regardless of the filtering setting. Vertical filtering works automatically
7536 * though - the U and V values of two rows are mixed nicely.
7538 * Apart of avoiding filtering issues, the code has to know which value it just
7539 * read, and where it can find the other one. To determine this, it checks if
7540 * it sampled an even or odd pixel, and shifts the 2nd read accordingly.
7542 * Handling horizontal filtering of U and V values requires reading a 2nd pair
7543 * of pixels, extracting U and V and mixing them. This is not implemented yet.
7545 * An alternative implementation idea is to load the texture as A8R8G8B8 texture,
7546 * with width / 2. This way one read gives all 3 values, finding U and V is easy
7547 * in an unfiltered situation. Finding the luminance on the other hand requires
7548 * finding out if it is an odd or even pixel. The real drawback of this approach
7549 * is filtering. This would have to be emulated completely in the shader, reading
7550 * up two 2 packed pixels in up to 2 rows and interpolating both horizontally and
7551 * vertically. Beyond that it would require adjustments to the texture handling
7552 * code to deal with the width scaling
7554 shader_addline(&buffer
, "!!ARBfp1.0\n");
7555 shader_addline(&buffer
, "TEMP luminance;\n");
7556 shader_addline(&buffer
, "TEMP temp;\n");
7557 shader_addline(&buffer
, "TEMP chroma;\n");
7558 shader_addline(&buffer
, "TEMP texcrd;\n");
7559 shader_addline(&buffer
, "TEMP texcrd2;\n");
7560 shader_addline(&buffer
, "PARAM coef = {1.0, 0.5, 2.0, 0.25};\n");
7561 shader_addline(&buffer
, "PARAM yuv_coef = {1.403, 0.344, 0.714, 1.770};\n");
7562 shader_addline(&buffer
, "PARAM size = program.local[%u];\n", ARBFP_BLIT_PARAM_SIZE
);
7564 switch (type
->fixup
)
7566 case COMPLEX_FIXUP_UYVY
:
7567 case COMPLEX_FIXUP_YUY2
:
7568 if (!gen_planar_yuv_read(&buffer
, type
, &luminance_component
))
7570 string_buffer_free(&buffer
);
7575 case COMPLEX_FIXUP_YV12
:
7576 if (!gen_yv12_read(&buffer
, type
, &luminance_component
))
7578 string_buffer_free(&buffer
);
7583 case COMPLEX_FIXUP_NV12
:
7584 if (!gen_nv12_read(&buffer
, type
, &luminance_component
))
7586 string_buffer_free(&buffer
);
7592 FIXME("Unsupported YUV fixup %#x\n", type
->fixup
);
7593 string_buffer_free(&buffer
);
7597 /* Calculate the final result. Formula is taken from
7598 * http://www.fourcc.org/fccyvrgb.php. Note that the chroma
7599 * ranges from -0.5 to 0.5
7601 shader_addline(&buffer
, "SUB chroma.xy, chroma, coef.y;\n");
7603 shader_addline(&buffer
, "MAD result.color.x, chroma.x, yuv_coef.x, luminance.%c;\n", luminance_component
);
7604 shader_addline(&buffer
, "MAD temp.x, -chroma.y, yuv_coef.y, luminance.%c;\n", luminance_component
);
7605 shader_addline(&buffer
, "MAD result.color.y, -chroma.x, yuv_coef.z, temp.x;\n");
7606 shader_addline(&buffer
, "MAD result.color.z, chroma.y, yuv_coef.w, luminance.%c;\n", luminance_component
);
7607 shader_addline(&buffer
, "END\n");
7609 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
7611 string_buffer_free(&buffer
);
7616 /* Context activation is done by the caller. */
7617 static GLuint
arbfp_gen_plain_shader(struct arbfp_blit_priv
*priv
,
7618 const struct wined3d_gl_info
*gl_info
, const struct arbfp_blit_type
*type
)
7621 struct wined3d_string_buffer buffer
;
7622 const char *tex_target
= arbfp_texture_target(type
->res_type
);
7625 if (!string_buffer_init(&buffer
))
7627 ERR("Failed to initialize shader buffer.\n");
7631 GL_EXTCALL(glGenProgramsARB(1, &shader
));
7634 string_buffer_free(&buffer
);
7637 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7639 shader_addline(&buffer
, "!!ARBfp1.0\n");
7641 if (type
->use_color_key
)
7643 shader_addline(&buffer
, "TEMP color;\n");
7644 shader_addline(&buffer
, "TEMP less, greater;\n");
7645 shader_addline(&buffer
, "PARAM color_key_low = program.local[%u];\n", ARBFP_BLIT_PARAM_COLOR_KEY_LOW
);
7646 shader_addline(&buffer
, "PARAM color_key_high = program.local[%u];\n", ARBFP_BLIT_PARAM_COLOR_KEY_HIGH
);
7647 shader_addline(&buffer
, "TEX color, fragment.texcoord[0], texture[0], %s;\n", tex_target
);
7648 shader_addline(&buffer
, "SLT less, color, color_key_low;\n"); /* below low key */
7649 shader_addline(&buffer
, "SGE greater, color, color_key_high;\n"); /* above high key */
7650 shader_addline(&buffer
, "ADD less, less, greater;\n"); /* or */
7651 shader_addline(&buffer
, "DP4 less.b, less, less;\n"); /* on any channel */
7652 shader_addline(&buffer
, "SGE less, -less.b, 0.0;\n"); /* logical not */
7653 shader_addline(&buffer
, "KIL -less;\n"); /* discard if true */
7654 shader_addline(&buffer
, "MOV result.color, color;\n");
7658 shader_addline(&buffer
, "TEX result.color, fragment.texcoord[0], texture[0], %s;\n", tex_target
);
7661 shader_addline(&buffer
, "END\n");
7663 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
7665 string_buffer_free(&buffer
);
7670 /* Context activation is done by the caller. */
7671 static HRESULT
arbfp_blit_set(void *blit_priv
, struct wined3d_context
*context
, const struct wined3d_surface
*surface
,
7672 const struct wined3d_color_key
*color_key
)
7675 float size
[4] = {(float) surface
->pow2Width
, (float) surface
->pow2Height
, 1.0f
, 1.0f
};
7676 const struct wined3d_texture
*texture
= surface
->container
;
7677 struct arbfp_blit_priv
*priv
= blit_priv
;
7678 enum complex_fixup fixup
;
7679 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
7680 struct wine_rb_entry
*entry
;
7681 struct arbfp_blit_type type
;
7682 struct arbfp_blit_desc
*desc
;
7683 struct wined3d_color float_color_key
[2];
7685 if (is_complex_fixup(texture
->resource
.format
->color_fixup
))
7686 fixup
= get_complex_fixup(texture
->resource
.format
->color_fixup
);
7688 fixup
= COMPLEX_FIXUP_NONE
;
7690 switch (texture
->target
)
7693 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_1D
;
7697 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_2D
;
7701 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_3D
;
7704 case GL_TEXTURE_CUBE_MAP_ARB
:
7705 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_CUBE
;
7708 case GL_TEXTURE_RECTANGLE_ARB
:
7709 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_RECT
;
7713 ERR("Unexpected GL texture type %#x.\n", texture
->target
);
7714 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_2D
;
7717 type
.use_color_key
= !!color_key
;
7720 entry
= wine_rb_get(&priv
->shaders
, &type
);
7723 desc
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_blit_desc
, entry
);
7724 shader
= desc
->shader
;
7730 case COMPLEX_FIXUP_NONE
:
7731 if (!is_identity_fixup(texture
->resource
.format
->color_fixup
))
7732 FIXME("Implement support for sign or swizzle fixups.\n");
7733 shader
= arbfp_gen_plain_shader(priv
, gl_info
, &type
);
7736 case COMPLEX_FIXUP_P8
:
7737 shader
= gen_p8_shader(priv
, gl_info
, &type
);
7740 case COMPLEX_FIXUP_YUY2
:
7741 case COMPLEX_FIXUP_UYVY
:
7742 case COMPLEX_FIXUP_YV12
:
7743 case COMPLEX_FIXUP_NV12
:
7744 shader
= gen_yuv_shader(priv
, gl_info
, &type
);
7750 FIXME("Unsupported complex fixup %#x, not setting a shader\n", fixup
);
7754 desc
= HeapAlloc(GetProcessHeap(), 0, sizeof(*desc
));
7759 desc
->shader
= shader
;
7760 if (wine_rb_put(&priv
->shaders
, &desc
->type
, &desc
->entry
) == -1)
7763 ERR("Out of memory\n");
7764 GL_EXTCALL(glDeleteProgramsARB(1, &shader
));
7765 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader))");
7766 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, 0));
7767 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, 0)");
7768 HeapFree(GetProcessHeap(), 0, desc
);
7769 return E_OUTOFMEMORY
;
7773 if (fixup
== COMPLEX_FIXUP_P8
)
7774 upload_palette(texture
, context
);
7776 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
7777 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
7778 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7779 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
7780 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARBFP_BLIT_PARAM_SIZE
, size
));
7781 checkGLcall("glProgramLocalParameter4fvARB");
7782 if (type
.use_color_key
)
7784 wined3d_format_get_float_color_key(texture
->resource
.format
, color_key
, float_color_key
);
7785 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
7786 ARBFP_BLIT_PARAM_COLOR_KEY_LOW
, &float_color_key
[0].r
));
7787 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
7788 ARBFP_BLIT_PARAM_COLOR_KEY_HIGH
, &float_color_key
[1].r
));
7789 checkGLcall("glProgramLocalParameter4fvARB");
7795 /* Context activation is done by the caller. */
7796 static void arbfp_blit_unset(const struct wined3d_gl_info
*gl_info
)
7798 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
7799 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
7802 static BOOL
arbfp_blit_supported(const struct wined3d_gl_info
*gl_info
,
7803 const struct wined3d_d3d_info
*d3d_info
, enum wined3d_blit_op blit_op
,
7804 const RECT
*src_rect
, DWORD src_usage
, enum wined3d_pool src_pool
, const struct wined3d_format
*src_format
,
7805 const RECT
*dst_rect
, DWORD dst_usage
, enum wined3d_pool dst_pool
, const struct wined3d_format
*dst_format
)
7807 enum complex_fixup src_fixup
;
7809 if (!gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
7814 case WINED3D_BLIT_OP_COLOR_BLIT_CKEY
:
7815 if (!d3d_info
->shader_color_key
)
7817 /* The conversion modifies the alpha channel so the color key might no longer match. */
7818 TRACE("Color keying not supported with converted textures.\n");
7821 case WINED3D_BLIT_OP_COLOR_BLIT_ALPHATEST
:
7822 case WINED3D_BLIT_OP_COLOR_BLIT
:
7826 TRACE("Unsupported blit_op=%d\n", blit_op
);
7830 if (src_pool
== WINED3D_POOL_SYSTEM_MEM
|| dst_pool
== WINED3D_POOL_SYSTEM_MEM
)
7833 src_fixup
= get_complex_fixup(src_format
->color_fixup
);
7834 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
7836 TRACE("Checking support for fixup:\n");
7837 dump_color_fixup_desc(src_format
->color_fixup
);
7840 if (!is_identity_fixup(dst_format
->color_fixup
))
7842 TRACE("Destination fixups are not supported\n");
7846 if (is_identity_fixup(src_format
->color_fixup
))
7852 /* We only support YUV conversions. */
7853 if (!is_complex_fixup(src_format
->color_fixup
))
7855 TRACE("[FAILED]\n");
7861 case COMPLEX_FIXUP_YUY2
:
7862 case COMPLEX_FIXUP_UYVY
:
7863 case COMPLEX_FIXUP_YV12
:
7864 case COMPLEX_FIXUP_NV12
:
7865 case COMPLEX_FIXUP_P8
:
7870 FIXME("Unsupported YUV fixup %#x\n", src_fixup
);
7871 TRACE("[FAILED]\n");
7876 static void arbfp_blit_surface(struct wined3d_device
*device
, enum wined3d_blit_op op
, DWORD filter
,
7877 struct wined3d_surface
*src_surface
, const RECT
*src_rect_in
,
7878 struct wined3d_surface
*dst_surface
, const RECT
*dst_rect_in
,
7879 const struct wined3d_color_key
*color_key
)
7881 unsigned int dst_sub_resource_idx
= surface_get_sub_resource_idx(dst_surface
);
7882 struct wined3d_texture
*src_texture
= src_surface
->container
;
7883 struct wined3d_texture
*dst_texture
= dst_surface
->container
;
7884 struct wined3d_context
*context
;
7885 RECT src_rect
= *src_rect_in
;
7886 RECT dst_rect
= *dst_rect_in
;
7887 struct wined3d_color_key alpha_test_key
;
7889 /* Activate the destination context, set it up for blitting */
7890 context
= context_acquire(device
, dst_surface
);
7892 /* Now load the surface */
7893 if (wined3d_settings
.offscreen_rendering_mode
!= ORM_FBO
7894 && (surface_get_sub_resource(src_surface
)->locations
7895 & (WINED3D_LOCATION_TEXTURE_RGB
| WINED3D_LOCATION_DRAWABLE
))
7896 == WINED3D_LOCATION_DRAWABLE
7897 && !wined3d_resource_is_offscreen(&src_texture
->resource
))
7899 /* Without FBO blits transferring from the drawable to the texture is
7900 * expensive, because we have to flip the data in sysmem. Since we can
7901 * flip in the blitter, we don't actually need that flip anyway. So we
7902 * use the surface's texture as scratch texture, and flip the source
7903 * rectangle instead. */
7904 surface_load_fb_texture(src_surface
, FALSE
, context
);
7906 src_rect
.top
= wined3d_texture_get_level_height(src_texture
, src_surface
->texture_level
) - src_rect
.top
;
7907 src_rect
.bottom
= wined3d_texture_get_level_height(src_texture
, src_surface
->texture_level
) - src_rect
.bottom
;
7910 wined3d_texture_load(src_texture
, context
, FALSE
);
7912 context_apply_blit_state(context
, device
);
7914 if (!wined3d_resource_is_offscreen(&dst_texture
->resource
))
7915 surface_translate_drawable_coords(dst_surface
, context
->win_handle
, &dst_rect
);
7917 if (op
== WINED3D_BLIT_OP_COLOR_BLIT_ALPHATEST
)
7919 const struct wined3d_format
*fmt
= src_texture
->resource
.format
;
7920 alpha_test_key
.color_space_low_value
= 0;
7921 alpha_test_key
.color_space_high_value
= ~(((1u << fmt
->alpha_size
) - 1) << fmt
->alpha_offset
);
7922 color_key
= &alpha_test_key
;
7925 arbfp_blit_set(device
->blit_priv
, context
, src_surface
, color_key
);
7927 /* Draw a textured quad */
7928 draw_textured_quad(src_surface
, context
, &src_rect
, &dst_rect
, filter
);
7930 /* Leave the opengl state valid for blitting */
7931 arbfp_blit_unset(context
->gl_info
);
7933 if (wined3d_settings
.strict_draw_ordering
7934 || (dst_texture
->swapchain
&& (dst_texture
->swapchain
->front_buffer
== dst_texture
)))
7935 context
->gl_info
->gl_ops
.gl
.p_glFlush(); /* Flush to ensure ordering across contexts. */
7937 context_release(context
);
7939 wined3d_texture_validate_location(dst_texture
, dst_sub_resource_idx
, dst_texture
->resource
.draw_binding
);
7940 wined3d_texture_invalidate_location(dst_texture
, dst_sub_resource_idx
, ~dst_texture
->resource
.draw_binding
);
7943 static HRESULT
arbfp_blit_color_fill(struct wined3d_device
*device
, struct wined3d_rendertarget_view
*view
,
7944 const RECT
*rect
, const struct wined3d_color
*color
)
7946 FIXME("Color filling not implemented by arbfp_blit\n");
7947 return WINED3DERR_INVALIDCALL
;
7950 static HRESULT
arbfp_blit_depth_fill(struct wined3d_device
*device
, struct wined3d_rendertarget_view
*view
,
7951 const RECT
*rect
, DWORD clear_flags
, float depth
, DWORD stencil
)
7953 FIXME("Depth/stencil filling not implemented by arbfp_blit.\n");
7954 return WINED3DERR_INVALIDCALL
;
7957 const struct blit_shader arbfp_blit
= {
7962 arbfp_blit_supported
,
7963 arbfp_blit_color_fill
,
7964 arbfp_blit_depth_fill
,