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_GEOMETRY
)
84 ERR("Geometry shaders are unsupported\n");
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
-= count_bits(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
] & (1 << shift
)) highest_constf
= i
;
811 if(use_nv_clip(gl_info
) && ctx
->target_version
>= NV2
)
813 if(ctx
->cur_vs_args
->super
.clip_enabled
)
814 clip_limit
= gl_info
->limits
.clipplanes
;
820 unsigned int mask
= ctx
->cur_vs_args
->clip
.boolclip
.clipplane_mask
;
821 clip_limit
= min(count_bits(mask
), 4);
823 *num_clipplanes
= min(clip_limit
, max_constantsF
- highest_constf
- 1);
824 max_constantsF
-= *num_clipplanes
;
825 if(*num_clipplanes
< clip_limit
)
827 WARN("Only %u clipplanes out of %u enabled\n", *num_clipplanes
, gl_info
->limits
.clipplanes
);
832 if (ctx
->target_version
>= NV2
) *num_clipplanes
= gl_info
->limits
.clipplanes
;
833 else *num_clipplanes
= min(gl_info
->limits
.clipplanes
, 4);
837 for (i
= 0, map
= reg_maps
->temporary
; map
; map
>>= 1, ++i
)
839 if (map
& 1) shader_addline(buffer
, "TEMP R%u;\n", i
);
842 for (i
= 0, map
= reg_maps
->address
; map
; map
>>= 1, ++i
)
844 if (map
& 1) shader_addline(buffer
, "ADDRESS A%u;\n", i
);
847 if (pshader
&& reg_maps
->shader_version
.major
== 1 && reg_maps
->shader_version
.minor
<= 3)
849 for (i
= 0, map
= reg_maps
->texcoord
; map
; map
>>= 1, ++i
)
851 if (map
& 1) shader_addline(buffer
, "TEMP T%u;\n", i
);
855 if (!shader
->load_local_constsF
)
857 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
860 value
= (const float *)lconst
->value
;
861 shader_addline(buffer
, "PARAM C%u = ", lconst
->idx
);
862 shader_arb_append_imm_vec4(buffer
, value
);
863 shader_addline(buffer
, ";\n");
867 /* After subtracting privately used constants from the hardware limit(they are loaded as
868 * local constants), make sure the shader doesn't violate the env constant limit
872 max_constantsF
= min(max_constantsF
, gl_info
->limits
.arb_ps_float_constants
);
876 max_constantsF
= min(max_constantsF
, gl_info
->limits
.arb_vs_float_constants
);
879 /* Avoid declaring more constants than needed */
880 max_constantsF
= min(max_constantsF
, shader
->limits
->constant_float
);
882 /* we use the array-based constants array if the local constants are marked for loading,
883 * because then we use indirect addressing, or when the local constant list is empty,
884 * because then we don't know if we're using indirect addressing or not. If we're hardcoding
885 * local constants do not declare the loaded constants as an array because ARB compilers usually
886 * do not optimize unused constants away
888 if (reg_maps
->usesrelconstF
)
890 /* Need to PARAM the environment parameters (constants) so we can use relative addressing */
891 shader_addline(buffer
, "PARAM C[%d] = { program.env[0..%d] };\n",
892 max_constantsF
, max_constantsF
- 1);
894 for(i
= 0; i
< max_constantsF
; i
++) {
897 mask
= 1 << (i
& 0x1f);
898 if (!shader_constant_is_local(shader
, i
) && (reg_maps
->constf
[idx
] & mask
))
900 shader_addline(buffer
, "PARAM C%d = program.env[%d];\n",i
, i
);
906 static const char * const shift_tab
[] = {
907 "dummy", /* 0 (none) */
908 "coefmul.x", /* 1 (x2) */
909 "coefmul.y", /* 2 (x4) */
910 "coefmul.z", /* 3 (x8) */
911 "coefmul.w", /* 4 (x16) */
912 "dummy", /* 5 (x32) */
913 "dummy", /* 6 (x64) */
914 "dummy", /* 7 (x128) */
915 "dummy", /* 8 (d256) */
916 "dummy", /* 9 (d128) */
917 "dummy", /* 10 (d64) */
918 "dummy", /* 11 (d32) */
919 "coefdiv.w", /* 12 (d16) */
920 "coefdiv.z", /* 13 (d8) */
921 "coefdiv.y", /* 14 (d4) */
922 "coefdiv.x" /* 15 (d2) */
925 static void shader_arb_get_write_mask(const struct wined3d_shader_instruction
*ins
,
926 const struct wined3d_shader_dst_param
*dst
, char *write_mask
)
928 char *ptr
= write_mask
;
930 if (dst
->write_mask
!= WINED3DSP_WRITEMASK_ALL
)
933 if (dst
->write_mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
934 if (dst
->write_mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
935 if (dst
->write_mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
936 if (dst
->write_mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
942 static void shader_arb_get_swizzle(const struct wined3d_shader_src_param
*param
, BOOL fixup
, char *swizzle_str
)
944 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
945 * but addressed as "rgba". To fix this we need to swap the register's x
946 * and z components. */
947 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
948 char *ptr
= swizzle_str
;
950 /* swizzle bits fields: wwzzyyxx */
951 DWORD swizzle
= param
->swizzle
;
952 DWORD swizzle_x
= swizzle
& 0x03;
953 DWORD swizzle_y
= (swizzle
>> 2) & 0x03;
954 DWORD swizzle_z
= (swizzle
>> 4) & 0x03;
955 DWORD swizzle_w
= (swizzle
>> 6) & 0x03;
957 /* If the swizzle is the default swizzle (ie, "xyzw"), we don't need to
958 * generate a swizzle string. Unless we need to our own swizzling. */
959 if (swizzle
!= WINED3DSP_NOSWIZZLE
|| fixup
)
962 if (swizzle_x
== swizzle_y
&& swizzle_x
== swizzle_z
&& swizzle_x
== swizzle_w
) {
963 *ptr
++ = swizzle_chars
[swizzle_x
];
965 *ptr
++ = swizzle_chars
[swizzle_x
];
966 *ptr
++ = swizzle_chars
[swizzle_y
];
967 *ptr
++ = swizzle_chars
[swizzle_z
];
968 *ptr
++ = swizzle_chars
[swizzle_w
];
975 static void shader_arb_request_a0(const struct wined3d_shader_instruction
*ins
, const char *src
)
977 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
978 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
980 if (!strcmp(priv
->addr_reg
, src
)) return;
982 strcpy(priv
->addr_reg
, src
);
983 shader_addline(buffer
, "ARL A0.x, %s;\n", src
);
986 static void shader_arb_get_src_param(const struct wined3d_shader_instruction
*ins
,
987 const struct wined3d_shader_src_param
*src
, unsigned int tmpreg
, char *outregstr
);
989 static void shader_arb_get_register_name(const struct wined3d_shader_instruction
*ins
,
990 const struct wined3d_shader_register
*reg
, char *register_name
, BOOL
*is_color
)
992 /* oPos, oFog and oPts in D3D */
993 static const char * const rastout_reg_names
[] = {"TMP_OUT", "TMP_FOGCOORD", "result.pointsize"};
994 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
995 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
996 BOOL pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
997 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1003 case WINED3DSPR_TEMP
:
1004 sprintf(register_name
, "R%u", reg
->idx
[0].offset
);
1007 case WINED3DSPR_INPUT
:
1010 if (reg_maps
->shader_version
.major
< 3)
1012 if (!reg
->idx
[0].offset
)
1013 strcpy(register_name
, "fragment.color.primary");
1015 strcpy(register_name
, "fragment.color.secondary");
1019 if (reg
->idx
[0].rel_addr
)
1022 shader_arb_get_src_param(ins
, reg
->idx
[0].rel_addr
, 0, rel_reg
);
1024 if (!strcmp(rel_reg
, "**aL_emul**"))
1026 DWORD idx
= ctx
->aL
+ reg
->idx
[0].offset
;
1027 if(idx
< MAX_REG_INPUT
)
1029 strcpy(register_name
, ctx
->ps_input
[idx
]);
1033 ERR("Pixel shader input register out of bounds: %u\n", idx
);
1034 sprintf(register_name
, "out_of_bounds_%u", idx
);
1037 else if (reg_maps
->input_registers
& 0x0300)
1039 /* There are two ways basically:
1041 * 1) Use the unrolling code that is used for loop emulation and unroll the loop.
1042 * That means trouble if the loop also contains a breakc or if the control values
1043 * aren't local constants.
1044 * 2) Generate an if block that checks if aL.y < 8, == 8 or == 9 and selects the
1045 * source dynamically. The trouble is that we cannot simply read aL.y because it
1046 * is an ADDRESS register. We could however push it, load .zw with a value and use
1047 * ADAC to load the condition code register and pop it again afterwards
1049 FIXME("Relative input register addressing with more than 8 registers\n");
1051 /* This is better than nothing for now */
1052 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1054 else if(ctx
->cur_ps_args
->super
.vp_mode
!= vertexshader
)
1056 /* This is problematic because we'd have to consult the ctx->ps_input strings
1057 * for where to find the varying. Some may be "0.0", others can be texcoords or
1058 * colors. This needs either a pipeline replacement to make the vertex shader feed
1059 * proper varyings, or loop unrolling
1061 * For now use the texcoords and hope for the best
1063 FIXME("Non-vertex shader varying input with indirect addressing\n");
1064 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1068 /* D3D supports indirect addressing only with aL in loop registers. The loop instruction
1069 * pulls GL_NV_fragment_program2 in
1071 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1076 if (reg
->idx
[0].offset
< MAX_REG_INPUT
)
1078 strcpy(register_name
, ctx
->ps_input
[reg
->idx
[0].offset
]);
1082 ERR("Pixel shader input register out of bounds: %u\n", reg
->idx
[0].offset
);
1083 sprintf(register_name
, "out_of_bounds_%u", reg
->idx
[0].offset
);
1090 if (ctx
->cur_vs_args
->super
.swizzle_map
& (1 << reg
->idx
[0].offset
))
1092 sprintf(register_name
, "vertex.attrib[%u]", reg
->idx
[0].offset
);
1096 case WINED3DSPR_CONST
:
1097 if (!pshader
&& reg
->idx
[0].rel_addr
)
1099 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1100 UINT rel_offset
= ctx
->target_version
== ARB
? shader_data
->rel_offset
: 0;
1103 if (reg_maps
->shader_version
.major
< 2)
1105 sprintf(rel_reg
, "A0.x");
1109 shader_arb_get_src_param(ins
, reg
->idx
[0].rel_addr
, 0, rel_reg
);
1110 if (ctx
->target_version
== ARB
)
1112 if (!strcmp(rel_reg
, "**aL_emul**"))
1116 shader_arb_request_a0(ins
, rel_reg
);
1117 sprintf(rel_reg
, "A0.x");
1122 sprintf(register_name
, "C[%u]", ctx
->aL
+ reg
->idx
[0].offset
);
1123 else if (reg
->idx
[0].offset
>= rel_offset
)
1124 sprintf(register_name
, "C[%s + %u]", rel_reg
, reg
->idx
[0].offset
- rel_offset
);
1126 sprintf(register_name
, "C[%s - %u]", rel_reg
, rel_offset
- reg
->idx
[0].offset
);
1130 if (reg_maps
->usesrelconstF
)
1131 sprintf(register_name
, "C[%u]", reg
->idx
[0].offset
);
1133 sprintf(register_name
, "C%u", reg
->idx
[0].offset
);
1137 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
1140 if (reg_maps
->shader_version
.major
== 1
1141 && reg_maps
->shader_version
.minor
<= 3)
1142 /* In ps <= 1.3, Tx is a temporary register as destination
1143 * to all instructions, and as source to most instructions.
1144 * For some instructions it is the texcoord input. Those
1145 * instructions know about the special use. */
1146 sprintf(register_name
, "T%u", reg
->idx
[0].offset
);
1148 /* In ps 1.4 and 2.x Tx is always a (read-only) varying. */
1149 sprintf(register_name
, "fragment.texcoord[%u]", reg
->idx
[0].offset
);
1153 if (reg_maps
->shader_version
.major
== 1 || ctx
->target_version
>= NV2
)
1154 sprintf(register_name
, "A%u", reg
->idx
[0].offset
);
1156 sprintf(register_name
, "A%u_SHADOW", reg
->idx
[0].offset
);
1160 case WINED3DSPR_COLOROUT
:
1161 if (ctx
->ps_post_process
&& !reg
->idx
[0].offset
)
1163 strcpy(register_name
, "TMP_COLOR");
1167 if (ctx
->cur_ps_args
->super
.srgb_correction
)
1168 FIXME("sRGB correction on higher render targets.\n");
1169 if (reg_maps
->rt_mask
> 1)
1170 sprintf(register_name
, "result.color[%u]", reg
->idx
[0].offset
);
1172 strcpy(register_name
, "result.color");
1176 case WINED3DSPR_RASTOUT
:
1177 if (reg
->idx
[0].offset
== 1)
1178 sprintf(register_name
, "%s", ctx
->fog_output
);
1180 sprintf(register_name
, "%s", rastout_reg_names
[reg
->idx
[0].offset
]);
1183 case WINED3DSPR_DEPTHOUT
:
1184 strcpy(register_name
, "result.depth");
1187 case WINED3DSPR_ATTROUT
:
1188 /* case WINED3DSPR_OUTPUT: */
1190 sprintf(register_name
, "oD[%u]", reg
->idx
[0].offset
);
1192 strcpy(register_name
, ctx
->color_output
[reg
->idx
[0].offset
]);
1195 case WINED3DSPR_TEXCRDOUT
:
1197 sprintf(register_name
, "oT[%u]", reg
->idx
[0].offset
);
1198 else if (reg_maps
->shader_version
.major
< 3)
1199 strcpy(register_name
, ctx
->texcrd_output
[reg
->idx
[0].offset
]);
1201 strcpy(register_name
, ctx
->vs_output
[reg
->idx
[0].offset
]);
1204 case WINED3DSPR_LOOP
:
1205 if(ctx
->target_version
>= NV2
)
1207 /* Pshader has an implicitly declared loop index counter A0.x that cannot be renamed */
1208 if(pshader
) sprintf(register_name
, "A0.x");
1209 else sprintf(register_name
, "aL.y");
1213 /* Unfortunately this code cannot return the value of ctx->aL here. An immediate value
1214 * would be valid, but if aL is used for indexing(its only use), there's likely an offset,
1215 * thus the result would be something like C[15 + 30], which is not valid in the ARB program
1216 * grammar. So return a marker for the emulated aL and intercept it in constant and varying
1219 sprintf(register_name
, "**aL_emul**");
1224 case WINED3DSPR_CONSTINT
:
1225 sprintf(register_name
, "I%u", reg
->idx
[0].offset
);
1228 case WINED3DSPR_MISCTYPE
:
1229 if (!reg
->idx
[0].offset
)
1230 sprintf(register_name
, "vpos");
1231 else if (reg
->idx
[0].offset
== 1)
1232 sprintf(register_name
, "fragment.facing.x");
1234 FIXME("Unknown MISCTYPE register index %u.\n", reg
->idx
[0].offset
);
1238 FIXME("Unhandled register type %#x[%u].\n", reg
->type
, reg
->idx
[0].offset
);
1239 sprintf(register_name
, "unrecognized_register[%u]", reg
->idx
[0].offset
);
1244 static void shader_arb_get_dst_param(const struct wined3d_shader_instruction
*ins
,
1245 const struct wined3d_shader_dst_param
*wined3d_dst
, char *str
)
1247 char register_name
[255];
1251 shader_arb_get_register_name(ins
, &wined3d_dst
->reg
, register_name
, &is_color
);
1252 strcpy(str
, register_name
);
1254 shader_arb_get_write_mask(ins
, wined3d_dst
, write_mask
);
1255 strcat(str
, write_mask
);
1258 static const char *shader_arb_get_fixup_swizzle(enum fixup_channel_source channel_source
)
1260 switch(channel_source
)
1262 case CHANNEL_SOURCE_ZERO
: return "0";
1263 case CHANNEL_SOURCE_ONE
: return "1";
1264 case CHANNEL_SOURCE_X
: return "x";
1265 case CHANNEL_SOURCE_Y
: return "y";
1266 case CHANNEL_SOURCE_Z
: return "z";
1267 case CHANNEL_SOURCE_W
: return "w";
1269 FIXME("Unhandled channel source %#x\n", channel_source
);
1274 struct color_fixup_masks
1280 static struct color_fixup_masks
calc_color_correction(struct color_fixup_desc fixup
, DWORD dst_mask
)
1282 struct color_fixup_masks masks
= {0, 0};
1284 if (is_complex_fixup(fixup
))
1286 enum complex_fixup complex_fixup
= get_complex_fixup(fixup
);
1287 FIXME("Complex fixup (%#x) not supported\n", complex_fixup
);
1291 if (fixup
.x_source
!= CHANNEL_SOURCE_X
)
1292 masks
.source
|= WINED3DSP_WRITEMASK_0
;
1293 if (fixup
.y_source
!= CHANNEL_SOURCE_Y
)
1294 masks
.source
|= WINED3DSP_WRITEMASK_1
;
1295 if (fixup
.z_source
!= CHANNEL_SOURCE_Z
)
1296 masks
.source
|= WINED3DSP_WRITEMASK_2
;
1297 if (fixup
.w_source
!= CHANNEL_SOURCE_W
)
1298 masks
.source
|= WINED3DSP_WRITEMASK_3
;
1299 masks
.source
&= dst_mask
;
1301 if (fixup
.x_sign_fixup
)
1302 masks
.sign
|= WINED3DSP_WRITEMASK_0
;
1303 if (fixup
.y_sign_fixup
)
1304 masks
.sign
|= WINED3DSP_WRITEMASK_1
;
1305 if (fixup
.z_sign_fixup
)
1306 masks
.sign
|= WINED3DSP_WRITEMASK_2
;
1307 if (fixup
.w_sign_fixup
)
1308 masks
.sign
|= WINED3DSP_WRITEMASK_3
;
1309 masks
.sign
&= dst_mask
;
1314 static void gen_color_correction(struct wined3d_string_buffer
*buffer
, const char *dst
,
1315 const char *src
, const char *one
, const char *two
,
1316 struct color_fixup_desc fixup
, struct color_fixup_masks masks
)
1318 const char *sign_fixup_src
= dst
;
1323 sign_fixup_src
= "TA";
1325 shader_addline(buffer
, "SWZ %s, %s, %s, %s, %s, %s;\n", sign_fixup_src
, src
,
1326 shader_arb_get_fixup_swizzle(fixup
.x_source
), shader_arb_get_fixup_swizzle(fixup
.y_source
),
1327 shader_arb_get_fixup_swizzle(fixup
.z_source
), shader_arb_get_fixup_swizzle(fixup
.w_source
));
1329 else if (masks
.sign
)
1331 sign_fixup_src
= src
;
1337 char *ptr
= reg_mask
;
1339 if (masks
.sign
!= WINED3DSP_WRITEMASK_ALL
)
1342 if (masks
.sign
& WINED3DSP_WRITEMASK_0
)
1344 if (masks
.sign
& WINED3DSP_WRITEMASK_1
)
1346 if (masks
.sign
& WINED3DSP_WRITEMASK_2
)
1348 if (masks
.sign
& WINED3DSP_WRITEMASK_3
)
1353 shader_addline(buffer
, "MAD %s%s, %s, %s, -%s;\n", dst
, reg_mask
, sign_fixup_src
, two
, one
);
1357 static const char *shader_arb_get_modifier(const struct wined3d_shader_instruction
*ins
)
1360 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1361 if (!ins
->dst_count
) return "";
1363 mod
= ins
->dst
[0].modifiers
;
1365 /* Silently ignore PARTIALPRECISION if its not supported */
1366 if(priv
->target_version
== ARB
) mod
&= ~WINED3DSPDM_PARTIALPRECISION
;
1368 if(mod
& WINED3DSPDM_MSAMPCENTROID
)
1370 FIXME("Unhandled modifier WINED3DSPDM_MSAMPCENTROID\n");
1371 mod
&= ~WINED3DSPDM_MSAMPCENTROID
;
1376 case WINED3DSPDM_SATURATE
| WINED3DSPDM_PARTIALPRECISION
:
1379 case WINED3DSPDM_SATURATE
:
1382 case WINED3DSPDM_PARTIALPRECISION
:
1389 FIXME("Unknown modifiers 0x%08x\n", mod
);
1394 #define TEX_PROJ 0x1
1395 #define TEX_BIAS 0x2
1397 #define TEX_DERIV 0x10
1399 static void shader_hw_sample(const struct wined3d_shader_instruction
*ins
, DWORD sampler_idx
,
1400 const char *dst_str
, const char *coord_reg
, WORD flags
, const char *dsx
, const char *dsy
)
1402 enum wined3d_shader_resource_type resource_type
= ins
->ctx
->reg_maps
->resource_info
[sampler_idx
].type
;
1403 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1404 const char *tex_type
;
1405 BOOL np2_fixup
= FALSE
;
1406 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1408 BOOL pshader
= shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
1409 const struct wined3d_shader
*shader
;
1410 const struct wined3d_device
*device
;
1411 const struct wined3d_gl_info
*gl_info
;
1412 const char *tex_dst
= dst_str
;
1413 struct color_fixup_masks masks
;
1415 /* D3D vertex shader sampler IDs are vertex samplers(0-3), not global d3d samplers */
1416 if(!pshader
) sampler_idx
+= MAX_FRAGMENT_SAMPLERS
;
1418 switch (resource_type
)
1420 case WINED3D_SHADER_RESOURCE_TEXTURE_1D
:
1424 case WINED3D_SHADER_RESOURCE_TEXTURE_2D
:
1425 shader
= ins
->ctx
->shader
;
1426 device
= shader
->device
;
1427 gl_info
= &device
->adapter
->gl_info
;
1429 if (pshader
&& priv
->cur_ps_args
->super
.np2_fixup
& (1 << sampler_idx
)
1430 && gl_info
->supported
[ARB_TEXTURE_RECTANGLE
])
1434 if (shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
))
1436 if (priv
->cur_np2fixup_info
->super
.active
& (1 << sampler_idx
))
1438 if (flags
) FIXME("Only ordinary sampling from NP2 textures is supported.\n");
1439 else np2_fixup
= TRUE
;
1444 case WINED3D_SHADER_RESOURCE_TEXTURE_3D
:
1448 case WINED3D_SHADER_RESOURCE_TEXTURE_CUBE
:
1453 ERR("Unexpected resource type %#x.\n", resource_type
);
1457 /* TEX, TXL, TXD and TXP do not support the "H" modifier,
1458 * so don't use shader_arb_get_modifier
1460 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
) mod
= "_SAT";
1463 /* Fragment samplers always have indentity mapping */
1464 if(sampler_idx
>= MAX_FRAGMENT_SAMPLERS
)
1466 sampler_idx
= priv
->cur_vs_args
->vertex
.samplers
[sampler_idx
- MAX_FRAGMENT_SAMPLERS
];
1471 masks
= calc_color_correction(priv
->cur_ps_args
->super
.color_fixup
[sampler_idx
],
1472 ins
->dst
[0].write_mask
);
1474 if (masks
.source
|| masks
.sign
)
1478 if (flags
& TEX_DERIV
)
1480 if(flags
& TEX_PROJ
) FIXME("Projected texture sampling with custom derivatives\n");
1481 if(flags
& TEX_BIAS
) FIXME("Biased texture sampling with custom derivatives\n");
1482 shader_addline(buffer
, "TXD%s %s, %s, %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
,
1483 dsx
, dsy
, sampler_idx
, tex_type
);
1485 else if(flags
& TEX_LOD
)
1487 if(flags
& TEX_PROJ
) FIXME("Projected texture sampling with explicit lod\n");
1488 if(flags
& TEX_BIAS
) FIXME("Biased texture sampling with explicit lod\n");
1489 shader_addline(buffer
, "TXL%s %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
,
1490 sampler_idx
, tex_type
);
1492 else if (flags
& TEX_BIAS
)
1494 /* Shouldn't be possible, but let's check for it */
1495 if(flags
& TEX_PROJ
) FIXME("Biased and Projected texture sampling\n");
1496 /* TXB takes the 4th component of the source vector automatically, as d3d. Nothing more to do */
1497 shader_addline(buffer
, "TXB%s %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
, sampler_idx
, tex_type
);
1499 else if (flags
& TEX_PROJ
)
1501 shader_addline(buffer
, "TXP%s %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
, sampler_idx
, tex_type
);
1507 const unsigned char idx
= priv
->cur_np2fixup_info
->super
.idx
[sampler_idx
];
1508 shader_addline(buffer
, "MUL TA, np2fixup[%u].%s, %s;\n", idx
>> 1,
1509 (idx
% 2) ? "zwxy" : "xyzw", coord_reg
);
1511 shader_addline(buffer
, "TEX%s %s, TA, texture[%u], %s;\n", mod
, tex_dst
, sampler_idx
, tex_type
);
1514 shader_addline(buffer
, "TEX%s %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
, sampler_idx
, tex_type
);
1519 gen_color_correction(buffer
, dst_str
, tex_dst
,
1520 arb_get_helper_value(WINED3D_SHADER_TYPE_PIXEL
, ARB_ONE
),
1521 arb_get_helper_value(WINED3D_SHADER_TYPE_PIXEL
, ARB_TWO
),
1522 priv
->cur_ps_args
->super
.color_fixup
[sampler_idx
], masks
);
1526 static void shader_arb_get_src_param(const struct wined3d_shader_instruction
*ins
,
1527 const struct wined3d_shader_src_param
*src
, unsigned int tmpreg
, char *outregstr
)
1529 /* Generate a line that does the input modifier computation and return the input register to use */
1530 BOOL is_color
= FALSE
, insert_line
;
1533 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1534 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1535 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
1536 const char *two
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_TWO
);
1538 /* Assume a new line will be added */
1541 /* Get register name */
1542 shader_arb_get_register_name(ins
, &src
->reg
, regstr
, &is_color
);
1543 shader_arb_get_swizzle(src
, is_color
, swzstr
);
1545 switch (src
->modifiers
)
1547 case WINED3DSPSM_NONE
:
1548 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
1549 insert_line
= FALSE
;
1551 case WINED3DSPSM_NEG
:
1552 sprintf(outregstr
, "-%s%s", regstr
, swzstr
);
1553 insert_line
= FALSE
;
1555 case WINED3DSPSM_BIAS
:
1556 shader_addline(buffer
, "ADD T%c, %s, -coefdiv.x;\n", 'A' + tmpreg
, regstr
);
1558 case WINED3DSPSM_BIASNEG
:
1559 shader_addline(buffer
, "ADD T%c, -%s, coefdiv.x;\n", 'A' + tmpreg
, regstr
);
1561 case WINED3DSPSM_SIGN
:
1562 shader_addline(buffer
, "MAD T%c, %s, %s, -%s;\n", 'A' + tmpreg
, regstr
, two
, one
);
1564 case WINED3DSPSM_SIGNNEG
:
1565 shader_addline(buffer
, "MAD T%c, %s, -%s, %s;\n", 'A' + tmpreg
, regstr
, two
, one
);
1567 case WINED3DSPSM_COMP
:
1568 shader_addline(buffer
, "SUB T%c, %s, %s;\n", 'A' + tmpreg
, one
, regstr
);
1570 case WINED3DSPSM_X2
:
1571 shader_addline(buffer
, "ADD T%c, %s, %s;\n", 'A' + tmpreg
, regstr
, regstr
);
1573 case WINED3DSPSM_X2NEG
:
1574 shader_addline(buffer
, "ADD T%c, -%s, -%s;\n", 'A' + tmpreg
, regstr
, regstr
);
1576 case WINED3DSPSM_DZ
:
1577 shader_addline(buffer
, "RCP T%c, %s.z;\n", 'A' + tmpreg
, regstr
);
1578 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
1580 case WINED3DSPSM_DW
:
1581 shader_addline(buffer
, "RCP T%c, %s.w;\n", 'A' + tmpreg
, regstr
);
1582 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
1584 case WINED3DSPSM_ABS
:
1585 if(ctx
->target_version
>= NV2
) {
1586 sprintf(outregstr
, "|%s%s|", regstr
, swzstr
);
1587 insert_line
= FALSE
;
1589 shader_addline(buffer
, "ABS T%c, %s;\n", 'A' + tmpreg
, regstr
);
1592 case WINED3DSPSM_ABSNEG
:
1593 if(ctx
->target_version
>= NV2
) {
1594 sprintf(outregstr
, "-|%s%s|", regstr
, swzstr
);
1596 shader_addline(buffer
, "ABS T%c, %s;\n", 'A' + tmpreg
, regstr
);
1597 sprintf(outregstr
, "-T%c%s", 'A' + tmpreg
, swzstr
);
1599 insert_line
= FALSE
;
1602 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
1603 insert_line
= FALSE
;
1606 /* Return modified or original register, with swizzle */
1608 sprintf(outregstr
, "T%c%s", 'A' + tmpreg
, swzstr
);
1611 static void pshader_hw_bem(const struct wined3d_shader_instruction
*ins
)
1613 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1614 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1615 DWORD sampler_code
= dst
->reg
.idx
[0].offset
;
1617 char src_name
[2][50];
1619 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1621 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed
1623 * Keep in mind that src_name[1] can be "TB" and src_name[0] can be "TA" because modifiers like _x2 are valid
1624 * with bem. So delay loading the first parameter until after the perturbation calculation which needs two
1627 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1628 shader_addline(buffer
, "SWZ TA, bumpenvmat%d, x, z, 0, 0;\n", sampler_code
);
1629 shader_addline(buffer
, "DP3 TC.r, TA, %s;\n", src_name
[1]);
1630 shader_addline(buffer
, "SWZ TA, bumpenvmat%d, y, w, 0, 0;\n", sampler_code
);
1631 shader_addline(buffer
, "DP3 TC.g, TA, %s;\n", src_name
[1]);
1633 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1634 shader_addline(buffer
, "ADD %s, %s, TC;\n", dst_name
, src_name
[0]);
1637 static DWORD
negate_modifiers(DWORD mod
, char *extra_char
)
1642 case WINED3DSPSM_NONE
: return WINED3DSPSM_NEG
;
1643 case WINED3DSPSM_NEG
: return WINED3DSPSM_NONE
;
1644 case WINED3DSPSM_BIAS
: return WINED3DSPSM_BIASNEG
;
1645 case WINED3DSPSM_BIASNEG
: return WINED3DSPSM_BIAS
;
1646 case WINED3DSPSM_SIGN
: return WINED3DSPSM_SIGNNEG
;
1647 case WINED3DSPSM_SIGNNEG
: return WINED3DSPSM_SIGN
;
1648 case WINED3DSPSM_COMP
: *extra_char
= '-'; return WINED3DSPSM_COMP
;
1649 case WINED3DSPSM_X2
: return WINED3DSPSM_X2NEG
;
1650 case WINED3DSPSM_X2NEG
: return WINED3DSPSM_X2
;
1651 case WINED3DSPSM_DZ
: *extra_char
= '-'; return WINED3DSPSM_DZ
;
1652 case WINED3DSPSM_DW
: *extra_char
= '-'; return WINED3DSPSM_DW
;
1653 case WINED3DSPSM_ABS
: return WINED3DSPSM_ABSNEG
;
1654 case WINED3DSPSM_ABSNEG
: return WINED3DSPSM_ABS
;
1656 FIXME("Unknown modifier %u\n", mod
);
1660 static void pshader_hw_cnd(const struct wined3d_shader_instruction
*ins
)
1662 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1663 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1665 char src_name
[3][50];
1666 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1667 ins
->ctx
->reg_maps
->shader_version
.minor
);
1669 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1670 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1672 if (shader_version
<= WINED3D_SHADER_VERSION(1, 3) && ins
->coissue
1673 && ins
->dst
->write_mask
!= WINED3DSP_WRITEMASK_3
)
1675 shader_addline(buffer
, "MOV%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
[1]);
1679 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
1682 /* src0 may have a negate srcmod set, so we can't blindly add "-" to the name */
1683 src0_copy
.modifiers
= negate_modifiers(src0_copy
.modifiers
, &extra_neg
);
1685 shader_arb_get_src_param(ins
, &src0_copy
, 0, src_name
[0]);
1686 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1687 shader_addline(buffer
, "ADD TA, %c%s, coefdiv.x;\n", extra_neg
, src_name
[0]);
1688 shader_addline(buffer
, "CMP%s %s, TA, %s, %s;\n", shader_arb_get_modifier(ins
),
1689 dst_name
, src_name
[1], src_name
[2]);
1693 static void pshader_hw_cmp(const struct wined3d_shader_instruction
*ins
)
1695 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1696 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1698 char src_name
[3][50];
1700 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1702 /* Generate input register names (with modifiers) */
1703 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1704 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1705 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1707 shader_addline(buffer
, "CMP%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1708 dst_name
, src_name
[0], src_name
[2], src_name
[1]);
1711 /** Process the WINED3DSIO_DP2ADD instruction in ARB.
1712 * dst = dot2(src0, src1) + src2 */
1713 static void pshader_hw_dp2add(const struct wined3d_shader_instruction
*ins
)
1715 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1716 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1718 char src_name
[3][50];
1719 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1721 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1722 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1723 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1725 if(ctx
->target_version
>= NV3
)
1727 /* GL_NV_fragment_program2 has a 1:1 matching instruction */
1728 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1729 shader_addline(buffer
, "DP2A%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1730 dst_name
, src_name
[0], src_name
[1], src_name
[2]);
1732 else if(ctx
->target_version
>= NV2
)
1734 /* dst.x = src2.?, src0.x, src1.x + src0.y * src1.y
1735 * dst.y = src2.?, src0.x, src1.z + src0.y * src1.w
1736 * dst.z = src2.?, src0.x, src1.x + src0.y * src1.y
1737 * dst.z = src2.?, src0.x, src1.z + src0.y * src1.w
1739 * Make sure that src1.zw = src1.xy, then we get a classic dp2add
1741 * .xyxy and other swizzles that we could get with this are not valid in
1742 * plain ARBfp, but luckily the NV extension grammar lifts this limitation.
1744 struct wined3d_shader_src_param tmp_param
= ins
->src
[1];
1745 DWORD swizzle
= tmp_param
.swizzle
& 0xf; /* Selects .xy */
1746 tmp_param
.swizzle
= swizzle
| (swizzle
<< 4); /* Creates .xyxy */
1748 shader_arb_get_src_param(ins
, &tmp_param
, 1, src_name
[1]);
1750 shader_addline(buffer
, "X2D%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1751 dst_name
, src_name
[2], src_name
[0], src_name
[1]);
1755 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1756 /* 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]
1757 * src_name[0] can be TA, but TA is a private temp for modifiers, so it is save to overwrite
1759 shader_addline(buffer
, "MOV TA, %s;\n", src_name
[0]);
1760 shader_addline(buffer
, "MOV TA.z, 0.0;\n");
1761 shader_addline(buffer
, "DP3 TA, TA, %s;\n", src_name
[1]);
1762 shader_addline(buffer
, "ADD%s %s, TA, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
[2]);
1766 /* Map the opcode 1-to-1 to the GL code */
1767 static void shader_hw_map2gl(const struct wined3d_shader_instruction
*ins
)
1769 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1770 const char *instruction
;
1771 char arguments
[256], dst_str
[50];
1773 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1775 switch (ins
->handler_idx
)
1777 case WINED3DSIH_ABS
: instruction
= "ABS"; break;
1778 case WINED3DSIH_ADD
: instruction
= "ADD"; break;
1779 case WINED3DSIH_CRS
: instruction
= "XPD"; break;
1780 case WINED3DSIH_DP3
: instruction
= "DP3"; break;
1781 case WINED3DSIH_DP4
: instruction
= "DP4"; break;
1782 case WINED3DSIH_DST
: instruction
= "DST"; break;
1783 case WINED3DSIH_FRC
: instruction
= "FRC"; break;
1784 case WINED3DSIH_LIT
: instruction
= "LIT"; break;
1785 case WINED3DSIH_LRP
: instruction
= "LRP"; break;
1786 case WINED3DSIH_MAD
: instruction
= "MAD"; break;
1787 case WINED3DSIH_MAX
: instruction
= "MAX"; break;
1788 case WINED3DSIH_MIN
: instruction
= "MIN"; break;
1789 case WINED3DSIH_MOV
: instruction
= "MOV"; break;
1790 case WINED3DSIH_MUL
: instruction
= "MUL"; break;
1791 case WINED3DSIH_SGE
: instruction
= "SGE"; break;
1792 case WINED3DSIH_SLT
: instruction
= "SLT"; break;
1793 case WINED3DSIH_SUB
: instruction
= "SUB"; break;
1794 case WINED3DSIH_MOVA
:instruction
= "ARR"; break;
1795 case WINED3DSIH_DSX
: instruction
= "DDX"; break;
1796 default: instruction
= "";
1797 FIXME("Unhandled opcode %#x\n", ins
->handler_idx
);
1801 /* Note that shader_arb_add_dst_param() adds spaces. */
1802 arguments
[0] = '\0';
1803 shader_arb_get_dst_param(ins
, dst
, dst_str
);
1804 for (i
= 0; i
< ins
->src_count
; ++i
)
1807 strcat(arguments
, ", ");
1808 shader_arb_get_src_param(ins
, &ins
->src
[i
], i
, operand
);
1809 strcat(arguments
, operand
);
1811 shader_addline(buffer
, "%s%s %s%s;\n", instruction
, shader_arb_get_modifier(ins
), dst_str
, arguments
);
1814 static void shader_hw_nop(const struct wined3d_shader_instruction
*ins
) {}
1816 static DWORD
shader_arb_select_component(DWORD swizzle
, DWORD component
)
1818 return ((swizzle
>> 2 * component
) & 0x3) * 0x55;
1821 static void shader_hw_mov(const struct wined3d_shader_instruction
*ins
)
1823 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
1824 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
1825 BOOL pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
1826 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1827 const char *zero
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ZERO
);
1828 const char *one
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ONE
);
1829 const char *two
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_TWO
);
1831 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1832 char src0_param
[256];
1834 if (ins
->handler_idx
== WINED3DSIH_MOVA
)
1836 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1838 const char *offset
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_VS_REL_OFFSET
);
1840 if(ctx
->target_version
>= NV2
) {
1841 shader_hw_map2gl(ins
);
1844 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_param
);
1845 shader_arb_get_write_mask(ins
, &ins
->dst
[0], write_mask
);
1847 /* This implements the mova formula used in GLSL. The first two instructions
1848 * prepare the sign() part. Note that it is fine to have my_sign(0.0) = 1.0
1852 * A0.x = arl(floor(abs(0.0) + 0.5) * 1.0) = floor(0.5) = 0.0 since arl does a floor
1854 * The ARL is performed when A0 is used - the requested component is read from A0_SHADOW into
1855 * A0.x. We can use the overwritten component of A0_shadow as temporary storage for the sign.
1857 shader_addline(buffer
, "SGE A0_SHADOW%s, %s, %s;\n", write_mask
, src0_param
, zero
);
1858 shader_addline(buffer
, "MAD A0_SHADOW%s, A0_SHADOW, %s, -%s;\n", write_mask
, two
, one
);
1860 shader_addline(buffer
, "ABS TA%s, %s;\n", write_mask
, src0_param
);
1861 shader_addline(buffer
, "ADD TA%s, TA, rel_addr_const.x;\n", write_mask
);
1862 shader_addline(buffer
, "FLR TA%s, TA;\n", write_mask
);
1863 if (shader_data
->rel_offset
)
1865 shader_addline(buffer
, "ADD TA%s, TA, %s;\n", write_mask
, offset
);
1867 shader_addline(buffer
, "MUL A0_SHADOW%s, TA, A0_SHADOW;\n", write_mask
);
1869 ((struct shader_arb_ctx_priv
*)ins
->ctx
->backend_data
)->addr_reg
[0] = '\0';
1871 else if (reg_maps
->shader_version
.major
== 1
1872 && !shader_is_pshader_version(reg_maps
->shader_version
.type
)
1873 && ins
->dst
[0].reg
.type
== WINED3DSPR_ADDR
)
1875 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1876 src0_param
[0] = '\0';
1878 if (shader_data
->rel_offset
&& ctx
->target_version
== ARB
)
1880 const char *offset
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_VS_REL_OFFSET
);
1881 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_param
);
1882 shader_addline(buffer
, "ADD TA.x, %s, %s;\n", src0_param
, offset
);
1883 shader_addline(buffer
, "ARL A0.x, TA.x;\n");
1887 /* Apple's ARB_vertex_program implementation does not accept an ARL source argument
1888 * with more than one component. Thus replicate the first source argument over all
1889 * 4 components. For example, .xyzw -> .x (or better: .xxxx), .zwxy -> .z, etc) */
1890 struct wined3d_shader_src_param tmp_src
= ins
->src
[0];
1891 tmp_src
.swizzle
= shader_arb_select_component(tmp_src
.swizzle
, 0);
1892 shader_arb_get_src_param(ins
, &tmp_src
, 0, src0_param
);
1893 shader_addline(buffer
, "ARL A0.x, %s;\n", src0_param
);
1896 else if (ins
->dst
[0].reg
.type
== WINED3DSPR_COLOROUT
&& !ins
->dst
[0].reg
.idx
[0].offset
&& pshader
)
1898 if (ctx
->ps_post_process
&& shader
->u
.ps
.color0_mov
)
1900 shader_addline(buffer
, "#mov handled in srgb write or fog code\n");
1903 shader_hw_map2gl(ins
);
1907 shader_hw_map2gl(ins
);
1911 static void pshader_hw_texkill(const struct wined3d_shader_instruction
*ins
)
1913 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1914 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1917 /* No swizzles are allowed in d3d's texkill. PS 1.x ignores the 4th component as documented,
1918 * but >= 2.0 honors it (undocumented, but tested by the d3d9 testsuite)
1920 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1922 if (ins
->ctx
->reg_maps
->shader_version
.major
>= 2)
1924 const char *kilsrc
= "TA";
1927 shader_arb_get_register_name(ins
, &dst
->reg
, reg_dest
, &is_color
);
1928 if(dst
->write_mask
== WINED3DSP_WRITEMASK_ALL
)
1934 /* Sigh. KIL doesn't support swizzles/writemasks. KIL passes a writemask, but ".xy" for example
1935 * is not valid as a swizzle in ARB (needs ".xyyy"). Use SWZ to load the register properly, and set
1936 * masked out components to 0(won't kill)
1938 char x
= '0', y
= '0', z
= '0', w
= '0';
1939 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
) x
= 'x';
1940 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
) y
= 'y';
1941 if(dst
->write_mask
& WINED3DSP_WRITEMASK_2
) z
= 'z';
1942 if(dst
->write_mask
& WINED3DSP_WRITEMASK_3
) w
= 'w';
1943 shader_addline(buffer
, "SWZ TA, %s, %c, %c, %c, %c;\n", reg_dest
, x
, y
, z
, w
);
1945 shader_addline(buffer
, "KIL %s;\n", kilsrc
);
1949 /* ARB fp doesn't like swizzles on the parameter of the KIL instruction. To mask the 4th component,
1950 * copy the register into our general purpose TMP variable, overwrite .w and pass TMP to KIL
1952 * ps_1_3 shaders use the texcoord incarnation of the Tx register. ps_1_4 shaders can use the same,
1953 * or pass in any temporary register(in shader phase 2)
1955 if (ins
->ctx
->reg_maps
->shader_version
.minor
<= 3)
1956 sprintf(reg_dest
, "fragment.texcoord[%u]", dst
->reg
.idx
[0].offset
);
1958 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1959 shader_addline(buffer
, "SWZ TA, %s, x, y, z, 1;\n", reg_dest
);
1960 shader_addline(buffer
, "KIL TA;\n");
1964 static void pshader_hw_tex(const struct wined3d_shader_instruction
*ins
)
1966 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1967 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1968 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1969 ins
->ctx
->reg_maps
->shader_version
.minor
);
1970 struct wined3d_shader_src_param src
;
1974 DWORD reg_sampler_code
;
1976 BOOL swizzle_coord
= FALSE
;
1978 /* All versions have a destination register */
1979 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1981 /* 1.0-1.4: Use destination register number as texture code.
1982 2.0+: Use provided sampler number as texure code. */
1983 if (shader_version
< WINED3D_SHADER_VERSION(2,0))
1984 reg_sampler_code
= dst
->reg
.idx
[0].offset
;
1986 reg_sampler_code
= ins
->src
[1].reg
.idx
[0].offset
;
1988 /* 1.0-1.3: Use the texcoord varying.
1989 1.4+: Use provided coordinate source register. */
1990 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
1991 sprintf(reg_coord
, "fragment.texcoord[%u]", reg_sampler_code
);
1993 /* TEX is the only instruction that can handle DW and DZ natively */
1995 if(src
.modifiers
== WINED3DSPSM_DW
) src
.modifiers
= WINED3DSPSM_NONE
;
1996 if(src
.modifiers
== WINED3DSPSM_DZ
) src
.modifiers
= WINED3DSPSM_NONE
;
1997 shader_arb_get_src_param(ins
, &src
, 0, reg_coord
);
2001 * 1.1, 1.2, 1.3: Use WINED3D_TSS_TEXTURETRANSFORMFLAGS
2002 * 1.4: Use WINED3DSPSM_DZ or WINED3DSPSM_DW on src[0]
2003 * 2.0+: Use WINED3DSI_TEXLD_PROJECT on the opcode
2005 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
2008 if (reg_sampler_code
< MAX_TEXTURES
)
2009 flags
= priv
->cur_ps_args
->super
.tex_transform
>> reg_sampler_code
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
;
2010 if (flags
& WINED3D_PSARGS_PROJECTED
)
2012 myflags
|= TEX_PROJ
;
2013 if ((flags
& ~WINED3D_PSARGS_PROJECTED
) == WINED3D_TTFF_COUNT3
)
2014 swizzle_coord
= TRUE
;
2017 else if (shader_version
< WINED3D_SHADER_VERSION(2,0))
2019 enum wined3d_shader_src_modifier src_mod
= ins
->src
[0].modifiers
;
2020 if (src_mod
== WINED3DSPSM_DZ
)
2022 swizzle_coord
= TRUE
;
2023 myflags
|= TEX_PROJ
;
2024 } else if(src_mod
== WINED3DSPSM_DW
) {
2025 myflags
|= TEX_PROJ
;
2028 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) myflags
|= TEX_PROJ
;
2029 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) myflags
|= TEX_BIAS
;
2034 /* TXP cannot handle DZ natively, so move the z coordinate to .w.
2035 * reg_coord is a read-only varying register, so we need a temp reg */
2036 shader_addline(ins
->ctx
->buffer
, "SWZ TA, %s, x, y, z, z;\n", reg_coord
);
2037 strcpy(reg_coord
, "TA");
2040 shader_hw_sample(ins
, reg_sampler_code
, reg_dest
, reg_coord
, myflags
, NULL
, NULL
);
2043 static void pshader_hw_texcoord(const struct wined3d_shader_instruction
*ins
)
2045 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2046 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2047 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
2048 ins
->ctx
->reg_maps
->shader_version
.minor
);
2051 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
2053 DWORD reg
= dst
->reg
.idx
[0].offset
;
2055 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2056 shader_addline(buffer
, "MOV_SAT %s, fragment.texcoord[%u];\n", dst_str
, reg
);
2060 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_src
);
2061 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2062 shader_addline(buffer
, "MOV %s, %s;\n", dst_str
, reg_src
);
2066 static void pshader_hw_texreg2ar(const struct wined3d_shader_instruction
*ins
)
2068 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2071 DWORD reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2075 /* Note that texreg2ar treats Tx as a temporary register, not as a varying */
2076 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2077 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2078 /* Move .x first in case src_str is "TA" */
2079 shader_addline(buffer
, "MOV TA.y, %s.x;\n", src_str
);
2080 shader_addline(buffer
, "MOV TA.x, %s.w;\n", src_str
);
2081 if (reg1
< MAX_TEXTURES
)
2083 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2084 flags
= priv
->cur_ps_args
->super
.tex_transform
>> reg1
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
;
2086 shader_hw_sample(ins
, reg1
, dst_str
, "TA", flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2089 static void pshader_hw_texreg2gb(const struct wined3d_shader_instruction
*ins
)
2091 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2093 DWORD reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2097 /* Note that texreg2gb treats Tx as a temporary register, not as a varying */
2098 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2099 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2100 shader_addline(buffer
, "MOV TA.x, %s.y;\n", src_str
);
2101 shader_addline(buffer
, "MOV TA.y, %s.z;\n", src_str
);
2102 shader_hw_sample(ins
, reg1
, dst_str
, "TA", 0, NULL
, NULL
);
2105 static void pshader_hw_texreg2rgb(const struct wined3d_shader_instruction
*ins
)
2107 DWORD reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2111 /* Note that texreg2rg treats Tx as a temporary register, not as a varying */
2112 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2113 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2114 shader_hw_sample(ins
, reg1
, dst_str
, src_str
, 0, NULL
, NULL
);
2117 static void pshader_hw_texbem(const struct wined3d_shader_instruction
*ins
)
2119 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2120 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2121 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2122 char reg_coord
[40], dst_reg
[50], src_reg
[50];
2123 DWORD reg_dest_code
;
2125 /* All versions have a destination register. The Tx where the texture coordinates come
2126 * from is the varying incarnation of the texture register
2128 reg_dest_code
= dst
->reg
.idx
[0].offset
;
2129 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_reg
);
2130 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_reg
);
2131 sprintf(reg_coord
, "fragment.texcoord[%u]", reg_dest_code
);
2133 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed
2134 * The Tx in which the perturbation map is stored is the tempreg incarnation of the texture register
2136 * GL_NV_fragment_program_option could handle this in one instruction via X2D:
2137 * X2D TA.xy, fragment.texcoord, T%u, bumpenvmat%u.xzyw
2139 * However, the NV extensions are never enabled for <= 2.0 shaders because of the performance penalty that
2140 * comes with it, and texbem is an 1.x only instruction. No 1.x instruction forces us to enable the NV
2143 shader_addline(buffer
, "SWZ TB, bumpenvmat%d, x, z, 0, 0;\n", reg_dest_code
);
2144 shader_addline(buffer
, "DP3 TA.x, TB, %s;\n", src_reg
);
2145 shader_addline(buffer
, "SWZ TB, bumpenvmat%d, y, w, 0, 0;\n", reg_dest_code
);
2146 shader_addline(buffer
, "DP3 TA.y, TB, %s;\n", src_reg
);
2148 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2149 * so we can't let the GL handle this.
2151 if ((priv
->cur_ps_args
->super
.tex_transform
>> reg_dest_code
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
)
2152 & WINED3D_PSARGS_PROJECTED
)
2154 shader_addline(buffer
, "RCP TB.w, %s.w;\n", reg_coord
);
2155 shader_addline(buffer
, "MUL TB.xy, %s, TB.w;\n", reg_coord
);
2156 shader_addline(buffer
, "ADD TA.xy, TA, TB;\n");
2158 shader_addline(buffer
, "ADD TA.xy, TA, %s;\n", reg_coord
);
2161 shader_hw_sample(ins
, reg_dest_code
, dst_reg
, "TA", 0, NULL
, NULL
);
2163 if (ins
->handler_idx
== WINED3DSIH_TEXBEML
)
2165 /* No src swizzles are allowed, so this is ok */
2166 shader_addline(buffer
, "MAD TA, %s.z, luminance%d.x, luminance%d.y;\n",
2167 src_reg
, reg_dest_code
, reg_dest_code
);
2168 shader_addline(buffer
, "MUL %s, %s, TA;\n", dst_reg
, dst_reg
);
2172 static void pshader_hw_texm3x2pad(const struct wined3d_shader_instruction
*ins
)
2174 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2175 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2176 char src0_name
[50], dst_name
[50];
2178 struct wined3d_shader_register tmp_reg
= ins
->dst
[0].reg
;
2180 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2181 /* The next instruction will be a texm3x2tex or texm3x2depth that writes to the uninitialized
2182 * T<reg+1> register. Use this register to store the calculated vector
2184 tmp_reg
.idx
[0].offset
= reg
+ 1;
2185 shader_arb_get_register_name(ins
, &tmp_reg
, dst_name
, &is_color
);
2186 shader_addline(buffer
, "DP3 %s.x, fragment.texcoord[%u], %s;\n", dst_name
, reg
, src0_name
);
2189 static void pshader_hw_texm3x2tex(const struct wined3d_shader_instruction
*ins
)
2191 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2193 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2194 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2200 /* We know that we're writing to the uninitialized T<reg> register, so use it for temporary storage */
2201 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2203 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2204 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2205 shader_addline(buffer
, "DP3 %s.y, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2206 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2207 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2210 static void pshader_hw_texm3x3pad(const struct wined3d_shader_instruction
*ins
)
2212 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2213 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2214 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2215 char src0_name
[50], dst_name
[50];
2216 struct wined3d_shader_register tmp_reg
= ins
->dst
[0].reg
;
2219 /* There are always 2 texm3x3pad instructions followed by one texm3x3[tex,vspec, ...] instruction, with
2220 * incrementing ins->dst[0].register_idx numbers. So the pad instruction already knows the final destination
2221 * register, and this register is uninitialized(otherwise the assembler complains that it is 'redeclared')
2223 tmp_reg
.idx
[0].offset
= reg
+ 2 - tex_mx
->current_row
;
2224 shader_arb_get_register_name(ins
, &tmp_reg
, dst_name
, &is_color
);
2226 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2227 shader_addline(buffer
, "DP3 %s.%c, fragment.texcoord[%u], %s;\n",
2228 dst_name
, 'x' + tex_mx
->current_row
, reg
, src0_name
);
2229 tex_mx
->texcoord_w
[tex_mx
->current_row
++] = reg
;
2232 static void pshader_hw_texm3x3tex(const struct wined3d_shader_instruction
*ins
)
2234 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2235 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2237 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2238 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2240 char src0_name
[50], dst_name
[50];
2243 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2244 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2245 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_name
, reg
, src0_name
);
2247 /* Sample the texture using the calculated coordinates */
2248 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2249 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2250 shader_hw_sample(ins
, reg
, dst_str
, dst_name
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2251 tex_mx
->current_row
= 0;
2254 static void pshader_hw_texm3x3vspec(const struct wined3d_shader_instruction
*ins
)
2256 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2257 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2259 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2260 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2266 /* Get the dst reg without writemask strings. We know this register is uninitialized, so we can use all
2267 * components for temporary data storage
2269 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2270 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2271 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2273 /* Construct the eye-ray vector from w coordinates */
2274 shader_addline(buffer
, "MOV TB.x, fragment.texcoord[%u].w;\n", tex_mx
->texcoord_w
[0]);
2275 shader_addline(buffer
, "MOV TB.y, fragment.texcoord[%u].w;\n", tex_mx
->texcoord_w
[1]);
2276 shader_addline(buffer
, "MOV TB.z, fragment.texcoord[%u].w;\n", reg
);
2278 /* Calculate reflection vector
2280 shader_addline(buffer
, "DP3 %s.w, %s, TB;\n", dst_reg
, dst_reg
);
2281 /* The .w is ignored when sampling, so I can use TB.w to calculate dot(N, N) */
2282 shader_addline(buffer
, "DP3 TB.w, %s, %s;\n", dst_reg
, dst_reg
);
2283 shader_addline(buffer
, "RCP TB.w, TB.w;\n");
2284 shader_addline(buffer
, "MUL %s.w, %s.w, TB.w;\n", dst_reg
, dst_reg
);
2285 shader_addline(buffer
, "MUL %s, %s.w, %s;\n", dst_reg
, dst_reg
, dst_reg
);
2286 shader_addline(buffer
, "MAD %s, coefmul.x, %s, -TB;\n", dst_reg
, dst_reg
);
2288 /* Sample the texture using the calculated coordinates */
2289 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2290 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2291 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2292 tex_mx
->current_row
= 0;
2295 static void pshader_hw_texm3x3spec(const struct wined3d_shader_instruction
*ins
)
2297 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2298 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2300 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2301 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2308 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2309 shader_arb_get_src_param(ins
, &ins
->src
[0], 1, src1_name
);
2310 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2311 /* Note: dst_reg.xy is input here, generated by two texm3x3pad instructions */
2312 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2314 /* Calculate reflection vector.
2317 * dst_reg.xyz = 2 * --------- * N - E
2320 * Which normalizes the normal vector
2322 shader_addline(buffer
, "DP3 %s.w, %s, %s;\n", dst_reg
, dst_reg
, src1_name
);
2323 shader_addline(buffer
, "DP3 TC.w, %s, %s;\n", dst_reg
, dst_reg
);
2324 shader_addline(buffer
, "RCP TC.w, TC.w;\n");
2325 shader_addline(buffer
, "MUL %s.w, %s.w, TC.w;\n", dst_reg
, dst_reg
);
2326 shader_addline(buffer
, "MUL %s, %s.w, %s;\n", dst_reg
, dst_reg
, dst_reg
);
2327 shader_addline(buffer
, "MAD %s, coefmul.x, %s, -%s;\n", dst_reg
, dst_reg
, src1_name
);
2329 /* Sample the texture using the calculated coordinates */
2330 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2331 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2332 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2333 tex_mx
->current_row
= 0;
2336 static void pshader_hw_texdepth(const struct wined3d_shader_instruction
*ins
)
2338 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2339 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2341 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2342 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2344 /* texdepth has an implicit destination, the fragment depth value. It's only parameter,
2345 * which is essentially an input, is the destination register because it is the first
2346 * parameter. According to the msdn, this must be register r5, but let's keep it more flexible
2347 * here(writemasks/swizzles are not valid on texdepth)
2349 shader_arb_get_dst_param(ins
, dst
, dst_name
);
2351 /* According to the msdn, the source register(must be r5) is unusable after
2352 * the texdepth instruction, so we're free to modify it
2354 shader_addline(buffer
, "MIN %s.y, %s.y, %s;\n", dst_name
, dst_name
, one
);
2356 /* How to deal with the special case dst_name.g == 0? if r != 0, then
2357 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
2358 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
2360 shader_addline(buffer
, "RCP %s.y, %s.y;\n", dst_name
, dst_name
);
2361 shader_addline(buffer
, "MUL TA.x, %s.x, %s.y;\n", dst_name
, dst_name
);
2362 shader_addline(buffer
, "MIN TA.x, TA.x, %s;\n", one
);
2363 shader_addline(buffer
, "MAX result.depth, TA.x, %s;\n", zero
);
2366 /** Process the WINED3DSIO_TEXDP3TEX instruction in ARB:
2367 * Take a 3-component dot product of the TexCoord[dstreg] and src,
2368 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
2369 static void pshader_hw_texdp3tex(const struct wined3d_shader_instruction
*ins
)
2371 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2372 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
[0].offset
;
2376 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2377 shader_addline(buffer
, "MOV TB, 0.0;\n");
2378 shader_addline(buffer
, "DP3 TB.x, fragment.texcoord[%u], %s;\n", sampler_idx
, src0
);
2380 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2381 shader_hw_sample(ins
, sampler_idx
, dst_str
, "TB", 0 /* Only one coord, can't be projected */, NULL
, NULL
);
2384 /** Process the WINED3DSIO_TEXDP3 instruction in ARB:
2385 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
2386 static void pshader_hw_texdp3(const struct wined3d_shader_instruction
*ins
)
2388 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2391 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2393 /* Handle output register */
2394 shader_arb_get_dst_param(ins
, dst
, dst_str
);
2395 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2396 shader_addline(buffer
, "DP3 %s, fragment.texcoord[%u], %s;\n", dst_str
, dst
->reg
.idx
[0].offset
, src0
);
2399 /** Process the WINED3DSIO_TEXM3X3 instruction in ARB
2400 * Perform the 3rd row of a 3x3 matrix multiply */
2401 static void pshader_hw_texm3x3(const struct wined3d_shader_instruction
*ins
)
2403 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2404 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2405 char dst_str
[50], dst_name
[50];
2409 shader_arb_get_dst_param(ins
, dst
, dst_str
);
2410 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2411 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2412 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_name
, dst
->reg
.idx
[0].offset
, src0
);
2413 shader_addline(buffer
, "MOV %s, %s;\n", dst_str
, dst_name
);
2416 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in ARB:
2417 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
2418 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
2419 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
2421 static void pshader_hw_texm3x2depth(const struct wined3d_shader_instruction
*ins
)
2423 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2424 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2425 char src0
[50], dst_name
[50];
2427 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2428 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2430 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2431 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2432 shader_addline(buffer
, "DP3 %s.y, fragment.texcoord[%u], %s;\n", dst_name
, dst
->reg
.idx
[0].offset
, src0
);
2434 /* How to deal with the special case dst_name.g == 0? if r != 0, then
2435 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
2436 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
2438 shader_addline(buffer
, "RCP %s.y, %s.y;\n", dst_name
, dst_name
);
2439 shader_addline(buffer
, "MUL %s.x, %s.x, %s.y;\n", dst_name
, dst_name
, dst_name
);
2440 shader_addline(buffer
, "MIN %s.x, %s.x, %s;\n", dst_name
, dst_name
, one
);
2441 shader_addline(buffer
, "MAX result.depth, %s.x, %s;\n", dst_name
, zero
);
2444 /** Handles transforming all WINED3DSIO_M?x? opcodes for
2445 Vertex/Pixel shaders to ARB_vertex_program codes */
2446 static void shader_hw_mnxn(const struct wined3d_shader_instruction
*ins
)
2449 int nComponents
= 0;
2450 struct wined3d_shader_dst_param tmp_dst
= {{0}};
2451 struct wined3d_shader_src_param tmp_src
[2] = {{{0}}};
2452 struct wined3d_shader_instruction tmp_ins
;
2454 memset(&tmp_ins
, 0, sizeof(tmp_ins
));
2456 /* Set constants for the temporary argument */
2457 tmp_ins
.ctx
= ins
->ctx
;
2458 tmp_ins
.dst_count
= 1;
2459 tmp_ins
.dst
= &tmp_dst
;
2460 tmp_ins
.src_count
= 2;
2461 tmp_ins
.src
= tmp_src
;
2463 switch(ins
->handler_idx
)
2465 case WINED3DSIH_M4x4
:
2467 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2469 case WINED3DSIH_M4x3
:
2471 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2473 case WINED3DSIH_M3x4
:
2475 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2477 case WINED3DSIH_M3x3
:
2479 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2481 case WINED3DSIH_M3x2
:
2483 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2486 FIXME("Unhandled opcode %#x\n", ins
->handler_idx
);
2490 tmp_dst
= ins
->dst
[0];
2491 tmp_src
[0] = ins
->src
[0];
2492 tmp_src
[1] = ins
->src
[1];
2493 for (i
= 0; i
< nComponents
; ++i
)
2495 tmp_dst
.write_mask
= WINED3DSP_WRITEMASK_0
<< i
;
2496 shader_hw_map2gl(&tmp_ins
);
2497 ++tmp_src
[1].reg
.idx
[0].offset
;
2501 static DWORD
abs_modifier(DWORD mod
, BOOL
*need_abs
)
2507 case WINED3DSPSM_NONE
: return WINED3DSPSM_ABS
;
2508 case WINED3DSPSM_NEG
: return WINED3DSPSM_ABS
;
2509 case WINED3DSPSM_BIAS
: *need_abs
= TRUE
; return WINED3DSPSM_BIAS
;
2510 case WINED3DSPSM_BIASNEG
: *need_abs
= TRUE
; return WINED3DSPSM_BIASNEG
;
2511 case WINED3DSPSM_SIGN
: *need_abs
= TRUE
; return WINED3DSPSM_SIGN
;
2512 case WINED3DSPSM_SIGNNEG
: *need_abs
= TRUE
; return WINED3DSPSM_SIGNNEG
;
2513 case WINED3DSPSM_COMP
: *need_abs
= TRUE
; return WINED3DSPSM_COMP
;
2514 case WINED3DSPSM_X2
: *need_abs
= TRUE
; return WINED3DSPSM_X2
;
2515 case WINED3DSPSM_X2NEG
: *need_abs
= TRUE
; return WINED3DSPSM_X2NEG
;
2516 case WINED3DSPSM_DZ
: *need_abs
= TRUE
; return WINED3DSPSM_DZ
;
2517 case WINED3DSPSM_DW
: *need_abs
= TRUE
; return WINED3DSPSM_DW
;
2518 case WINED3DSPSM_ABS
: return WINED3DSPSM_ABS
;
2519 case WINED3DSPSM_ABSNEG
: return WINED3DSPSM_ABS
;
2521 FIXME("Unknown modifier %u\n", mod
);
2525 static void shader_hw_scalar_op(const struct wined3d_shader_instruction
*ins
)
2527 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2528 const char *instruction
;
2529 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
2530 BOOL need_abs
= FALSE
;
2535 switch(ins
->handler_idx
)
2537 case WINED3DSIH_RSQ
: instruction
= "RSQ"; break;
2538 case WINED3DSIH_RCP
: instruction
= "RCP"; break;
2539 case WINED3DSIH_EXPP
:
2540 if (ins
->ctx
->reg_maps
->shader_version
.major
< 2)
2542 instruction
= "EXP";
2546 case WINED3DSIH_EXP
:
2547 instruction
= "EX2";
2549 case WINED3DSIH_LOG
:
2550 case WINED3DSIH_LOGP
:
2551 /* The precision requirements suggest that LOGP matches ARBvp's LOG
2552 * instruction, but notice that the output of those instructions is
2554 src0_copy
.modifiers
= abs_modifier(src0_copy
.modifiers
, &need_abs
);
2555 instruction
= "LG2";
2557 default: instruction
= "";
2558 FIXME("Unhandled opcode %#x\n", ins
->handler_idx
);
2562 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
2564 src0_copy
.swizzle
= shader_arb_select_component(src0_copy
.swizzle
, 3);
2566 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
); /* Destination */
2567 shader_arb_get_src_param(ins
, &src0_copy
, 0, src
);
2571 shader_addline(buffer
, "ABS TA.w, %s;\n", src
);
2572 shader_addline(buffer
, "%s%s %s, TA.w;\n", instruction
, shader_arb_get_modifier(ins
), dst
);
2576 shader_addline(buffer
, "%s%s %s, %s;\n", instruction
, shader_arb_get_modifier(ins
), dst
, src
);
2581 static void shader_hw_nrm(const struct wined3d_shader_instruction
*ins
)
2583 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2586 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2587 BOOL pshader
= shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2588 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2590 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2591 shader_arb_get_src_param(ins
, &ins
->src
[0], 1 /* Use TB */, src_name
);
2593 /* In D3D, NRM of a vector with length zero returns zero. Catch this situation, as
2594 * otherwise NRM or RSQ would return NaN */
2595 if(pshader
&& priv
->target_version
>= NV3
)
2597 /* GL_NV_fragment_program2's NRM needs protection against length zero vectors too
2599 * TODO: Find out if DP3+NRM+MOV is really faster than DP3+RSQ+MUL
2601 shader_addline(buffer
, "DP3C TA, %s, %s;\n", src_name
, src_name
);
2602 shader_addline(buffer
, "NRM%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
);
2603 shader_addline(buffer
, "MOV %s (EQ), %s;\n", dst_name
, zero
);
2605 else if(priv
->target_version
>= NV2
)
2607 shader_addline(buffer
, "DP3C TA.x, %s, %s;\n", src_name
, src_name
);
2608 shader_addline(buffer
, "RSQ TA.x (NE), TA.x;\n");
2609 shader_addline(buffer
, "MUL%s %s, %s, TA.x;\n", shader_arb_get_modifier(ins
), dst_name
,
2614 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2616 shader_addline(buffer
, "DP3 TA.x, %s, %s;\n", src_name
, src_name
);
2617 /* Pass any non-zero value to RSQ if the input vector has a length of zero. The
2618 * RSQ result doesn't matter, as long as multiplying it by 0 returns 0.
2620 shader_addline(buffer
, "SGE TA.y, -TA.x, %s;\n", zero
);
2621 shader_addline(buffer
, "MAD TA.x, %s, TA.y, TA.x;\n", one
);
2623 shader_addline(buffer
, "RSQ TA.x, TA.x;\n");
2624 /* dst.w = src[0].w * 1 / (src.x^2 + src.y^2 + src.z^2)^(1/2) according to msdn*/
2625 shader_addline(buffer
, "MUL%s %s, %s, TA.x;\n", shader_arb_get_modifier(ins
), dst_name
,
2630 static void shader_hw_lrp(const struct wined3d_shader_instruction
*ins
)
2632 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2634 char src_name
[3][50];
2636 /* ARB_fragment_program has a convenient LRP instruction */
2637 if(shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)) {
2638 shader_hw_map2gl(ins
);
2642 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2643 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
2644 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
2645 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
2647 shader_addline(buffer
, "SUB TA, %s, %s;\n", src_name
[1], src_name
[2]);
2648 shader_addline(buffer
, "MAD%s %s, %s, TA, %s;\n", shader_arb_get_modifier(ins
),
2649 dst_name
, src_name
[0], src_name
[2]);
2652 static void shader_hw_sincos(const struct wined3d_shader_instruction
*ins
)
2654 /* This instruction exists in ARB, but the d3d instruction takes two extra parameters which
2655 * must contain fixed constants. So we need a separate function to filter those constants and
2658 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2659 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2660 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2662 char src_name0
[50], src_name1
[50], src_name2
[50];
2665 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
2666 if(shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)) {
2667 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2668 /* No modifiers are supported on SCS */
2669 shader_addline(buffer
, "SCS %s, %s;\n", dst_name
, src_name0
);
2671 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
)
2673 shader_arb_get_register_name(ins
, &dst
->reg
, src_name0
, &is_color
);
2674 shader_addline(buffer
, "MOV_SAT %s, %s;\n", dst_name
, src_name0
);
2676 } else if(priv
->target_version
>= NV2
) {
2677 shader_arb_get_register_name(ins
, &dst
->reg
, dst_name
, &is_color
);
2679 /* Sincos writemask must be .x, .y or .xy */
2680 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
)
2681 shader_addline(buffer
, "COS%s %s.x, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name0
);
2682 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
)
2683 shader_addline(buffer
, "SIN%s %s.y, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name0
);
2685 /* Approximate sine and cosine with a taylor series, as per math textbook. The application passes 8
2686 * helper constants(D3DSINCOSCONST1 and D3DSINCOSCONST2) in src1 and src2.
2688 * sin(x) = x - x^3/3! + x^5/5! - x^7/7! + ...
2689 * cos(x) = 1 - x^2/2! + x^4/4! - x^6/6! + ...
2691 * The constants we get are:
2693 * +1 +1, -1 -1 +1 +1 -1 -1
2694 * ---- , ---- , ---- , ----- , ----- , ----- , ------
2695 * 1!*2 2!*4 3!*8 4!*16 5!*32 6!*64 7!*128
2697 * If used with x^2, x^3, x^4 etc they calculate sin(x/2) and cos(x/2):
2701 * (x/2)^4 = x^4 / 16
2702 * (x/2)^5 = x^5 / 32
2705 * To get the final result:
2706 * sin(x) = 2 * sin(x/2) * cos(x/2)
2707 * cos(x) = cos(x/2)^2 - sin(x/2)^2
2708 * (from sin(x+y) and cos(x+y) rules)
2710 * As per MSDN, dst.z is undefined after the operation, and so is
2711 * dst.x and dst.y if they're masked out by the writemask. Ie
2712 * sincos dst.y, src1, c0, c1
2713 * returns the sine in dst.y. dst.x and dst.z are undefined, dst.w is not touched. The assembler
2714 * vsa.exe also stops with an error if the dest register is the same register as the source
2715 * register. This means we can use dest.xyz as temporary storage. The assembler vsa.exe output also
2716 * indicates that sincos consumes 8 instruction slots in vs_2_0(and, strangely, in vs_3_0).
2718 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
2719 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name2
);
2720 shader_arb_get_register_name(ins
, &dst
->reg
, dst_name
, &is_color
);
2722 shader_addline(buffer
, "MUL %s.x, %s, %s;\n", dst_name
, src_name0
, src_name0
); /* x ^ 2 */
2723 shader_addline(buffer
, "MUL TA.y, %s.x, %s;\n", dst_name
, src_name0
); /* x ^ 3 */
2724 shader_addline(buffer
, "MUL %s.y, TA.y, %s;\n", dst_name
, src_name0
); /* x ^ 4 */
2725 shader_addline(buffer
, "MUL TA.z, %s.y, %s;\n", dst_name
, src_name0
); /* x ^ 5 */
2726 shader_addline(buffer
, "MUL %s.z, TA.z, %s;\n", dst_name
, src_name0
); /* x ^ 6 */
2727 shader_addline(buffer
, "MUL TA.w, %s.z, %s;\n", dst_name
, src_name0
); /* x ^ 7 */
2731 * Unfortunately we don't get the constants in a DP4-capable form. Is there a way to
2732 * properly merge that with MULs in the code above?
2733 * The swizzles .yz and xw however fit into the .yzxw swizzle added to ps_2_0. Maybe
2734 * we can merge the sine and cosine MAD rows to calculate them together.
2736 shader_addline(buffer
, "MUL TA.x, %s, %s.w;\n", src_name0
, src_name2
); /* x^1, +1/(1!*2) */
2737 shader_addline(buffer
, "MAD TA.x, TA.y, %s.x, TA.x;\n", src_name2
); /* -1/(3!*8) */
2738 shader_addline(buffer
, "MAD TA.x, TA.z, %s.w, TA.x;\n", src_name1
); /* +1/(5!*32) */
2739 shader_addline(buffer
, "MAD TA.x, TA.w, %s.x, TA.x;\n", src_name1
); /* -1/(7!*128) */
2742 shader_addline(buffer
, "MAD TA.y, %s.x, %s.y, %s.z;\n", dst_name
, src_name2
, src_name2
); /* -1/(2!*4), +1.0 */
2743 shader_addline(buffer
, "MAD TA.y, %s.y, %s.z, TA.y;\n", dst_name
, src_name1
); /* +1/(4!*16) */
2744 shader_addline(buffer
, "MAD TA.y, %s.z, %s.y, TA.y;\n", dst_name
, src_name1
); /* -1/(6!*64) */
2746 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
) {
2748 shader_addline(buffer
, "MUL TA.z, TA.y, TA.y;\n");
2749 shader_addline(buffer
, "MAD %s.x, -TA.x, TA.x, TA.z;\n", dst_name
);
2751 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
) {
2753 shader_addline(buffer
, "MUL %s.y, TA.x, TA.y;\n", dst_name
);
2754 shader_addline(buffer
, "ADD %s.y, %s.y, %s.y;\n", dst_name
, dst_name
, dst_name
);
2759 static void shader_hw_sgn(const struct wined3d_shader_instruction
*ins
)
2761 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2764 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
2766 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2767 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
);
2769 /* SGN is only valid in vertex shaders */
2770 if(ctx
->target_version
>= NV2
) {
2771 shader_addline(buffer
, "SSG%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
);
2775 /* If SRC > 0.0, -SRC < SRC = TRUE, otherwise false.
2776 * if SRC < 0.0, SRC < -SRC = TRUE. If neither is true, src = 0.0
2778 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
) {
2779 shader_addline(buffer
, "SLT %s, -%s, %s;\n", dst_name
, src_name
, src_name
);
2781 /* src contains TA? Write to the dest first. This won't overwrite our destination.
2782 * Then use TA, and calculate the final result
2784 * Not reading from TA? Store the first result in TA to avoid overwriting the
2785 * destination if src reg = dst reg
2787 if(strstr(src_name
, "TA"))
2789 shader_addline(buffer
, "SLT %s, %s, -%s;\n", dst_name
, src_name
, src_name
);
2790 shader_addline(buffer
, "SLT TA, -%s, %s;\n", src_name
, src_name
);
2791 shader_addline(buffer
, "ADD %s, %s, -TA;\n", dst_name
, dst_name
);
2795 shader_addline(buffer
, "SLT TA, -%s, %s;\n", src_name
, src_name
);
2796 shader_addline(buffer
, "SLT %s, %s, -%s;\n", dst_name
, src_name
, src_name
);
2797 shader_addline(buffer
, "ADD %s, TA, -%s;\n", dst_name
, dst_name
);
2802 static void shader_hw_dsy(const struct wined3d_shader_instruction
*ins
)
2804 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2810 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
);
2811 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src
);
2812 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2814 shader_addline(buffer
, "DDY %s, %s;\n", dst
, src
);
2815 shader_addline(buffer
, "MUL%s %s, %s, ycorrection.y;\n", shader_arb_get_modifier(ins
), dst
, dst_name
);
2818 static void shader_hw_pow(const struct wined3d_shader_instruction
*ins
)
2820 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2821 char src0
[50], src1
[50], dst
[50];
2822 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
2823 BOOL need_abs
= FALSE
;
2824 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2825 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2827 /* POW operates on the absolute value of the input */
2828 src0_copy
.modifiers
= abs_modifier(src0_copy
.modifiers
, &need_abs
);
2830 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
);
2831 shader_arb_get_src_param(ins
, &src0_copy
, 0, src0
);
2832 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src1
);
2835 shader_addline(buffer
, "ABS TA.x, %s;\n", src0
);
2837 shader_addline(buffer
, "MOV TA.x, %s;\n", src0
);
2839 if (priv
->target_version
>= NV2
)
2841 shader_addline(buffer
, "MOVC TA.y, %s;\n", src1
);
2842 shader_addline(buffer
, "POW%s %s, TA.x, TA.y;\n", shader_arb_get_modifier(ins
), dst
);
2843 shader_addline(buffer
, "MOV %s (EQ.y), %s;\n", dst
, one
);
2847 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2848 const char *flt_eps
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_EPS
);
2850 shader_addline(buffer
, "ABS TA.y, %s;\n", src1
);
2851 shader_addline(buffer
, "SGE TA.y, -TA.y, %s;\n", zero
);
2852 /* Possibly add flt_eps to avoid getting float special values */
2853 shader_addline(buffer
, "MAD TA.z, TA.y, %s, %s;\n", flt_eps
, src1
);
2854 shader_addline(buffer
, "POW%s TA.x, TA.x, TA.z;\n", shader_arb_get_modifier(ins
));
2855 shader_addline(buffer
, "MAD TA.x, -TA.x, TA.y, TA.x;\n");
2856 shader_addline(buffer
, "MAD %s, TA.y, %s, TA.x;\n", dst
, one
);
2860 static void shader_hw_loop(const struct wined3d_shader_instruction
*ins
)
2862 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2864 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2867 shader_arb_get_src_param(ins
, &ins
->src
[1], 0, src_name
);
2871 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2872 struct list
*e
= list_head(&priv
->control_frames
);
2873 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2875 if(priv
->loop_depth
> 1) shader_addline(buffer
, "PUSHA aL;\n");
2876 /* The constant loader makes sure to load -1 into iX.w */
2877 shader_addline(buffer
, "ARLC aL, %s.xywz;\n", src_name
);
2878 shader_addline(buffer
, "BRA loop_%u_end (LE.x);\n", control_frame
->no
.loop
);
2879 shader_addline(buffer
, "loop_%u_start:\n", control_frame
->no
.loop
);
2883 shader_addline(buffer
, "LOOP %s;\n", src_name
);
2887 static void shader_hw_rep(const struct wined3d_shader_instruction
*ins
)
2889 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2891 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2893 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
);
2895 /* The constant loader makes sure to load -1 into iX.w */
2898 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2899 struct list
*e
= list_head(&priv
->control_frames
);
2900 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2902 if(priv
->loop_depth
> 1) shader_addline(buffer
, "PUSHA aL;\n");
2904 shader_addline(buffer
, "ARLC aL, %s.xywz;\n", src_name
);
2905 shader_addline(buffer
, "BRA loop_%u_end (LE.x);\n", control_frame
->no
.loop
);
2906 shader_addline(buffer
, "loop_%u_start:\n", control_frame
->no
.loop
);
2910 shader_addline(buffer
, "REP %s;\n", src_name
);
2914 static void shader_hw_endloop(const struct wined3d_shader_instruction
*ins
)
2916 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2917 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2921 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2922 struct list
*e
= list_head(&priv
->control_frames
);
2923 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2925 shader_addline(buffer
, "ARAC aL.xy, aL;\n");
2926 shader_addline(buffer
, "BRA loop_%u_start (GT.x);\n", control_frame
->no
.loop
);
2927 shader_addline(buffer
, "loop_%u_end:\n", control_frame
->no
.loop
);
2929 if(priv
->loop_depth
> 1) shader_addline(buffer
, "POPA aL;\n");
2933 shader_addline(buffer
, "ENDLOOP;\n");
2937 static void shader_hw_endrep(const struct wined3d_shader_instruction
*ins
)
2939 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2940 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2944 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2945 struct list
*e
= list_head(&priv
->control_frames
);
2946 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2948 shader_addline(buffer
, "ARAC aL.xy, aL;\n");
2949 shader_addline(buffer
, "BRA loop_%u_start (GT.x);\n", control_frame
->no
.loop
);
2950 shader_addline(buffer
, "loop_%u_end:\n", control_frame
->no
.loop
);
2952 if(priv
->loop_depth
> 1) shader_addline(buffer
, "POPA aL;\n");
2956 shader_addline(buffer
, "ENDREP;\n");
2960 static const struct control_frame
*find_last_loop(const struct shader_arb_ctx_priv
*priv
)
2962 struct control_frame
*control_frame
;
2964 LIST_FOR_EACH_ENTRY(control_frame
, &priv
->control_frames
, struct control_frame
, entry
)
2966 if(control_frame
->type
== LOOP
|| control_frame
->type
== REP
) return control_frame
;
2968 ERR("Could not find loop for break\n");
2972 static void shader_hw_break(const struct wined3d_shader_instruction
*ins
)
2974 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2975 const struct control_frame
*control_frame
= find_last_loop(ins
->ctx
->backend_data
);
2976 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2980 shader_addline(buffer
, "BRA loop_%u_end;\n", control_frame
->no
.loop
);
2984 shader_addline(buffer
, "BRK;\n");
2988 static const char *get_compare(enum wined3d_shader_rel_op op
)
2992 case WINED3D_SHADER_REL_OP_GT
: return "GT";
2993 case WINED3D_SHADER_REL_OP_EQ
: return "EQ";
2994 case WINED3D_SHADER_REL_OP_GE
: return "GE";
2995 case WINED3D_SHADER_REL_OP_LT
: return "LT";
2996 case WINED3D_SHADER_REL_OP_NE
: return "NE";
2997 case WINED3D_SHADER_REL_OP_LE
: return "LE";
2999 FIXME("Unrecognized operator %#x.\n", op
);
3004 static enum wined3d_shader_rel_op
invert_compare(enum wined3d_shader_rel_op op
)
3008 case WINED3D_SHADER_REL_OP_GT
: return WINED3D_SHADER_REL_OP_LE
;
3009 case WINED3D_SHADER_REL_OP_EQ
: return WINED3D_SHADER_REL_OP_NE
;
3010 case WINED3D_SHADER_REL_OP_GE
: return WINED3D_SHADER_REL_OP_LT
;
3011 case WINED3D_SHADER_REL_OP_LT
: return WINED3D_SHADER_REL_OP_GE
;
3012 case WINED3D_SHADER_REL_OP_NE
: return WINED3D_SHADER_REL_OP_EQ
;
3013 case WINED3D_SHADER_REL_OP_LE
: return WINED3D_SHADER_REL_OP_GT
;
3015 FIXME("Unrecognized operator %#x.\n", op
);
3020 static void shader_hw_breakc(const struct wined3d_shader_instruction
*ins
)
3022 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3023 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3024 const struct control_frame
*control_frame
= find_last_loop(ins
->ctx
->backend_data
);
3027 const char *comp
= get_compare(ins
->flags
);
3029 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
3030 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
3034 /* SUBC CC, src0, src1" works only in pixel shaders, so use TA to throw
3035 * away the subtraction result
3037 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3038 shader_addline(buffer
, "BRA loop_%u_end (%s.x);\n", control_frame
->no
.loop
, comp
);
3042 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3043 shader_addline(buffer
, "BRK (%s.x);\n", comp
);
3047 static void shader_hw_ifc(const struct wined3d_shader_instruction
*ins
)
3049 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3050 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3051 struct list
*e
= list_head(&priv
->control_frames
);
3052 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3056 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3058 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
3059 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
3063 /* Invert the flag. We jump to the else label if the condition is NOT true */
3064 comp
= get_compare(invert_compare(ins
->flags
));
3065 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3066 shader_addline(buffer
, "BRA ifc_%u_else (%s.x);\n", control_frame
->no
.ifc
, comp
);
3070 comp
= get_compare(ins
->flags
);
3071 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3072 shader_addline(buffer
, "IF %s.x;\n", comp
);
3076 static void shader_hw_else(const struct wined3d_shader_instruction
*ins
)
3078 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3079 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3080 struct list
*e
= list_head(&priv
->control_frames
);
3081 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3082 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3086 shader_addline(buffer
, "BRA ifc_%u_endif;\n", control_frame
->no
.ifc
);
3087 shader_addline(buffer
, "ifc_%u_else:\n", control_frame
->no
.ifc
);
3088 control_frame
->had_else
= TRUE
;
3092 shader_addline(buffer
, "ELSE;\n");
3096 static void shader_hw_endif(const struct wined3d_shader_instruction
*ins
)
3098 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3099 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3100 struct list
*e
= list_head(&priv
->control_frames
);
3101 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3102 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3106 if(control_frame
->had_else
)
3108 shader_addline(buffer
, "ifc_%u_endif:\n", control_frame
->no
.ifc
);
3112 shader_addline(buffer
, "#No else branch. else is endif\n");
3113 shader_addline(buffer
, "ifc_%u_else:\n", control_frame
->no
.ifc
);
3118 shader_addline(buffer
, "ENDIF;\n");
3122 static void shader_hw_texldd(const struct wined3d_shader_instruction
*ins
)
3124 DWORD sampler_idx
= ins
->src
[1].reg
.idx
[0].offset
;
3126 char reg_src
[3][40];
3127 WORD flags
= TEX_DERIV
;
3129 shader_arb_get_dst_param(ins
, &ins
->dst
[0], reg_dest
);
3130 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_src
[0]);
3131 shader_arb_get_src_param(ins
, &ins
->src
[2], 1, reg_src
[1]);
3132 shader_arb_get_src_param(ins
, &ins
->src
[3], 2, reg_src
[2]);
3134 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) flags
|= TEX_PROJ
;
3135 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) flags
|= TEX_BIAS
;
3137 shader_hw_sample(ins
, sampler_idx
, reg_dest
, reg_src
[0], flags
, reg_src
[1], reg_src
[2]);
3140 static void shader_hw_texldl(const struct wined3d_shader_instruction
*ins
)
3142 DWORD sampler_idx
= ins
->src
[1].reg
.idx
[0].offset
;
3145 WORD flags
= TEX_LOD
;
3147 shader_arb_get_dst_param(ins
, &ins
->dst
[0], reg_dest
);
3148 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_coord
);
3150 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) flags
|= TEX_PROJ
;
3151 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) flags
|= TEX_BIAS
;
3153 shader_hw_sample(ins
, sampler_idx
, reg_dest
, reg_coord
, flags
, NULL
, NULL
);
3156 static void shader_hw_label(const struct wined3d_shader_instruction
*ins
)
3158 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3159 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3161 priv
->in_main_func
= FALSE
;
3162 /* Call instructions activate the NV extensions, not labels and rets. If there is an uncalled
3163 * subroutine, don't generate a label that will make GL complain
3165 if(priv
->target_version
== ARB
) return;
3167 shader_addline(buffer
, "l%u:\n", ins
->src
[0].reg
.idx
[0].offset
);
3170 static void vshader_add_footer(struct shader_arb_ctx_priv
*priv_ctx
,
3171 const struct arb_vshader_private
*shader_data
, const struct arb_vs_compile_args
*args
,
3172 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
,
3173 struct wined3d_string_buffer
*buffer
)
3177 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
3178 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
3179 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
3180 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
3182 if (args
->super
.fog_src
== VS_FOG_Z
)
3184 shader_addline(buffer
, "MOV result.fogcoord, TMP_OUT.z;\n");
3190 /* posFixup.x is always 1.0, so we can safely use it */
3191 shader_addline(buffer
, "ADD result.fogcoord, posFixup.x, -posFixup.x;\n");
3195 /* Clamp fogcoord */
3196 const char *zero
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ZERO
);
3197 const char *one
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ONE
);
3199 shader_addline(buffer
, "MIN TMP_FOGCOORD.x, TMP_FOGCOORD.x, %s;\n", one
);
3200 shader_addline(buffer
, "MAX result.fogcoord.x, TMP_FOGCOORD.x, %s;\n", zero
);
3204 /* Clipplanes are always stored without y inversion */
3205 if (use_nv_clip(gl_info
) && priv_ctx
->target_version
>= NV2
)
3207 if (args
->super
.clip_enabled
)
3209 for (i
= 0; i
< priv_ctx
->vs_clipplanes
; i
++)
3211 shader_addline(buffer
, "DP4 result.clip[%u].x, TMP_OUT, state.clip[%u].plane;\n", i
, i
);
3215 else if (args
->clip
.boolclip
.clip_texcoord
)
3217 unsigned int cur_clip
= 0;
3218 char component
[4] = {'x', 'y', 'z', 'w'};
3219 const char *zero
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_ZERO
);
3221 for (i
= 0; i
< gl_info
->limits
.clipplanes
; ++i
)
3223 if (args
->clip
.boolclip
.clipplane_mask
& (1 << i
))
3225 shader_addline(buffer
, "DP4 TA.%c, TMP_OUT, state.clip[%u].plane;\n",
3226 component
[cur_clip
++], i
);
3232 shader_addline(buffer
, "MOV TA, %s;\n", zero
);
3235 shader_addline(buffer
, "MOV TA.yzw, %s;\n", zero
);
3238 shader_addline(buffer
, "MOV TA.zw, %s;\n", zero
);
3241 shader_addline(buffer
, "MOV TA.w, %s;\n", zero
);
3244 shader_addline(buffer
, "MOV result.texcoord[%u], TA;\n",
3245 args
->clip
.boolclip
.clip_texcoord
- 1);
3248 /* Write the final position.
3250 * OpenGL coordinates specify the center of the pixel while d3d coords specify
3251 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
3252 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
3253 * contains 1.0 to allow a mad, but arb vs swizzles are too restricted for that.
3255 shader_addline(buffer
, "MUL TA, posFixup, TMP_OUT.w;\n");
3256 shader_addline(buffer
, "ADD TMP_OUT.x, TMP_OUT.x, TA.z;\n");
3257 shader_addline(buffer
, "MAD TMP_OUT.y, TMP_OUT.y, posFixup.y, TA.w;\n");
3259 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
3260 * and the glsl equivalent
3262 if (need_helper_const(shader_data
, reg_maps
, gl_info
))
3264 const char *two
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_TWO
);
3265 shader_addline(buffer
, "MAD TMP_OUT.z, TMP_OUT.z, %s, -TMP_OUT.w;\n", two
);
3269 shader_addline(buffer
, "ADD TMP_OUT.z, TMP_OUT.z, TMP_OUT.z;\n");
3270 shader_addline(buffer
, "ADD TMP_OUT.z, TMP_OUT.z, -TMP_OUT.w;\n");
3273 shader_addline(buffer
, "MOV result.position, TMP_OUT;\n");
3275 priv_ctx
->footer_written
= TRUE
;
3278 static void shader_hw_ret(const struct wined3d_shader_instruction
*ins
)
3280 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3281 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3282 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
3283 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3285 if(priv
->target_version
== ARB
) return;
3289 if (priv
->in_main_func
) vshader_add_footer(priv
, shader
->backend_data
,
3290 priv
->cur_vs_args
, ins
->ctx
->reg_maps
, ins
->ctx
->gl_info
, buffer
);
3293 shader_addline(buffer
, "RET;\n");
3296 static void shader_hw_call(const struct wined3d_shader_instruction
*ins
)
3298 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3299 shader_addline(buffer
, "CAL l%u;\n", ins
->src
[0].reg
.idx
[0].offset
);
3302 static BOOL
shader_arb_compile(const struct wined3d_gl_info
*gl_info
, GLenum target
, const char *src
)
3304 const char *ptr
, *line
;
3307 if (TRACE_ON(d3d_shader
))
3310 while ((line
= get_line(&ptr
))) TRACE_(d3d_shader
)(" %.*s", (int)(ptr
- line
), line
);
3313 GL_EXTCALL(glProgramStringARB(target
, GL_PROGRAM_FORMAT_ASCII_ARB
, strlen(src
), src
));
3314 checkGLcall("glProgramStringARB()");
3316 if (FIXME_ON(d3d_shader
))
3318 gl_info
->gl_ops
.gl
.p_glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
3321 FIXME_(d3d_shader
)("Program error at position %d: %s\n\n", pos
,
3322 debugstr_a((const char *)gl_info
->gl_ops
.gl
.p_glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
3324 while ((line
= get_line(&ptr
))) FIXME_(d3d_shader
)(" %.*s", (int)(ptr
- line
), line
);
3325 FIXME_(d3d_shader
)("\n");
3331 if (WARN_ON(d3d_perf
))
3333 GL_EXTCALL(glGetProgramivARB(target
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
3334 checkGLcall("glGetProgramivARB()");
3336 WARN_(d3d_perf
)("Program exceeds native resource limits.\n");
3342 /* Context activation is done by the caller. */
3343 static GLuint
create_arb_blt_vertex_program(const struct wined3d_gl_info
*gl_info
)
3345 GLuint program_id
= 0;
3347 static const char blt_vprogram
[] =
3349 "PARAM c[1] = { { 1, 0.5 } };\n"
3350 "MOV result.position, vertex.position;\n"
3351 "MOV result.color, c[0].x;\n"
3352 "MOV result.texcoord[0], vertex.texcoord[0];\n"
3355 GL_EXTCALL(glGenProgramsARB(1, &program_id
));
3356 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, program_id
));
3357 shader_arb_compile(gl_info
, GL_VERTEX_PROGRAM_ARB
, blt_vprogram
);
3362 /* Context activation is done by the caller. */
3363 static GLuint
create_arb_blt_fragment_program(const struct wined3d_gl_info
*gl_info
,
3364 enum wined3d_gl_resource_type tex_type
, BOOL masked
)
3366 GLuint program_id
= 0;
3367 const char *fprogram
;
3369 static const char * const blt_fprograms_full
[WINED3D_GL_RES_TYPE_COUNT
] =
3371 /* WINED3D_GL_RES_TYPE_TEX_1D */
3373 /* WINED3D_GL_RES_TYPE_TEX_2D */
3376 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
3377 "MOV result.depth.z, R0.x;\n"
3379 /* WINED3D_GL_RES_TYPE_TEX_3D */
3381 /* WINED3D_GL_RES_TYPE_TEX_CUBE */
3384 "TEX R0.x, fragment.texcoord[0], texture[0], CUBE;\n"
3385 "MOV result.depth.z, R0.x;\n"
3387 /* WINED3D_GL_RES_TYPE_TEX_RECT */
3390 "TEX R0.x, fragment.texcoord[0], texture[0], RECT;\n"
3391 "MOV result.depth.z, R0.x;\n"
3395 static const char * const blt_fprograms_masked
[WINED3D_GL_RES_TYPE_COUNT
] =
3397 /* WINED3D_GL_RES_TYPE_TEX_1D */
3399 /* WINED3D_GL_RES_TYPE_TEX_2D */
3401 "PARAM mask = program.local[0];\n"
3403 "SLT R0.xy, fragment.position, mask.zwzw;\n"
3404 "MUL R0.x, R0.x, R0.y;\n"
3406 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
3407 "MOV result.depth.z, R0.x;\n"
3409 /* WINED3D_GL_RES_TYPE_TEX_3D */
3411 /* WINED3D_GL_RES_TYPE_TEX_CUBE */
3413 "PARAM mask = program.local[0];\n"
3415 "SLT R0.xy, fragment.position, mask.zwzw;\n"
3416 "MUL R0.x, R0.x, R0.y;\n"
3418 "TEX R0.x, fragment.texcoord[0], texture[0], CUBE;\n"
3419 "MOV result.depth.z, R0.x;\n"
3421 /* WINED3D_GL_RES_TYPE_TEX_RECT */
3423 "PARAM mask = program.local[0];\n"
3425 "SLT R0.xy, fragment.position, mask.zwzw;\n"
3426 "MUL R0.x, R0.x, R0.y;\n"
3428 "TEX R0.x, fragment.texcoord[0], texture[0], RECT;\n"
3429 "MOV result.depth.z, R0.x;\n"
3433 fprogram
= masked
? blt_fprograms_masked
[tex_type
] : blt_fprograms_full
[tex_type
];
3436 FIXME("tex_type %#x not supported, falling back to 2D\n", tex_type
);
3437 tex_type
= WINED3D_GL_RES_TYPE_TEX_2D
;
3438 fprogram
= masked
? blt_fprograms_masked
[tex_type
] : blt_fprograms_full
[tex_type
];
3441 GL_EXTCALL(glGenProgramsARB(1, &program_id
));
3442 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, program_id
));
3443 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, fprogram
);
3448 static void arbfp_add_sRGB_correction(struct wined3d_string_buffer
*buffer
, const char *fragcolor
,
3449 const char *tmp1
, const char *tmp2
, const char *tmp3
, const char *tmp4
, BOOL condcode
)
3451 /* Perform sRGB write correction. See GLX_EXT_framebuffer_sRGB */
3455 /* Sigh. MOVC CC doesn't work, so use one of the temps as dummy dest */
3456 shader_addline(buffer
, "SUBC %s, %s.x, srgb_consts1.x;\n", tmp1
, fragcolor
);
3457 /* Calculate the > 0.0031308 case */
3458 shader_addline(buffer
, "POW %s.x (GE), %s.x, srgb_consts0.x;\n", fragcolor
, fragcolor
);
3459 shader_addline(buffer
, "POW %s.y (GE), %s.y, srgb_consts0.x;\n", fragcolor
, fragcolor
);
3460 shader_addline(buffer
, "POW %s.z (GE), %s.z, srgb_consts0.x;\n", fragcolor
, fragcolor
);
3461 shader_addline(buffer
, "MUL %s.xyz (GE), %s, srgb_consts0.y;\n", fragcolor
, fragcolor
);
3462 shader_addline(buffer
, "SUB %s.xyz (GE), %s, srgb_consts0.z;\n", fragcolor
, fragcolor
);
3463 /* Calculate the < case */
3464 shader_addline(buffer
, "MUL %s.xyz (LT), srgb_consts0.w, %s;\n", fragcolor
, fragcolor
);
3468 /* Calculate the > 0.0031308 case */
3469 shader_addline(buffer
, "POW %s.x, %s.x, srgb_consts0.x;\n", tmp1
, fragcolor
);
3470 shader_addline(buffer
, "POW %s.y, %s.y, srgb_consts0.x;\n", tmp1
, fragcolor
);
3471 shader_addline(buffer
, "POW %s.z, %s.z, srgb_consts0.x;\n", tmp1
, fragcolor
);
3472 shader_addline(buffer
, "MUL %s, %s, srgb_consts0.y;\n", tmp1
, tmp1
);
3473 shader_addline(buffer
, "SUB %s, %s, srgb_consts0.z;\n", tmp1
, tmp1
);
3474 /* Calculate the < case */
3475 shader_addline(buffer
, "MUL %s, srgb_consts0.w, %s;\n", tmp2
, fragcolor
);
3476 /* Get 1.0 / 0.0 masks for > 0.0031308 and < 0.0031308 */
3477 shader_addline(buffer
, "SLT %s, srgb_consts1.x, %s;\n", tmp3
, fragcolor
);
3478 shader_addline(buffer
, "SGE %s, srgb_consts1.x, %s;\n", tmp4
, fragcolor
);
3479 /* Store the components > 0.0031308 in the destination */
3480 shader_addline(buffer
, "MUL %s.xyz, %s, %s;\n", fragcolor
, tmp1
, tmp3
);
3481 /* Add the components that are < 0.0031308 */
3482 shader_addline(buffer
, "MAD %s.xyz, %s, %s, %s;\n", fragcolor
, tmp2
, tmp4
, fragcolor
);
3483 /* Move everything into result.color at once. Nvidia hardware cannot handle partial
3484 * result.color writes(.rgb first, then .a), or handle overwriting already written
3485 * components. The assembler uses a temporary register in this case, which is usually
3486 * not allocated from one of our registers that were used earlier.
3489 /* [0.0;1.0] clamping. Not needed, this is done implicitly */
3492 static const DWORD
*find_loop_control_values(const struct wined3d_shader
*shader
, DWORD idx
)
3494 const struct wined3d_shader_lconst
*constant
;
3496 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsI
, struct wined3d_shader_lconst
, entry
)
3498 if (constant
->idx
== idx
)
3500 return constant
->value
;
3506 static void init_ps_input(const struct wined3d_shader
*shader
,
3507 const struct arb_ps_compile_args
*args
, struct shader_arb_ctx_priv
*priv
)
3509 static const char * const texcoords
[8] =
3511 "fragment.texcoord[0]", "fragment.texcoord[1]", "fragment.texcoord[2]", "fragment.texcoord[3]",
3512 "fragment.texcoord[4]", "fragment.texcoord[5]", "fragment.texcoord[6]", "fragment.texcoord[7]"
3515 const struct wined3d_shader_signature_element
*input
;
3516 const char *semantic_name
;
3519 switch(args
->super
.vp_mode
)
3521 case pretransformed
:
3523 /* The pixelshader has to collect the varyings on its own. In any case properly load
3524 * color0 and color1. In the case of pretransformed vertices also load texcoords. Set
3525 * other attribs to 0.0.
3527 * For fixedfunction this behavior is correct, according to the tests. For pretransformed
3528 * we'd either need a replacement shader that can load other attribs like BINORMAL, or
3529 * load the texcoord attrib pointers to match the pixel shader signature
3531 for (i
= 0; i
< shader
->input_signature
.element_count
; ++i
)
3533 input
= &shader
->input_signature
.elements
[i
];
3534 if (!(semantic_name
= input
->semantic_name
))
3536 semantic_idx
= input
->semantic_idx
;
3538 if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_COLOR
))
3541 priv
->ps_input
[input
->register_idx
] = "fragment.color.primary";
3542 else if (semantic_idx
== 1)
3543 priv
->ps_input
[input
->register_idx
] = "fragment.color.secondary";
3545 priv
->ps_input
[input
->register_idx
] = "0.0";
3547 else if (args
->super
.vp_mode
== fixedfunction
)
3549 priv
->ps_input
[input
->register_idx
] = "0.0";
3551 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
3553 if (semantic_idx
< 8)
3554 priv
->ps_input
[input
->register_idx
] = texcoords
[semantic_idx
];
3556 priv
->ps_input
[input
->register_idx
] = "0.0";
3558 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_FOG
))
3561 priv
->ps_input
[input
->register_idx
] = "fragment.fogcoord";
3563 priv
->ps_input
[input
->register_idx
] = "0.0";
3567 priv
->ps_input
[input
->register_idx
] = "0.0";
3570 TRACE("v%u, semantic %s%u is %s\n", input
->register_idx
,
3571 semantic_name
, semantic_idx
, priv
->ps_input
[input
->register_idx
]);
3576 /* That one is easy. The vertex shaders provide v0-v7 in fragment.texcoord and v8 and v9 in
3579 for(i
= 0; i
< 8; i
++)
3581 priv
->ps_input
[i
] = texcoords
[i
];
3583 priv
->ps_input
[8] = "fragment.color.primary";
3584 priv
->ps_input
[9] = "fragment.color.secondary";
3589 static void arbfp_add_linear_fog(struct wined3d_string_buffer
*buffer
,
3590 const char *fragcolor
, const char *tmp
)
3592 shader_addline(buffer
, "SUB %s.x, state.fog.params.z, fragment.fogcoord.x;\n", tmp
);
3593 shader_addline(buffer
, "MUL_SAT %s.x, %s.x, state.fog.params.w;\n", tmp
, tmp
);
3594 shader_addline(buffer
, "LRP %s.rgb, %s.x, %s, state.fog.color;\n", fragcolor
, tmp
, fragcolor
);
3597 /* Context activation is done by the caller. */
3598 static GLuint
shader_arb_generate_pshader(const struct wined3d_shader
*shader
,
3599 const struct wined3d_gl_info
*gl_info
, struct wined3d_string_buffer
*buffer
,
3600 const struct arb_ps_compile_args
*args
, struct arb_ps_compiled_shader
*compiled
)
3602 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
3603 const DWORD
*function
= shader
->function
;
3606 DWORD next_local
= 0;
3607 struct shader_arb_ctx_priv priv_ctx
;
3608 BOOL dcl_td
= FALSE
;
3609 BOOL want_nv_prog
= FALSE
;
3610 struct arb_pshader_private
*shader_priv
= shader
->backend_data
;
3612 BOOL custom_linear_fog
= FALSE
;
3616 unsigned int i
, found
= 0;
3618 for (i
= 0, map
= reg_maps
->temporary
; map
; map
>>= 1, ++i
)
3621 || (shader
->u
.ps
.color0_mov
&& i
== shader
->u
.ps
.color0_reg
)
3622 || (reg_maps
->shader_version
.major
< 2 && !i
))
3625 sprintf(srgbtmp
[found
], "R%u", i
);
3627 if (found
== 4) break;
3632 sprintf(srgbtmp
[0], "TA");
3633 sprintf(srgbtmp
[1], "TB");
3634 sprintf(srgbtmp
[2], "TC");
3635 sprintf(srgbtmp
[3], "TD");
3639 sprintf(srgbtmp
[1], "TA");
3640 sprintf(srgbtmp
[2], "TB");
3641 sprintf(srgbtmp
[3], "TC");
3644 sprintf(srgbtmp
[2], "TA");
3645 sprintf(srgbtmp
[3], "TB");
3648 sprintf(srgbtmp
[3], "TA");
3654 /* Create the hw ARB shader */
3655 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
3656 priv_ctx
.cur_ps_args
= args
;
3657 priv_ctx
.compiled_fprog
= compiled
;
3658 priv_ctx
.cur_np2fixup_info
= &compiled
->np2fixup_info
;
3659 init_ps_input(shader
, args
, &priv_ctx
);
3660 list_init(&priv_ctx
.control_frames
);
3661 priv_ctx
.ps_post_process
= args
->super
.srgb_correction
;
3663 /* Avoid enabling NV_fragment_program* if we do not need it.
3665 * Enabling GL_NV_fragment_program_option causes the driver to occupy a temporary register,
3666 * and it slows down the shader execution noticeably(about 5%). Usually our instruction emulation
3667 * is faster than what we gain from using higher native instructions. There are some things though
3668 * that cannot be emulated. In that case enable the extensions.
3669 * If the extension is enabled, instruction handlers that support both ways will use it.
3671 * Testing shows no performance difference between OPTION NV_fragment_program2 and NV_fragment_program.
3672 * So enable the best we can get.
3674 if(reg_maps
->usesdsx
|| reg_maps
->usesdsy
|| reg_maps
->loop_depth
> 0 || reg_maps
->usestexldd
||
3675 reg_maps
->usestexldl
|| reg_maps
->usesfacing
|| reg_maps
->usesifc
|| reg_maps
->usescall
)
3677 want_nv_prog
= TRUE
;
3680 shader_addline(buffer
, "!!ARBfp1.0\n");
3681 if (want_nv_prog
&& gl_info
->supported
[NV_FRAGMENT_PROGRAM2
])
3683 shader_addline(buffer
, "OPTION NV_fragment_program2;\n");
3684 priv_ctx
.target_version
= NV3
;
3686 else if (want_nv_prog
&& gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
])
3688 shader_addline(buffer
, "OPTION NV_fragment_program;\n");
3689 priv_ctx
.target_version
= NV2
;
3693 /* This is an error - either we're advertising the wrong shader version, or aren't enforcing some
3696 ERR("The shader requires instructions that are not available in plain GL_ARB_fragment_program\n");
3699 priv_ctx
.target_version
= ARB
;
3702 if (reg_maps
->rt_mask
> 1)
3704 shader_addline(buffer
, "OPTION ARB_draw_buffers;\n");
3707 if (reg_maps
->shader_version
.major
< 3)
3709 switch (args
->super
.fog
)
3711 case WINED3D_FFP_PS_FOG_OFF
:
3713 case WINED3D_FFP_PS_FOG_LINEAR
:
3714 if (gl_info
->quirks
& WINED3D_QUIRK_BROKEN_ARB_FOG
)
3716 custom_linear_fog
= TRUE
;
3717 priv_ctx
.ps_post_process
= TRUE
;
3720 shader_addline(buffer
, "OPTION ARB_fog_linear;\n");
3722 case WINED3D_FFP_PS_FOG_EXP
:
3723 shader_addline(buffer
, "OPTION ARB_fog_exp;\n");
3725 case WINED3D_FFP_PS_FOG_EXP2
:
3726 shader_addline(buffer
, "OPTION ARB_fog_exp2;\n");
3731 /* For now always declare the temps. At least the Nvidia assembler optimizes completely
3732 * unused temps away(but occupies them for the whole shader if they're used once). Always
3733 * declaring them avoids tricky bookkeeping work
3735 shader_addline(buffer
, "TEMP TA;\n"); /* Used for modifiers */
3736 shader_addline(buffer
, "TEMP TB;\n"); /* Used for modifiers */
3737 shader_addline(buffer
, "TEMP TC;\n"); /* Used for modifiers */
3738 if(dcl_td
) shader_addline(buffer
, "TEMP TD;\n"); /* Used for sRGB writing */
3739 shader_addline(buffer
, "PARAM coefdiv = { 0.5, 0.25, 0.125, 0.0625 };\n");
3740 shader_addline(buffer
, "PARAM coefmul = { 2, 4, 8, 16 };\n");
3741 wined3d_ftoa(eps
, ftoa_tmp
);
3742 shader_addline(buffer
, "PARAM ps_helper_const = { 0.0, 1.0, %s, 0.0 };\n", ftoa_tmp
);
3744 if (reg_maps
->shader_version
.major
< 2)
3746 strcpy(fragcolor
, "R0");
3750 if (priv_ctx
.ps_post_process
)
3752 if (shader
->u
.ps
.color0_mov
)
3754 sprintf(fragcolor
, "R%u", shader
->u
.ps
.color0_reg
);
3758 shader_addline(buffer
, "TEMP TMP_COLOR;\n");
3759 strcpy(fragcolor
, "TMP_COLOR");
3762 strcpy(fragcolor
, "result.color");
3766 if (args
->super
.srgb_correction
)
3768 shader_addline(buffer
, "PARAM srgb_consts0 = ");
3769 shader_arb_append_imm_vec4(buffer
, wined3d_srgb_const0
);
3770 shader_addline(buffer
, ";\n");
3771 shader_addline(buffer
, "PARAM srgb_consts1 = ");
3772 shader_arb_append_imm_vec4(buffer
, wined3d_srgb_const1
);
3773 shader_addline(buffer
, ";\n");
3776 /* Base Declarations */
3777 shader_generate_arb_declarations(shader
, reg_maps
, buffer
, gl_info
, NULL
, &priv_ctx
);
3779 for (i
= 0, map
= reg_maps
->bumpmat
; map
; map
>>= 1, ++i
)
3781 unsigned char bump_const
;
3783 if (!(map
& 1)) continue;
3785 bump_const
= compiled
->numbumpenvmatconsts
;
3786 compiled
->bumpenvmatconst
[bump_const
].const_num
= WINED3D_CONST_NUM_UNUSED
;
3787 compiled
->bumpenvmatconst
[bump_const
].texunit
= i
;
3788 compiled
->luminanceconst
[bump_const
].const_num
= WINED3D_CONST_NUM_UNUSED
;
3789 compiled
->luminanceconst
[bump_const
].texunit
= i
;
3791 /* We can fit the constants into the constant limit for sure because texbem, texbeml, bem and beml are only supported
3792 * in 1.x shaders, and GL_ARB_fragment_program has a constant limit of 24 constants. So in the worst case we're loading
3793 * 8 shader constants, 8 bump matrices and 8 luminance parameters and are perfectly fine. (No NP2 fixup on bumpmapped
3794 * textures due to conditional NP2 restrictions)
3796 * Use local constants to load the bump env parameters, not program.env. This avoids collisions with d3d constants of
3797 * shaders in newer shader models. Since the bump env parameters have to share their space with NP2 fixup constants,
3798 * their location is shader dependent anyway and they cannot be loaded globally.
3800 compiled
->bumpenvmatconst
[bump_const
].const_num
= next_local
++;
3801 shader_addline(buffer
, "PARAM bumpenvmat%d = program.local[%d];\n",
3802 i
, compiled
->bumpenvmatconst
[bump_const
].const_num
);
3803 compiled
->numbumpenvmatconsts
= bump_const
+ 1;
3805 if (!(reg_maps
->luminanceparams
& (1 << i
))) continue;
3807 compiled
->luminanceconst
[bump_const
].const_num
= next_local
++;
3808 shader_addline(buffer
, "PARAM luminance%d = program.local[%d];\n",
3809 i
, compiled
->luminanceconst
[bump_const
].const_num
);
3812 for(i
= 0; i
< MAX_CONST_I
; i
++)
3814 compiled
->int_consts
[i
] = WINED3D_CONST_NUM_UNUSED
;
3815 if (reg_maps
->integer_constants
& (1 << i
) && priv_ctx
.target_version
>= NV2
)
3817 const DWORD
*control_values
= find_loop_control_values(shader
, i
);
3821 shader_addline(buffer
, "PARAM I%u = {%u, %u, %u, -1};\n", i
,
3822 control_values
[0], control_values
[1], control_values
[2]);
3826 compiled
->int_consts
[i
] = next_local
;
3827 compiled
->num_int_consts
++;
3828 shader_addline(buffer
, "PARAM I%u = program.local[%u];\n", i
, next_local
++);
3833 if(reg_maps
->vpos
|| reg_maps
->usesdsy
)
3835 compiled
->ycorrection
= next_local
;
3836 shader_addline(buffer
, "PARAM ycorrection = program.local[%u];\n", next_local
++);
3840 shader_addline(buffer
, "TEMP vpos;\n");
3841 /* ycorrection.x: Backbuffer height(onscreen) or 0(offscreen).
3842 * ycorrection.y: -1.0(onscreen), 1.0(offscreen)
3843 * ycorrection.z: 1.0
3844 * ycorrection.w: 0.0
3846 shader_addline(buffer
, "MAD vpos, fragment.position, ycorrection.zyww, ycorrection.wxww;\n");
3847 shader_addline(buffer
, "FLR vpos.xy, vpos;\n");
3852 compiled
->ycorrection
= WINED3D_CONST_NUM_UNUSED
;
3855 /* Load constants to fixup NP2 texcoords if there are still free constants left:
3856 * Constants (texture dimensions) for the NP2 fixup are loaded as local program parameters. This will consume
3857 * at most 8 (MAX_FRAGMENT_SAMPLERS / 2) parameters, which is highly unlikely, since the application had to
3858 * use 16 NP2 textures at the same time. In case that we run out of constants the fixup is simply not
3859 * applied / activated. This will probably result in wrong rendering of the texture, but will save us from
3860 * shader compilation errors and the subsequent errors when drawing with this shader. */
3861 if (priv_ctx
.cur_ps_args
->super
.np2_fixup
) {
3862 unsigned char cur_fixup_sampler
= 0;
3864 struct arb_ps_np2fixup_info
* const fixup
= priv_ctx
.cur_np2fixup_info
;
3865 const WORD map
= priv_ctx
.cur_ps_args
->super
.np2_fixup
;
3866 const UINT max_lconsts
= gl_info
->limits
.arb_ps_local_constants
;
3868 fixup
->offset
= next_local
;
3869 fixup
->super
.active
= 0;
3871 for (i
= 0; i
< MAX_FRAGMENT_SAMPLERS
; ++i
) {
3872 if (!(map
& (1 << i
))) continue;
3874 if (fixup
->offset
+ (cur_fixup_sampler
>> 1) < max_lconsts
) {
3875 fixup
->super
.active
|= (1 << i
);
3876 fixup
->super
.idx
[i
] = cur_fixup_sampler
++;
3878 FIXME("No free constant found to load NP2 fixup data into shader. "
3879 "Sampling from this texture will probably look wrong.\n");
3884 fixup
->super
.num_consts
= (cur_fixup_sampler
+ 1) >> 1;
3885 if (fixup
->super
.num_consts
) {
3886 shader_addline(buffer
, "PARAM np2fixup[%u] = { program.env[%u..%u] };\n",
3887 fixup
->super
.num_consts
, fixup
->offset
, fixup
->super
.num_consts
+ fixup
->offset
- 1);
3891 if (shader_priv
->clipplane_emulation
!= ~0U && args
->clip
)
3893 shader_addline(buffer
, "KIL fragment.texcoord[%u];\n", shader_priv
->clipplane_emulation
);
3896 /* Base Shader Body */
3897 shader_generate_main(shader
, buffer
, reg_maps
, function
, &priv_ctx
);
3899 if(args
->super
.srgb_correction
) {
3900 arbfp_add_sRGB_correction(buffer
, fragcolor
, srgbtmp
[0], srgbtmp
[1], srgbtmp
[2], srgbtmp
[3],
3901 priv_ctx
.target_version
>= NV2
);
3904 if (custom_linear_fog
)
3905 arbfp_add_linear_fog(buffer
, fragcolor
, "TA");
3907 if(strcmp(fragcolor
, "result.color")) {
3908 shader_addline(buffer
, "MOV result.color, %s;\n", fragcolor
);
3910 shader_addline(buffer
, "END\n");
3912 /* TODO: change to resource.glObjectHandle or something like that */
3913 GL_EXTCALL(glGenProgramsARB(1, &retval
));
3915 TRACE("Creating a hw pixel shader, prg=%d\n", retval
);
3916 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, retval
));
3918 TRACE("Created hw pixel shader, prg=%d\n", retval
);
3919 if (!shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
->buffer
))
3925 static int compare_sig(const struct wined3d_shader_signature
*sig1
, const struct wined3d_shader_signature
*sig2
)
3930 if (sig1
->element_count
!= sig2
->element_count
)
3931 return sig1
->element_count
< sig2
->element_count
? -1 : 1;
3933 for (i
= 0; i
< sig1
->element_count
; ++i
)
3935 const struct wined3d_shader_signature_element
*e1
, *e2
;
3937 e1
= &sig1
->elements
[i
];
3938 e2
= &sig2
->elements
[i
];
3940 if (!e1
->semantic_name
|| !e2
->semantic_name
)
3942 /* Compare pointers, not contents. One string is NULL (element
3943 * does not exist), the other one is not NULL. */
3944 if (e1
->semantic_name
!= e2
->semantic_name
)
3945 return e1
->semantic_name
< e2
->semantic_name
? -1 : 1;
3949 if ((ret
= strcmp(e1
->semantic_name
, e2
->semantic_name
)))
3951 if (e1
->semantic_idx
!= e2
->semantic_idx
)
3952 return e1
->semantic_idx
< e2
->semantic_idx
? -1 : 1;
3953 if (e1
->sysval_semantic
!= e2
->sysval_semantic
)
3954 return e1
->sysval_semantic
< e2
->sysval_semantic
? -1 : 1;
3955 if (e1
->component_type
!= e2
->component_type
)
3956 return e1
->component_type
< e2
->component_type
? -1 : 1;
3957 if (e1
->register_idx
!= e2
->register_idx
)
3958 return e1
->register_idx
< e2
->register_idx
? -1 : 1;
3959 if (e1
->mask
!= e2
->mask
)
3960 return e1
->mask
< e2
->mask
? -1 : 1;
3965 static void clone_sig(struct wined3d_shader_signature
*new, const struct wined3d_shader_signature
*sig
)
3970 new->element_count
= sig
->element_count
;
3971 new->elements
= HeapAlloc(GetProcessHeap(), 0, sizeof(*new->elements
) * new->element_count
);
3972 for (i
= 0; i
< sig
->element_count
; ++i
)
3974 new->elements
[i
] = sig
->elements
[i
];
3976 if (!new->elements
[i
].semantic_name
)
3979 /* Clone the semantic string */
3980 name
= HeapAlloc(GetProcessHeap(), 0, strlen(sig
->elements
[i
].semantic_name
) + 1);
3981 strcpy(name
, sig
->elements
[i
].semantic_name
);
3982 new->elements
[i
].semantic_name
= name
;
3986 static DWORD
find_input_signature(struct shader_arb_priv
*priv
, const struct wined3d_shader_signature
*sig
)
3988 struct wine_rb_entry
*entry
= wine_rb_get(&priv
->signature_tree
, sig
);
3989 struct ps_signature
*found_sig
;
3993 found_sig
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
3994 TRACE("Found existing signature %u\n", found_sig
->idx
);
3995 return found_sig
->idx
;
3997 found_sig
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*found_sig
));
3998 clone_sig(&found_sig
->sig
, sig
);
3999 found_sig
->idx
= priv
->ps_sig_number
++;
4000 TRACE("New signature stored and assigned number %u\n", found_sig
->idx
);
4001 if(wine_rb_put(&priv
->signature_tree
, sig
, &found_sig
->entry
) == -1)
4003 ERR("Failed to insert program entry.\n");
4005 return found_sig
->idx
;
4008 static void init_output_registers(const struct wined3d_shader
*shader
,
4009 const struct wined3d_shader_signature
*ps_input_sig
,
4010 struct shader_arb_ctx_priv
*priv_ctx
, struct arb_vs_compiled_shader
*compiled
)
4013 static const char * const texcoords
[8] =
4015 "result.texcoord[0]", "result.texcoord[1]", "result.texcoord[2]", "result.texcoord[3]",
4016 "result.texcoord[4]", "result.texcoord[5]", "result.texcoord[6]", "result.texcoord[7]"
4018 /* Write generic input varyings 0 to 7 to result.texcoord[], varying 8 to result.color.primary
4019 * and varying 9 to result.color.secondary
4021 static const char * const decl_idx_to_string
[MAX_REG_INPUT
] =
4023 "result.texcoord[0]", "result.texcoord[1]", "result.texcoord[2]", "result.texcoord[3]",
4024 "result.texcoord[4]", "result.texcoord[5]", "result.texcoord[6]", "result.texcoord[7]",
4025 "result.color.primary", "result.color.secondary"
4030 TRACE("Pixel shader uses builtin varyings\n");
4031 /* Map builtins to builtins */
4032 for(i
= 0; i
< 8; i
++)
4034 priv_ctx
->texcrd_output
[i
] = texcoords
[i
];
4036 priv_ctx
->color_output
[0] = "result.color.primary";
4037 priv_ctx
->color_output
[1] = "result.color.secondary";
4038 priv_ctx
->fog_output
= "TMP_FOGCOORD";
4040 /* Map declared regs to builtins. Use "TA" to /dev/null unread output */
4041 for (i
= 0; i
< shader
->output_signature
.element_count
; ++i
)
4043 const struct wined3d_shader_signature_element
*output
= &shader
->output_signature
.elements
[i
];
4045 if (!output
->semantic_name
)
4048 if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_POSITION
))
4050 TRACE("o%u is TMP_OUT\n", output
->register_idx
);
4051 if (!output
->semantic_idx
)
4052 priv_ctx
->vs_output
[output
->register_idx
] = "TMP_OUT";
4054 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4056 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_PSIZE
))
4058 TRACE("o%u is result.pointsize\n", output
->register_idx
);
4059 if (!output
->semantic_idx
)
4060 priv_ctx
->vs_output
[output
->register_idx
] = "result.pointsize";
4062 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4064 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_COLOR
))
4066 TRACE("o%u is result.color.?, idx %u\n", output
->register_idx
, output
->semantic_idx
);
4067 if (!output
->semantic_idx
)
4068 priv_ctx
->vs_output
[output
->register_idx
] = "result.color.primary";
4069 else if (output
->semantic_idx
== 1)
4070 priv_ctx
->vs_output
[output
->register_idx
] = "result.color.secondary";
4071 else priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4073 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
4075 TRACE("o%u is result.texcoord[%u]\n", output
->register_idx
, output
->semantic_idx
);
4076 if (output
->semantic_idx
>= 8)
4077 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4079 priv_ctx
->vs_output
[output
->register_idx
] = texcoords
[output
->semantic_idx
];
4081 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_FOG
))
4083 TRACE("o%u is result.fogcoord\n", output
->register_idx
);
4084 if (output
->semantic_idx
> 0)
4085 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4087 priv_ctx
->vs_output
[output
->register_idx
] = "result.fogcoord";
4091 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4097 TRACE("Pixel shader uses declared varyings\n");
4099 /* Map builtin to declared. /dev/null the results by default to the TA temp reg */
4100 for(i
= 0; i
< 8; i
++)
4102 priv_ctx
->texcrd_output
[i
] = "TA";
4104 priv_ctx
->color_output
[0] = "TA";
4105 priv_ctx
->color_output
[1] = "TA";
4106 priv_ctx
->fog_output
= "TA";
4108 for (i
= 0; i
< ps_input_sig
->element_count
; ++i
)
4110 const struct wined3d_shader_signature_element
*input
= &ps_input_sig
->elements
[i
];
4112 if (!input
->semantic_name
)
4115 /* If a declared input register is not written by builtin arguments, don't write to it.
4116 * GL_NV_vertex_program makes sure the input defaults to 0.0, which is correct with D3D
4118 * Don't care about POSITION and PSIZE here - this is a builtin vertex shader, position goes
4119 * to TMP_OUT in any case
4121 if (shader_match_semantic(input
->semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
4123 if (input
->semantic_idx
< 8)
4124 priv_ctx
->texcrd_output
[input
->semantic_idx
] = decl_idx_to_string
[input
->register_idx
];
4126 else if (shader_match_semantic(input
->semantic_name
, WINED3D_DECL_USAGE_COLOR
))
4128 if (input
->semantic_idx
< 2)
4129 priv_ctx
->color_output
[input
->semantic_idx
] = decl_idx_to_string
[input
->register_idx
];
4131 else if (shader_match_semantic(input
->semantic_name
, WINED3D_DECL_USAGE_FOG
))
4133 if (!input
->semantic_idx
)
4134 priv_ctx
->fog_output
= decl_idx_to_string
[input
->register_idx
];
4141 if (!strcmp(decl_idx_to_string
[input
->register_idx
], "result.color.primary")
4142 || !strcmp(decl_idx_to_string
[input
->register_idx
], "result.color.secondary"))
4144 compiled
->need_color_unclamp
= TRUE
;
4148 /* Map declared to declared */
4149 for (i
= 0; i
< shader
->output_signature
.element_count
; ++i
)
4151 const struct wined3d_shader_signature_element
*output
= &shader
->output_signature
.elements
[i
];
4153 /* Write unread output to TA to throw them away */
4154 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4156 if (!output
->semantic_name
)
4159 if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_POSITION
) && !output
->semantic_idx
)
4161 priv_ctx
->vs_output
[output
->register_idx
] = "TMP_OUT";
4164 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_PSIZE
) && !output
->semantic_idx
)
4166 priv_ctx
->vs_output
[output
->register_idx
] = "result.pointsize";
4170 for (j
= 0; j
< ps_input_sig
->element_count
; ++j
)
4172 const struct wined3d_shader_signature_element
*input
= &ps_input_sig
->elements
[j
];
4174 if (!input
->semantic_name
)
4177 if (!strcmp(input
->semantic_name
, output
->semantic_name
)
4178 && input
->semantic_idx
== output
->semantic_idx
)
4180 priv_ctx
->vs_output
[output
->register_idx
] = decl_idx_to_string
[input
->register_idx
];
4182 if (!strcmp(priv_ctx
->vs_output
[output
->register_idx
], "result.color.primary")
4183 || !strcmp(priv_ctx
->vs_output
[output
->register_idx
], "result.color.secondary"))
4185 compiled
->need_color_unclamp
= TRUE
;
4192 /* Context activation is done by the caller. */
4193 static GLuint
shader_arb_generate_vshader(const struct wined3d_shader
*shader
,
4194 const struct wined3d_gl_info
*gl_info
, struct wined3d_string_buffer
*buffer
,
4195 const struct arb_vs_compile_args
*args
, struct arb_vs_compiled_shader
*compiled
,
4196 const struct wined3d_shader_signature
*ps_input_sig
)
4198 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
4199 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
4200 struct shader_arb_priv
*priv
= shader
->device
->shader_priv
;
4201 const DWORD
*function
= shader
->function
;
4203 DWORD next_local
= 0;
4204 struct shader_arb_ctx_priv priv_ctx
;
4207 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
4208 priv_ctx
.cur_vs_args
= args
;
4209 list_init(&priv_ctx
.control_frames
);
4210 init_output_registers(shader
, ps_input_sig
, &priv_ctx
, compiled
);
4212 /* Create the hw ARB shader */
4213 shader_addline(buffer
, "!!ARBvp1.0\n");
4215 /* Always enable the NV extension if available. Unlike fragment shaders, there is no
4216 * mesurable performance penalty, and we can always make use of it for clipplanes.
4218 if (gl_info
->supported
[NV_VERTEX_PROGRAM3
])
4220 shader_addline(buffer
, "OPTION NV_vertex_program3;\n");
4221 priv_ctx
.target_version
= NV3
;
4222 shader_addline(buffer
, "ADDRESS aL;\n");
4224 else if (gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
])
4226 shader_addline(buffer
, "OPTION NV_vertex_program2;\n");
4227 priv_ctx
.target_version
= NV2
;
4228 shader_addline(buffer
, "ADDRESS aL;\n");
4230 priv_ctx
.target_version
= ARB
;
4233 shader_addline(buffer
, "TEMP TMP_OUT;\n");
4235 shader_addline(buffer
, "TEMP TMP_FOGCOORD;\n");
4236 if (need_helper_const(shader_data
, reg_maps
, gl_info
))
4239 wined3d_ftoa(eps
, ftoa_tmp
);
4240 shader_addline(buffer
, "PARAM helper_const = { 0.0, 1.0, 2.0, %s};\n", ftoa_tmp
);
4242 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
))
4244 shader_addline(buffer
, "PARAM rel_addr_const = { 0.5, %d.0, 0.0, 0.0 };\n", shader_data
->rel_offset
);
4245 shader_addline(buffer
, "TEMP A0_SHADOW;\n");
4248 shader_addline(buffer
, "TEMP TA;\n");
4249 shader_addline(buffer
, "TEMP TB;\n");
4251 /* Base Declarations */
4252 shader_generate_arb_declarations(shader
, reg_maps
, buffer
, gl_info
,
4253 &priv_ctx
.vs_clipplanes
, &priv_ctx
);
4255 for(i
= 0; i
< MAX_CONST_I
; i
++)
4257 compiled
->int_consts
[i
] = WINED3D_CONST_NUM_UNUSED
;
4258 if(reg_maps
->integer_constants
& (1 << i
) && priv_ctx
.target_version
>= NV2
)
4260 const DWORD
*control_values
= find_loop_control_values(shader
, i
);
4264 shader_addline(buffer
, "PARAM I%u = {%u, %u, %u, -1};\n", i
,
4265 control_values
[0], control_values
[1], control_values
[2]);
4269 compiled
->int_consts
[i
] = next_local
;
4270 compiled
->num_int_consts
++;
4271 shader_addline(buffer
, "PARAM I%u = program.local[%u];\n", i
, next_local
++);
4276 /* We need a constant to fixup the final position */
4277 shader_addline(buffer
, "PARAM posFixup = program.local[%u];\n", next_local
);
4278 compiled
->pos_fixup
= next_local
++;
4280 /* Initialize output parameters. GL_ARB_vertex_program does not require special initialization values
4281 * for output parameters. D3D in theory does not do that either, but some applications depend on a
4282 * proper initialization of the secondary color, and programs using the fixed function pipeline without
4283 * a replacement shader depend on the texcoord.w being set properly.
4285 * GL_NV_vertex_program defines that all output values are initialized to {0.0, 0.0, 0.0, 1.0}. This
4286 * assertion is in effect even when using GL_ARB_vertex_program without any NV specific additions. So
4287 * skip this if NV_vertex_program is supported. Otherwise, initialize the secondary color. For the tex-
4288 * coords, we have a flag in the opengl caps. Many cards do not require the texcoord being set, and
4289 * this can eat a number of instructions, so skip it unless this cap is set as well
4291 if (!gl_info
->supported
[NV_VERTEX_PROGRAM
])
4293 const char *color_init
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_0001
);
4294 shader_addline(buffer
, "MOV result.color.secondary, %s;\n", color_init
);
4296 if (gl_info
->quirks
& WINED3D_QUIRK_SET_TEXCOORD_W
&& !priv
->ffp_proj_control
)
4299 const char *one
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_ONE
);
4300 for(i
= 0; i
< MAX_REG_TEXCRD
; i
++)
4302 if (reg_maps
->texcoord_mask
[i
] && reg_maps
->texcoord_mask
[i
] != WINED3DSP_WRITEMASK_ALL
)
4303 shader_addline(buffer
, "MOV result.texcoord[%u].w, %s\n", i
, one
);
4308 /* The shader starts with the main function */
4309 priv_ctx
.in_main_func
= TRUE
;
4310 /* Base Shader Body */
4311 shader_generate_main(shader
, buffer
, reg_maps
, function
, &priv_ctx
);
4313 if (!priv_ctx
.footer_written
) vshader_add_footer(&priv_ctx
,
4314 shader_data
, args
, reg_maps
, gl_info
, buffer
);
4316 shader_addline(buffer
, "END\n");
4318 /* TODO: change to resource.glObjectHandle or something like that */
4319 GL_EXTCALL(glGenProgramsARB(1, &ret
));
4321 TRACE("Creating a hw vertex shader, prg=%d\n", ret
);
4322 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, ret
));
4324 TRACE("Created hw vertex shader, prg=%d\n", ret
);
4325 if (!shader_arb_compile(gl_info
, GL_VERTEX_PROGRAM_ARB
, buffer
->buffer
))
4331 /* Context activation is done by the caller. */
4332 static struct arb_ps_compiled_shader
*find_arb_pshader(struct wined3d_shader
*shader
,
4333 const struct arb_ps_compile_args
*args
)
4335 struct wined3d_device
*device
= shader
->device
;
4336 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
4337 const struct wined3d_d3d_info
*d3d_info
= &device
->adapter
->d3d_info
;
4340 struct arb_ps_compiled_shader
*new_array
;
4341 struct wined3d_string_buffer buffer
;
4342 struct arb_pshader_private
*shader_data
;
4345 if (!shader
->backend_data
)
4347 struct shader_arb_priv
*priv
= device
->shader_priv
;
4349 shader
->backend_data
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
4350 shader_data
= shader
->backend_data
;
4351 shader_data
->clamp_consts
= shader
->reg_maps
.shader_version
.major
== 1;
4353 if (shader
->reg_maps
.shader_version
.major
< 3)
4354 shader_data
->input_signature_idx
= ~0U;
4356 shader_data
->input_signature_idx
= find_input_signature(priv
, &shader
->input_signature
);
4358 TRACE("Shader got assigned input signature index %u\n", shader_data
->input_signature_idx
);
4360 if (!d3d_info
->vs_clipping
)
4361 shader_data
->clipplane_emulation
= shader_find_free_input_register(&shader
->reg_maps
,
4362 d3d_info
->limits
.ffp_blend_stages
- 1);
4364 shader_data
->clipplane_emulation
= ~0U;
4366 shader_data
= shader
->backend_data
;
4368 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4369 * so a linear search is more performant than a hashmap or a binary search
4370 * (cache coherency etc)
4372 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4374 if (!memcmp(&shader_data
->gl_shaders
[i
].args
, args
, sizeof(*args
)))
4375 return &shader_data
->gl_shaders
[i
];
4378 TRACE("No matching GL shader found, compiling a new shader\n");
4379 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4380 if (shader_data
->num_gl_shaders
)
4382 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4383 new_array
= HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, shader_data
->gl_shaders
,
4384 new_size
* sizeof(*shader_data
->gl_shaders
));
4386 new_array
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
->gl_shaders
));
4391 ERR("Out of memory\n");
4394 shader_data
->gl_shaders
= new_array
;
4395 shader_data
->shader_array_size
= new_size
;
4398 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4400 pixelshader_update_resource_types(shader
, args
->super
.tex_types
);
4402 if (!string_buffer_init(&buffer
))
4404 ERR("Failed to initialize shader buffer.\n");
4408 ret
= shader_arb_generate_pshader(shader
, gl_info
, &buffer
, args
,
4409 &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
]);
4410 string_buffer_free(&buffer
);
4411 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].prgId
= ret
;
4413 return &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++];
4416 static inline BOOL
vs_args_equal(const struct arb_vs_compile_args
*stored
, const struct arb_vs_compile_args
*new,
4417 const DWORD use_map
, BOOL skip_int
) {
4418 if((stored
->super
.swizzle_map
& use_map
) != new->super
.swizzle_map
) return FALSE
;
4419 if(stored
->super
.clip_enabled
!= new->super
.clip_enabled
) return FALSE
;
4420 if(stored
->super
.fog_src
!= new->super
.fog_src
) return FALSE
;
4421 if(stored
->clip
.boolclip_compare
!= new->clip
.boolclip_compare
) return FALSE
;
4422 if(stored
->ps_signature
!= new->ps_signature
) return FALSE
;
4423 if(stored
->vertex
.samplers_compare
!= new->vertex
.samplers_compare
) return FALSE
;
4424 if(skip_int
) return TRUE
;
4426 return !memcmp(stored
->loop_ctrl
, new->loop_ctrl
, sizeof(stored
->loop_ctrl
));
4429 static struct arb_vs_compiled_shader
*find_arb_vshader(struct wined3d_shader
*shader
,
4430 const struct wined3d_gl_info
*gl_info
, DWORD use_map
, const struct arb_vs_compile_args
*args
,
4431 const struct wined3d_shader_signature
*ps_input_sig
)
4435 struct arb_vs_compiled_shader
*new_array
;
4436 struct wined3d_string_buffer buffer
;
4437 struct arb_vshader_private
*shader_data
;
4440 if (!shader
->backend_data
)
4442 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
4444 shader
->backend_data
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
4445 shader_data
= shader
->backend_data
;
4447 if ((gl_info
->quirks
& WINED3D_QUIRK_ARB_VS_OFFSET_LIMIT
)
4448 && reg_maps
->min_rel_offset
<= reg_maps
->max_rel_offset
)
4450 if (reg_maps
->max_rel_offset
- reg_maps
->min_rel_offset
> 127)
4452 FIXME("The difference between the minimum and maximum relative offset is > 127.\n");
4453 FIXME("Which this OpenGL implementation does not support. Try using GLSL.\n");
4454 FIXME("Min: %u, Max: %u.\n", reg_maps
->min_rel_offset
, reg_maps
->max_rel_offset
);
4456 else if (reg_maps
->max_rel_offset
- reg_maps
->min_rel_offset
> 63)
4457 shader_data
->rel_offset
= reg_maps
->min_rel_offset
+ 63;
4458 else if (reg_maps
->max_rel_offset
> 63)
4459 shader_data
->rel_offset
= reg_maps
->min_rel_offset
;
4462 shader_data
= shader
->backend_data
;
4464 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4465 * so a linear search is more performant than a hashmap or a binary search
4466 * (cache coherency etc)
4468 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
4469 if (vs_args_equal(&shader_data
->gl_shaders
[i
].args
, args
,
4470 use_map
, gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]))
4472 return &shader_data
->gl_shaders
[i
];
4476 TRACE("No matching GL shader found, compiling a new shader\n");
4478 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4479 if (shader_data
->num_gl_shaders
)
4481 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4482 new_array
= HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, shader_data
->gl_shaders
,
4483 new_size
* sizeof(*shader_data
->gl_shaders
));
4485 new_array
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
->gl_shaders
));
4490 ERR("Out of memory\n");
4493 shader_data
->gl_shaders
= new_array
;
4494 shader_data
->shader_array_size
= new_size
;
4497 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4499 if (!string_buffer_init(&buffer
))
4501 ERR("Failed to initialize shader buffer.\n");
4505 ret
= shader_arb_generate_vshader(shader
, gl_info
, &buffer
, args
,
4506 &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
],
4508 string_buffer_free(&buffer
);
4509 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].prgId
= ret
;
4511 return &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++];
4514 static void find_arb_ps_compile_args(const struct wined3d_state
*state
,
4515 const struct wined3d_context
*context
, const struct wined3d_shader
*shader
,
4516 struct arb_ps_compile_args
*args
)
4518 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4519 const struct wined3d_d3d_info
*d3d_info
= context
->d3d_info
;
4523 find_ps_compile_args(state
, shader
, context
->stream_info
.position_transformed
, &args
->super
, gl_info
);
4525 /* This forces all local boolean constants to 1 to make them stateblock independent */
4526 args
->bools
= shader
->reg_maps
.local_bool_consts
;
4528 for(i
= 0; i
< MAX_CONST_B
; i
++)
4530 if (state
->ps_consts_b
[i
])
4531 args
->bools
|= ( 1 << i
);
4534 /* Only enable the clip plane emulation KIL if at least one clipplane is enabled. The KIL instruction
4535 * is quite expensive because it forces the driver to disable early Z discards. It is cheaper to
4536 * duplicate the shader than have a no-op KIL instruction in every shader
4538 if (!d3d_info
->vs_clipping
&& use_vs(state
)
4539 && state
->render_states
[WINED3D_RS_CLIPPING
]
4540 && state
->render_states
[WINED3D_RS_CLIPPLANEENABLE
])
4545 /* Skip if unused or local, or supported natively */
4546 int_skip
= ~shader
->reg_maps
.integer_constants
| shader
->reg_maps
.local_int_consts
;
4547 if (int_skip
== 0xffff || gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
])
4549 memset(args
->loop_ctrl
, 0, sizeof(args
->loop_ctrl
));
4553 for(i
= 0; i
< MAX_CONST_I
; i
++)
4555 if(int_skip
& (1 << i
))
4557 args
->loop_ctrl
[i
][0] = 0;
4558 args
->loop_ctrl
[i
][1] = 0;
4559 args
->loop_ctrl
[i
][2] = 0;
4563 args
->loop_ctrl
[i
][0] = state
->ps_consts_i
[i
* 4];
4564 args
->loop_ctrl
[i
][1] = state
->ps_consts_i
[i
* 4 + 1];
4565 args
->loop_ctrl
[i
][2] = state
->ps_consts_i
[i
* 4 + 2];
4570 static void find_arb_vs_compile_args(const struct wined3d_state
*state
,
4571 const struct wined3d_context
*context
, const struct wined3d_shader
*shader
,
4572 struct arb_vs_compile_args
*args
)
4574 const struct wined3d_device
*device
= shader
->device
;
4575 const struct wined3d_adapter
*adapter
= device
->adapter
;
4576 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4577 const struct wined3d_d3d_info
*d3d_info
= context
->d3d_info
;
4581 find_vs_compile_args(state
, shader
, context
->stream_info
.swizzle_map
, &args
->super
);
4583 args
->clip
.boolclip_compare
= 0;
4586 const struct wined3d_shader
*ps
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
4587 const struct arb_pshader_private
*shader_priv
= ps
->backend_data
;
4588 args
->ps_signature
= shader_priv
->input_signature_idx
;
4590 args
->clip
.boolclip
.clip_texcoord
= shader_priv
->clipplane_emulation
+ 1;
4594 args
->ps_signature
= ~0;
4595 if (!d3d_info
->vs_clipping
&& adapter
->fragment_pipe
== &arbfp_fragment_pipeline
)
4596 args
->clip
.boolclip
.clip_texcoord
= ffp_clip_emul(context
) ? d3d_info
->limits
.ffp_blend_stages
: 0;
4597 /* Otherwise: Setting boolclip_compare set clip_texcoord to 0 */
4600 if (args
->clip
.boolclip
.clip_texcoord
)
4602 if (state
->render_states
[WINED3D_RS_CLIPPING
])
4603 args
->clip
.boolclip
.clipplane_mask
= (unsigned char)state
->render_states
[WINED3D_RS_CLIPPLANEENABLE
];
4604 /* clipplane_mask was set to 0 by setting boolclip_compare to 0 */
4607 /* This forces all local boolean constants to 1 to make them stateblock independent */
4608 args
->clip
.boolclip
.bools
= shader
->reg_maps
.local_bool_consts
;
4609 /* TODO: Figure out if it would be better to store bool constants as bitmasks in the stateblock */
4610 for(i
= 0; i
< MAX_CONST_B
; i
++)
4612 if (state
->vs_consts_b
[i
])
4613 args
->clip
.boolclip
.bools
|= ( 1 << i
);
4616 args
->vertex
.samplers
[0] = context
->tex_unit_map
[MAX_FRAGMENT_SAMPLERS
+ 0];
4617 args
->vertex
.samplers
[1] = context
->tex_unit_map
[MAX_FRAGMENT_SAMPLERS
+ 1];
4618 args
->vertex
.samplers
[2] = context
->tex_unit_map
[MAX_FRAGMENT_SAMPLERS
+ 2];
4619 args
->vertex
.samplers
[3] = 0;
4621 /* Skip if unused or local */
4622 int_skip
= ~shader
->reg_maps
.integer_constants
| shader
->reg_maps
.local_int_consts
;
4623 /* This is about flow control, not clipping. */
4624 if (int_skip
== 0xffff || gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
])
4626 memset(args
->loop_ctrl
, 0, sizeof(args
->loop_ctrl
));
4630 for(i
= 0; i
< MAX_CONST_I
; i
++)
4632 if(int_skip
& (1 << i
))
4634 args
->loop_ctrl
[i
][0] = 0;
4635 args
->loop_ctrl
[i
][1] = 0;
4636 args
->loop_ctrl
[i
][2] = 0;
4640 args
->loop_ctrl
[i
][0] = state
->vs_consts_i
[i
* 4];
4641 args
->loop_ctrl
[i
][1] = state
->vs_consts_i
[i
* 4 + 1];
4642 args
->loop_ctrl
[i
][2] = state
->vs_consts_i
[i
* 4 + 2];
4647 /* Context activation is done by the caller. */
4648 static void shader_arb_select(void *shader_priv
, struct wined3d_context
*context
,
4649 const struct wined3d_state
*state
)
4651 struct shader_arb_priv
*priv
= shader_priv
;
4652 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4655 /* Deal with pixel shaders first so the vertex shader arg function has the input signature ready */
4658 struct wined3d_shader
*ps
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
4659 struct arb_ps_compile_args compile_args
;
4660 struct arb_ps_compiled_shader
*compiled
;
4662 TRACE("Using pixel shader %p.\n", ps
);
4663 find_arb_ps_compile_args(state
, context
, ps
, &compile_args
);
4664 compiled
= find_arb_pshader(ps
, &compile_args
);
4665 priv
->current_fprogram_id
= compiled
->prgId
;
4666 priv
->compiled_fprog
= compiled
;
4668 /* Bind the fragment program */
4669 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, priv
->current_fprogram_id
));
4670 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->current_fprogram_id);");
4672 if (!priv
->use_arbfp_fixed_func
)
4673 priv
->fragment_pipe
->enable_extension(gl_info
, FALSE
);
4675 /* Enable OpenGL fragment programs. */
4676 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
4677 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
4679 TRACE("Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", priv
->current_fprogram_id
);
4681 /* Pixel Shader 1.x constants are clamped to [-1;1], Pixel Shader 2.0 constants are not. If switching between
4682 * a 1.x and newer shader, reload the first 8 constants
4684 if (priv
->last_ps_const_clamped
!= ((struct arb_pshader_private
*)ps
->backend_data
)->clamp_consts
)
4686 priv
->last_ps_const_clamped
= ((struct arb_pshader_private
*)ps
->backend_data
)->clamp_consts
;
4687 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, 8);
4688 for(i
= 0; i
< 8; i
++)
4690 priv
->pshader_const_dirty
[i
] = 1;
4692 /* Also takes care of loading local constants */
4693 shader_arb_load_constants_internal(shader_priv
, context
, state
, TRUE
, FALSE
, TRUE
);
4697 UINT rt_height
= state
->fb
->render_targets
[0]->height
;
4698 shader_arb_ps_local_constants(compiled
, context
, state
, rt_height
);
4701 /* Force constant reloading for the NP2 fixup (see comment in shader_glsl_select for more info) */
4702 if (compiled
->np2fixup_info
.super
.active
)
4703 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_NP2_FIXUP
;
4705 if (ps
->load_local_constsF
)
4706 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_F
;
4710 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
] && !priv
->use_arbfp_fixed_func
)
4712 /* Disable only if we're not using arbfp fixed function fragment
4713 * processing. If this is used, keep GL_FRAGMENT_PROGRAM_ARB
4714 * enabled, and the fixed function pipeline will bind the fixed
4715 * function replacement shader. */
4716 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4717 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4718 priv
->current_fprogram_id
= 0;
4720 priv
->fragment_pipe
->enable_extension(gl_info
, TRUE
);
4725 struct wined3d_shader
*vs
= state
->shader
[WINED3D_SHADER_TYPE_VERTEX
];
4726 struct arb_vs_compile_args compile_args
;
4727 struct arb_vs_compiled_shader
*compiled
;
4728 const struct wined3d_shader_signature
*ps_input_sig
;
4730 TRACE("Using vertex shader %p\n", vs
);
4731 find_arb_vs_compile_args(state
, context
, vs
, &compile_args
);
4733 /* Instead of searching for the signature in the signature list, read the one from the
4734 * current pixel shader. It's maybe not the shader where the signature came from, but it
4735 * is the same signature and faster to find. */
4736 if (compile_args
.ps_signature
== ~0U)
4737 ps_input_sig
= NULL
;
4739 ps_input_sig
= &state
->shader
[WINED3D_SHADER_TYPE_PIXEL
]->input_signature
;
4741 compiled
= find_arb_vshader(vs
, context
->gl_info
, context
->stream_info
.use_map
,
4742 &compile_args
, ps_input_sig
);
4743 priv
->current_vprogram_id
= compiled
->prgId
;
4744 priv
->compiled_vprog
= compiled
;
4746 /* Bind the vertex program */
4747 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->current_vprogram_id
));
4748 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->current_vprogram_id);");
4750 priv
->vertex_pipe
->vp_enable(gl_info
, FALSE
);
4752 /* Enable OpenGL vertex programs */
4753 gl_info
->gl_ops
.gl
.p_glEnable(GL_VERTEX_PROGRAM_ARB
);
4754 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
4755 TRACE("Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", priv
->current_vprogram_id
);
4756 shader_arb_vs_local_constants(compiled
, context
, state
);
4758 if(priv
->last_vs_color_unclamp
!= compiled
->need_color_unclamp
) {
4759 priv
->last_vs_color_unclamp
= compiled
->need_color_unclamp
;
4761 if (gl_info
->supported
[ARB_COLOR_BUFFER_FLOAT
])
4763 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, !compiled
->need_color_unclamp
));
4764 checkGLcall("glClampColorARB");
4766 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
4770 if (vs
->load_local_constsF
)
4771 context
->constant_update_mask
|= WINED3D_SHADER_CONST_VS_F
;
4775 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
4777 priv
->current_vprogram_id
= 0;
4778 gl_info
->gl_ops
.gl
.p_glDisable(GL_VERTEX_PROGRAM_ARB
);
4779 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4781 priv
->vertex_pipe
->vp_enable(gl_info
, TRUE
);
4786 /* Context activation is done by the caller. */
4787 static void shader_arb_disable(void *shader_priv
, struct wined3d_context
*context
)
4789 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4790 struct shader_arb_priv
*priv
= shader_priv
;
4792 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
4794 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4795 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4796 priv
->current_fprogram_id
= 0;
4798 priv
->fragment_pipe
->enable_extension(gl_info
, FALSE
);
4800 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
4802 priv
->current_vprogram_id
= 0;
4803 gl_info
->gl_ops
.gl
.p_glDisable(GL_VERTEX_PROGRAM_ARB
);
4804 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4806 priv
->vertex_pipe
->vp_enable(gl_info
, FALSE
);
4808 if (gl_info
->supported
[ARB_COLOR_BUFFER_FLOAT
] && priv
->last_vs_color_unclamp
)
4810 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, GL_FIXED_ONLY_ARB
));
4811 checkGLcall("glClampColorARB");
4812 priv
->last_vs_color_unclamp
= FALSE
;
4815 context
->shader_update_mask
= (1 << WINED3D_SHADER_TYPE_PIXEL
)
4816 | (1 << WINED3D_SHADER_TYPE_VERTEX
)
4817 | (1 << WINED3D_SHADER_TYPE_GEOMETRY
);
4820 /* Context activation is done by the caller. */
4821 static void shader_arb_select_depth_blt(void *shader_priv
, const struct wined3d_gl_info
*gl_info
,
4822 enum wined3d_gl_resource_type tex_type
, const SIZE
*ds_mask_size
)
4824 const float mask
[] = {0.0f
, 0.0f
, (float)ds_mask_size
->cx
, (float)ds_mask_size
->cy
};
4825 BOOL masked
= ds_mask_size
->cx
&& ds_mask_size
->cy
;
4826 struct shader_arb_priv
*priv
= shader_priv
;
4827 GLuint
*blt_fprogram
;
4829 if (!priv
->depth_blt_vprogram_id
) priv
->depth_blt_vprogram_id
= create_arb_blt_vertex_program(gl_info
);
4830 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->depth_blt_vprogram_id
));
4831 gl_info
->gl_ops
.gl
.p_glEnable(GL_VERTEX_PROGRAM_ARB
);
4833 blt_fprogram
= masked
? &priv
->depth_blt_fprogram_id_masked
[tex_type
] : &priv
->depth_blt_fprogram_id_full
[tex_type
];
4834 if (!*blt_fprogram
) *blt_fprogram
= create_arb_blt_fragment_program(gl_info
, tex_type
, masked
);
4835 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, *blt_fprogram
));
4836 if (masked
) GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, 0, mask
));
4837 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
4840 /* Context activation is done by the caller. */
4841 static void shader_arb_deselect_depth_blt(void *shader_priv
, const struct wined3d_gl_info
*gl_info
)
4843 struct shader_arb_priv
*priv
= shader_priv
;
4845 if (priv
->current_vprogram_id
) {
4846 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->current_vprogram_id
));
4847 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
4849 TRACE("Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB.\n", priv
->current_vprogram_id
);
4853 gl_info
->gl_ops
.gl
.p_glDisable(GL_VERTEX_PROGRAM_ARB
);
4854 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4857 if (priv
->current_fprogram_id
) {
4858 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, priv
->current_fprogram_id
));
4859 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
4861 TRACE("Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB.\n", priv
->current_fprogram_id
);
4863 else if(!priv
->use_arbfp_fixed_func
)
4865 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4866 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4870 static void shader_arb_destroy(struct wined3d_shader
*shader
)
4872 struct wined3d_device
*device
= shader
->device
;
4873 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
4875 if (shader_is_pshader_version(shader
->reg_maps
.shader_version
.type
))
4877 struct arb_pshader_private
*shader_data
= shader
->backend_data
;
4880 if(!shader_data
) return; /* This can happen if a shader was never compiled */
4882 if (shader_data
->num_gl_shaders
)
4884 struct wined3d_context
*context
= context_acquire(device
, NULL
);
4886 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4888 GL_EXTCALL(glDeleteProgramsARB(1, &shader_data
->gl_shaders
[i
].prgId
));
4889 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader_data->gl_shaders[i].prgId))");
4892 context_release(context
);
4895 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
);
4896 HeapFree(GetProcessHeap(), 0, shader_data
);
4897 shader
->backend_data
= NULL
;
4901 struct arb_vshader_private
*shader_data
= shader
->backend_data
;
4904 if(!shader_data
) return; /* This can happen if a shader was never compiled */
4906 if (shader_data
->num_gl_shaders
)
4908 struct wined3d_context
*context
= context_acquire(device
, NULL
);
4910 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4912 GL_EXTCALL(glDeleteProgramsARB(1, &shader_data
->gl_shaders
[i
].prgId
));
4913 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader_data->gl_shaders[i].prgId))");
4916 context_release(context
);
4919 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
);
4920 HeapFree(GetProcessHeap(), 0, shader_data
);
4921 shader
->backend_data
= NULL
;
4925 static int sig_tree_compare(const void *key
, const struct wine_rb_entry
*entry
)
4927 struct ps_signature
*e
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
4928 return compare_sig(key
, &e
->sig
);
4931 static const struct wine_rb_functions sig_tree_functions
=
4939 static HRESULT
shader_arb_alloc(struct wined3d_device
*device
, const struct wined3d_vertex_pipe_ops
*vertex_pipe
,
4940 const struct fragment_pipeline
*fragment_pipe
)
4942 struct shader_arb_priv
*priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*priv
));
4943 struct fragment_caps fragment_caps
;
4944 void *vertex_priv
, *fragment_priv
;
4945 const struct wined3d_d3d_info
*d3d_info
= &device
->adapter
->d3d_info
;
4947 if (!(vertex_priv
= vertex_pipe
->vp_alloc(&arb_program_shader_backend
, priv
)))
4949 ERR("Failed to initialize vertex pipe.\n");
4950 HeapFree(GetProcessHeap(), 0, priv
);
4954 if (!(fragment_priv
= fragment_pipe
->alloc_private(&arb_program_shader_backend
, priv
)))
4956 ERR("Failed to initialize fragment pipe.\n");
4957 vertex_pipe
->vp_free(device
);
4958 HeapFree(GetProcessHeap(), 0, priv
);
4962 priv
->vshader_const_dirty
= HeapAlloc(GetProcessHeap(), 0,
4963 sizeof(*priv
->vshader_const_dirty
) * d3d_info
->limits
.vs_uniform_count
);
4964 if (!priv
->vshader_const_dirty
)
4966 memset(priv
->vshader_const_dirty
, 1,
4967 sizeof(*priv
->vshader_const_dirty
) * d3d_info
->limits
.vs_uniform_count
);
4969 priv
->pshader_const_dirty
= HeapAlloc(GetProcessHeap(), 0,
4970 sizeof(*priv
->pshader_const_dirty
) * d3d_info
->limits
.ps_uniform_count
);
4971 if (!priv
->pshader_const_dirty
)
4973 memset(priv
->pshader_const_dirty
, 1,
4974 sizeof(*priv
->pshader_const_dirty
) * d3d_info
->limits
.ps_uniform_count
);
4976 if(wine_rb_init(&priv
->signature_tree
, &sig_tree_functions
) == -1)
4978 ERR("RB tree init failed\n");
4982 priv
->vertex_pipe
= vertex_pipe
;
4983 priv
->fragment_pipe
= fragment_pipe
;
4984 fragment_pipe
->get_caps(&device
->adapter
->gl_info
, &fragment_caps
);
4985 priv
->ffp_proj_control
= fragment_caps
.wined3d_caps
& WINED3D_FRAGMENT_CAP_PROJ_CONTROL
;
4987 device
->vertex_priv
= vertex_priv
;
4988 device
->fragment_priv
= fragment_priv
;
4989 device
->shader_priv
= priv
;
4994 HeapFree(GetProcessHeap(), 0, priv
->pshader_const_dirty
);
4995 HeapFree(GetProcessHeap(), 0, priv
->vshader_const_dirty
);
4996 fragment_pipe
->free_private(device
);
4997 vertex_pipe
->vp_free(device
);
4998 HeapFree(GetProcessHeap(), 0, priv
);
4999 return E_OUTOFMEMORY
;
5002 static void release_signature(struct wine_rb_entry
*entry
, void *context
)
5004 struct ps_signature
*sig
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
5007 for (i
= 0; i
< sig
->sig
.element_count
; ++i
)
5009 HeapFree(GetProcessHeap(), 0, (char *)sig
->sig
.elements
[i
].semantic_name
);
5011 HeapFree(GetProcessHeap(), 0, sig
->sig
.elements
);
5012 HeapFree(GetProcessHeap(), 0, sig
);
5015 /* Context activation is done by the caller. */
5016 static void shader_arb_free(struct wined3d_device
*device
)
5018 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
5019 struct shader_arb_priv
*priv
= device
->shader_priv
;
5022 if (priv
->depth_blt_vprogram_id
)
5023 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_vprogram_id
));
5025 for (i
= 0; i
< WINED3D_GL_RES_TYPE_COUNT
; ++i
)
5027 if (priv
->depth_blt_fprogram_id_full
[i
])
5029 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_fprogram_id_full
[i
]));
5031 if (priv
->depth_blt_fprogram_id_masked
[i
])
5033 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_fprogram_id_masked
[i
]));
5037 wine_rb_destroy(&priv
->signature_tree
, release_signature
, NULL
);
5038 HeapFree(GetProcessHeap(), 0, priv
->pshader_const_dirty
);
5039 HeapFree(GetProcessHeap(), 0, priv
->vshader_const_dirty
);
5040 priv
->fragment_pipe
->free_private(device
);
5041 priv
->vertex_pipe
->vp_free(device
);
5042 HeapFree(GetProcessHeap(), 0, device
->shader_priv
);
5045 static BOOL
shader_arb_allocate_context_data(struct wined3d_context
*context
)
5050 static void shader_arb_free_context_data(struct wined3d_context
*context
)
5052 struct shader_arb_priv
*priv
;
5054 if (!context
->swapchain
)
5057 priv
= context
->swapchain
->device
->shader_priv
;
5058 if (priv
->last_context
== context
)
5059 priv
->last_context
= NULL
;
5062 static void shader_arb_get_caps(const struct wined3d_gl_info
*gl_info
, struct shader_caps
*caps
)
5064 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
5069 /* 96 is the minimum allowed value of MAX_PROGRAM_ENV_PARAMETERS_ARB
5070 * for vertex programs. If the native limit is less than that it's
5071 * not very useful, and e.g. Mesa swrast returns 0, probably to
5072 * indicate it's a software implementation. */
5073 if (gl_info
->limits
.arb_vs_native_constants
< 96)
5074 vs_consts
= gl_info
->limits
.arb_vs_float_constants
;
5076 vs_consts
= min(gl_info
->limits
.arb_vs_float_constants
, gl_info
->limits
.arb_vs_native_constants
);
5078 if (gl_info
->supported
[NV_VERTEX_PROGRAM3
])
5081 TRACE("Hardware vertex shader version 3.0 enabled (NV_VERTEX_PROGRAM3)\n");
5083 else if (vs_consts
>= 256)
5085 /* Shader Model 2.0 requires at least 256 vertex shader constants */
5087 TRACE("Hardware vertex shader version 2.0 enabled (ARB_PROGRAM)\n");
5092 TRACE("Hardware vertex shader version 1.1 enabled (ARB_PROGRAM)\n");
5094 caps
->vs_version
= min(wined3d_settings
.max_sm_vs
, vs_version
);
5095 caps
->vs_uniform_count
= vs_consts
;
5099 caps
->vs_version
= 0;
5100 caps
->vs_uniform_count
= 0;
5103 caps
->gs_version
= 0;
5105 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
5110 /* Similar as above for vertex programs, but the minimum for fragment
5111 * programs is 24. */
5112 if (gl_info
->limits
.arb_ps_native_constants
< 24)
5113 ps_consts
= gl_info
->limits
.arb_ps_float_constants
;
5115 ps_consts
= min(gl_info
->limits
.arb_ps_float_constants
, gl_info
->limits
.arb_ps_native_constants
);
5117 if (gl_info
->supported
[NV_FRAGMENT_PROGRAM2
])
5120 TRACE("Hardware pixel shader version 3.0 enabled (NV_FRAGMENT_PROGRAM2)\n");
5122 else if (ps_consts
>= 32)
5124 /* Shader Model 2.0 requires at least 32 pixel shader constants */
5126 TRACE("Hardware pixel shader version 2.0 enabled (ARB_PROGRAM)\n");
5131 TRACE("Hardware pixel shader version 1.4 enabled (ARB_PROGRAM)\n");
5133 caps
->ps_version
= min(wined3d_settings
.max_sm_ps
, ps_version
);
5134 caps
->ps_uniform_count
= ps_consts
;
5135 caps
->ps_1x_max_value
= 8.0f
;
5139 caps
->ps_version
= 0;
5140 caps
->ps_uniform_count
= 0;
5141 caps
->ps_1x_max_value
= 0.0f
;
5144 caps
->wined3d_caps
= WINED3D_SHADER_CAP_SRGB_WRITE
;
5145 if (use_nv_clip(gl_info
))
5146 caps
->wined3d_caps
|= WINED3D_SHADER_CAP_VS_CLIPPING
;
5149 static BOOL
shader_arb_color_fixup_supported(struct color_fixup_desc fixup
)
5151 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
5153 TRACE("Checking support for color_fixup:\n");
5154 dump_color_fixup_desc(fixup
);
5157 /* We support everything except complex conversions. */
5158 if (!is_complex_fixup(fixup
))
5164 TRACE("[FAILED]\n");
5168 static void shader_arb_add_instruction_modifiers(const struct wined3d_shader_instruction
*ins
) {
5170 char write_mask
[20], regstr
[50];
5171 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
5172 BOOL is_color
= FALSE
;
5173 const struct wined3d_shader_dst_param
*dst
;
5175 if (!ins
->dst_count
) return;
5179 if (!shift
) return; /* Saturate alone is handled by the instructions */
5181 shader_arb_get_write_mask(ins
, dst
, write_mask
);
5182 shader_arb_get_register_name(ins
, &dst
->reg
, regstr
, &is_color
);
5184 /* Generate a line that does the output modifier computation
5185 * FIXME: _SAT vs shift? _SAT alone is already handled in the instructions, if this
5186 * maps problems in e.g. _d4_sat modify shader_arb_get_modifier
5188 shader_addline(buffer
, "MUL%s %s%s, %s, %s;\n", shader_arb_get_modifier(ins
),
5189 regstr
, write_mask
, regstr
, shift_tab
[shift
]);
5192 static const SHADER_HANDLER shader_arb_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
5194 /* WINED3DSIH_ABS */ shader_hw_map2gl
,
5195 /* WINED3DSIH_ADD */ shader_hw_map2gl
,
5196 /* WINED3DSIH_AND */ NULL
,
5197 /* WINED3DSIH_BEM */ pshader_hw_bem
,
5198 /* WINED3DSIH_BREAK */ shader_hw_break
,
5199 /* WINED3DSIH_BREAKC */ shader_hw_breakc
,
5200 /* WINED3DSIH_BREAKP */ NULL
,
5201 /* WINED3DSIH_CALL */ shader_hw_call
,
5202 /* WINED3DSIH_CALLNZ */ NULL
,
5203 /* WINED3DSIH_CMP */ pshader_hw_cmp
,
5204 /* WINED3DSIH_CND */ pshader_hw_cnd
,
5205 /* WINED3DSIH_CRS */ shader_hw_map2gl
,
5206 /* WINED3DSIH_CUT */ NULL
,
5207 /* WINED3DSIH_DCL */ shader_hw_nop
,
5208 /* WINED3DSIH_DCL_CONSTANT_BUFFER */ shader_hw_nop
,
5209 /* WINED3DSIH_DCL_INPUT_PRIMITIVE */ shader_hw_nop
,
5210 /* WINED3DSIH_DCL_OUTPUT_TOPOLOGY */ shader_hw_nop
,
5211 /* WINED3DSIH_DCL_VERTICES_OUT */ shader_hw_nop
,
5212 /* WINED3DSIH_DEF */ shader_hw_nop
,
5213 /* WINED3DSIH_DEFB */ shader_hw_nop
,
5214 /* WINED3DSIH_DEFI */ shader_hw_nop
,
5215 /* WINED3DSIH_DIV */ NULL
,
5216 /* WINED3DSIH_DP2 */ NULL
,
5217 /* WINED3DSIH_DP2ADD */ pshader_hw_dp2add
,
5218 /* WINED3DSIH_DP3 */ shader_hw_map2gl
,
5219 /* WINED3DSIH_DP4 */ shader_hw_map2gl
,
5220 /* WINED3DSIH_DST */ shader_hw_map2gl
,
5221 /* WINED3DSIH_DSX */ shader_hw_map2gl
,
5222 /* WINED3DSIH_DSY */ shader_hw_dsy
,
5223 /* WINED3DSIH_ELSE */ shader_hw_else
,
5224 /* WINED3DSIH_EMIT */ NULL
,
5225 /* WINED3DSIH_ENDIF */ shader_hw_endif
,
5226 /* WINED3DSIH_ENDLOOP */ shader_hw_endloop
,
5227 /* WINED3DSIH_ENDREP */ shader_hw_endrep
,
5228 /* WINED3DSIH_EQ */ NULL
,
5229 /* WINED3DSIH_EXP */ shader_hw_scalar_op
,
5230 /* WINED3DSIH_EXPP */ shader_hw_scalar_op
,
5231 /* WINED3DSIH_FRC */ shader_hw_map2gl
,
5232 /* WINED3DSIH_FTOI */ NULL
,
5233 /* WINED3DSIH_GE */ NULL
,
5234 /* WINED3DSIH_IADD */ NULL
,
5235 /* WINED3DSIH_IEQ */ NULL
,
5236 /* WINED3DSIH_IF */ NULL
/* Hardcoded into the shader */,
5237 /* WINED3DSIH_IFC */ shader_hw_ifc
,
5238 /* WINED3DSIH_IGE */ NULL
,
5239 /* WINED3DSIH_IMUL */ NULL
,
5240 /* WINED3DSIH_ISHL */ NULL
,
5241 /* WINED3DSIH_ITOF */ NULL
,
5242 /* WINED3DSIH_LABEL */ shader_hw_label
,
5243 /* WINED3DSIH_LD */ NULL
,
5244 /* WINED3DSIH_LIT */ shader_hw_map2gl
,
5245 /* WINED3DSIH_LOG */ shader_hw_scalar_op
,
5246 /* WINED3DSIH_LOGP */ shader_hw_scalar_op
,
5247 /* WINED3DSIH_LOOP */ shader_hw_loop
,
5248 /* WINED3DSIH_LRP */ shader_hw_lrp
,
5249 /* WINED3DSIH_LT */ NULL
,
5250 /* WINED3DSIH_M3x2 */ shader_hw_mnxn
,
5251 /* WINED3DSIH_M3x3 */ shader_hw_mnxn
,
5252 /* WINED3DSIH_M3x4 */ shader_hw_mnxn
,
5253 /* WINED3DSIH_M4x3 */ shader_hw_mnxn
,
5254 /* WINED3DSIH_M4x4 */ shader_hw_mnxn
,
5255 /* WINED3DSIH_MAD */ shader_hw_map2gl
,
5256 /* WINED3DSIH_MAX */ shader_hw_map2gl
,
5257 /* WINED3DSIH_MIN */ shader_hw_map2gl
,
5258 /* WINED3DSIH_MOV */ shader_hw_mov
,
5259 /* WINED3DSIH_MOVA */ shader_hw_mov
,
5260 /* WINED3DSIH_MOVC */ NULL
,
5261 /* WINED3DSIH_MUL */ shader_hw_map2gl
,
5262 /* WINED3DSIH_NE */ NULL
,
5263 /* WINED3DSIH_NOP */ shader_hw_nop
,
5264 /* WINED3DSIH_NRM */ shader_hw_nrm
,
5265 /* WINED3DSIH_OR */ NULL
,
5266 /* WINED3DSIH_PHASE */ shader_hw_nop
,
5267 /* WINED3DSIH_POW */ shader_hw_pow
,
5268 /* WINED3DSIH_RCP */ shader_hw_scalar_op
,
5269 /* WINED3DSIH_REP */ shader_hw_rep
,
5270 /* WINED3DSIH_RET */ shader_hw_ret
,
5271 /* WINED3DSIH_ROUND_NI */ NULL
,
5272 /* WINED3DSIH_RSQ */ shader_hw_scalar_op
,
5273 /* WINED3DSIH_SAMPLE */ NULL
,
5274 /* WINED3DSIH_SAMPLE_GRAD */ NULL
,
5275 /* WINED3DSIH_SAMPLE_LOD */ NULL
,
5276 /* WINED3DSIH_SETP */ NULL
,
5277 /* WINED3DSIH_SGE */ shader_hw_map2gl
,
5278 /* WINED3DSIH_SGN */ shader_hw_sgn
,
5279 /* WINED3DSIH_SINCOS */ shader_hw_sincos
,
5280 /* WINED3DSIH_SLT */ shader_hw_map2gl
,
5281 /* WINED3DSIH_SQRT */ NULL
,
5282 /* WINED3DSIH_SUB */ shader_hw_map2gl
,
5283 /* WINED3DSIH_TEX */ pshader_hw_tex
,
5284 /* WINED3DSIH_TEXBEM */ pshader_hw_texbem
,
5285 /* WINED3DSIH_TEXBEML */ pshader_hw_texbem
,
5286 /* WINED3DSIH_TEXCOORD */ pshader_hw_texcoord
,
5287 /* WINED3DSIH_TEXDEPTH */ pshader_hw_texdepth
,
5288 /* WINED3DSIH_TEXDP3 */ pshader_hw_texdp3
,
5289 /* WINED3DSIH_TEXDP3TEX */ pshader_hw_texdp3tex
,
5290 /* WINED3DSIH_TEXKILL */ pshader_hw_texkill
,
5291 /* WINED3DSIH_TEXLDD */ shader_hw_texldd
,
5292 /* WINED3DSIH_TEXLDL */ shader_hw_texldl
,
5293 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_hw_texm3x2depth
,
5294 /* WINED3DSIH_TEXM3x2PAD */ pshader_hw_texm3x2pad
,
5295 /* WINED3DSIH_TEXM3x2TEX */ pshader_hw_texm3x2tex
,
5296 /* WINED3DSIH_TEXM3x3 */ pshader_hw_texm3x3
,
5297 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
5298 /* WINED3DSIH_TEXM3x3PAD */ pshader_hw_texm3x3pad
,
5299 /* WINED3DSIH_TEXM3x3SPEC */ pshader_hw_texm3x3spec
,
5300 /* WINED3DSIH_TEXM3x3TEX */ pshader_hw_texm3x3tex
,
5301 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_hw_texm3x3vspec
,
5302 /* WINED3DSIH_TEXREG2AR */ pshader_hw_texreg2ar
,
5303 /* WINED3DSIH_TEXREG2GB */ pshader_hw_texreg2gb
,
5304 /* WINED3DSIH_TEXREG2RGB */ pshader_hw_texreg2rgb
,
5305 /* WINED3DSIH_UDIV */ NULL
,
5306 /* WINED3DSIH_UGE */ NULL
,
5307 /* WINED3DSIH_USHR */ NULL
,
5308 /* WINED3DSIH_UTOF */ NULL
,
5309 /* WINED3DSIH_XOR */ NULL
,
5312 static BOOL
get_bool_const(const struct wined3d_shader_instruction
*ins
,
5313 const struct wined3d_shader
*shader
, DWORD idx
)
5315 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
5316 BOOL vshader
= shader_is_vshader_version(reg_maps
->shader_version
.type
);
5317 const struct wined3d_shader_lconst
*constant
;
5319 WORD flag
= (1 << idx
);
5320 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5322 if (reg_maps
->local_bool_consts
& flag
)
5324 /* What good is an if(bool) with a hardcoded local constant? I don't know, but handle it */
5325 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsB
, struct wined3d_shader_lconst
, entry
)
5327 if (constant
->idx
== idx
)
5329 return constant
->value
[0];
5332 ERR("Local constant not found\n");
5337 if(vshader
) bools
= priv
->cur_vs_args
->clip
.boolclip
.bools
;
5338 else bools
= priv
->cur_ps_args
->bools
;
5339 return bools
& flag
;
5343 static void get_loop_control_const(const struct wined3d_shader_instruction
*ins
,
5344 const struct wined3d_shader
*shader
, UINT idx
, struct wined3d_shader_loop_control
*loop_control
)
5346 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
5347 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5349 /* Integer constants can either be a local constant, or they can be stored in the shader
5350 * type specific compile args. */
5351 if (reg_maps
->local_int_consts
& (1 << idx
))
5353 const struct wined3d_shader_lconst
*constant
;
5355 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsI
, struct wined3d_shader_lconst
, entry
)
5357 if (constant
->idx
== idx
)
5359 loop_control
->count
= constant
->value
[0];
5360 loop_control
->start
= constant
->value
[1];
5361 /* Step is signed. */
5362 loop_control
->step
= (int)constant
->value
[2];
5366 /* If this happens the flag was set incorrectly */
5367 ERR("Local constant not found\n");
5368 loop_control
->count
= 0;
5369 loop_control
->start
= 0;
5370 loop_control
->step
= 0;
5374 switch (reg_maps
->shader_version
.type
)
5376 case WINED3D_SHADER_TYPE_VERTEX
:
5377 /* Count and aL start value are unsigned */
5378 loop_control
->count
= priv
->cur_vs_args
->loop_ctrl
[idx
][0];
5379 loop_control
->start
= priv
->cur_vs_args
->loop_ctrl
[idx
][1];
5380 /* Step is signed. */
5381 loop_control
->step
= ((char)priv
->cur_vs_args
->loop_ctrl
[idx
][2]);
5384 case WINED3D_SHADER_TYPE_PIXEL
:
5385 loop_control
->count
= priv
->cur_ps_args
->loop_ctrl
[idx
][0];
5386 loop_control
->start
= priv
->cur_ps_args
->loop_ctrl
[idx
][1];
5387 loop_control
->step
= ((char)priv
->cur_ps_args
->loop_ctrl
[idx
][2]);
5391 FIXME("Unhandled shader type %#x.\n", reg_maps
->shader_version
.type
);
5396 static void record_instruction(struct list
*list
, const struct wined3d_shader_instruction
*ins
)
5399 struct wined3d_shader_dst_param
*dst_param
= NULL
;
5400 struct wined3d_shader_src_param
*src_param
= NULL
, *rel_addr
= NULL
;
5401 struct recorded_instruction
*rec
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*rec
));
5404 ERR("Out of memory\n");
5409 dst_param
= HeapAlloc(GetProcessHeap(), 0, sizeof(*dst_param
));
5410 if(!dst_param
) goto free
;
5411 *dst_param
= *ins
->dst
;
5412 if (ins
->dst
->reg
.idx
[0].rel_addr
)
5414 rel_addr
= HeapAlloc(GetProcessHeap(), 0, sizeof(*rel_addr
));
5417 *rel_addr
= *ins
->dst
->reg
.idx
[0].rel_addr
;
5418 dst_param
->reg
.idx
[0].rel_addr
= rel_addr
;
5420 rec
->ins
.dst
= dst_param
;
5422 src_param
= HeapAlloc(GetProcessHeap(), 0, sizeof(*src_param
) * ins
->src_count
);
5425 for (i
= 0; i
< ins
->src_count
; ++i
)
5427 src_param
[i
] = ins
->src
[i
];
5428 if (ins
->src
[i
].reg
.idx
[0].rel_addr
)
5430 rel_addr
= HeapAlloc(GetProcessHeap(), 0, sizeof(*rel_addr
));
5433 *rel_addr
= *ins
->src
[i
].reg
.idx
[0].rel_addr
;
5434 src_param
[i
].reg
.idx
[0].rel_addr
= rel_addr
;
5437 rec
->ins
.src
= src_param
;
5438 list_add_tail(list
, &rec
->entry
);
5442 ERR("Out of memory\n");
5445 HeapFree(GetProcessHeap(), 0, (void *)dst_param
->reg
.idx
[0].rel_addr
);
5446 HeapFree(GetProcessHeap(), 0, dst_param
);
5450 for(i
= 0; i
< ins
->src_count
; i
++)
5452 HeapFree(GetProcessHeap(), 0, (void *)src_param
[i
].reg
.idx
[0].rel_addr
);
5454 HeapFree(GetProcessHeap(), 0, src_param
);
5456 HeapFree(GetProcessHeap(), 0, rec
);
5459 static void free_recorded_instruction(struct list
*list
)
5461 struct recorded_instruction
*rec_ins
, *entry2
;
5464 LIST_FOR_EACH_ENTRY_SAFE(rec_ins
, entry2
, list
, struct recorded_instruction
, entry
)
5466 list_remove(&rec_ins
->entry
);
5467 if (rec_ins
->ins
.dst
)
5469 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.dst
->reg
.idx
[0].rel_addr
);
5470 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.dst
);
5472 if (rec_ins
->ins
.src
)
5474 for (i
= 0; i
< rec_ins
->ins
.src_count
; ++i
)
5476 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.src
[i
].reg
.idx
[0].rel_addr
);
5478 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.src
);
5480 HeapFree(GetProcessHeap(), 0, rec_ins
);
5484 static void pop_control_frame(const struct wined3d_shader_instruction
*ins
)
5486 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5487 struct control_frame
*control_frame
;
5489 if (ins
->handler_idx
== WINED3DSIH_ENDLOOP
|| ins
->handler_idx
== WINED3DSIH_ENDREP
)
5491 struct list
*e
= list_head(&priv
->control_frames
);
5492 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5493 list_remove(&control_frame
->entry
);
5494 HeapFree(GetProcessHeap(), 0, control_frame
);
5497 else if (ins
->handler_idx
== WINED3DSIH_ENDIF
)
5499 /* Non-ifc ENDIFs were already handled previously. */
5500 struct list
*e
= list_head(&priv
->control_frames
);
5501 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5502 list_remove(&control_frame
->entry
);
5503 HeapFree(GetProcessHeap(), 0, control_frame
);
5507 static void shader_arb_handle_instruction(const struct wined3d_shader_instruction
*ins
) {
5508 SHADER_HANDLER hw_fct
;
5509 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5510 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
5511 struct control_frame
*control_frame
;
5512 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
5515 if(ins
->handler_idx
== WINED3DSIH_LOOP
|| ins
->handler_idx
== WINED3DSIH_REP
)
5517 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
5518 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5520 if(ins
->handler_idx
== WINED3DSIH_LOOP
) control_frame
->type
= LOOP
;
5521 if(ins
->handler_idx
== WINED3DSIH_REP
) control_frame
->type
= REP
;
5523 if(priv
->target_version
>= NV2
)
5525 control_frame
->no
.loop
= priv
->num_loops
++;
5530 /* Don't bother recording when we're in a not used if branch */
5536 if(!priv
->recording
)
5538 list_init(&priv
->record
);
5539 priv
->recording
= TRUE
;
5540 control_frame
->outer_loop
= TRUE
;
5541 get_loop_control_const(ins
, shader
, ins
->src
[0].reg
.idx
[0].offset
, &control_frame
->loop_control
);
5542 return; /* Instruction is handled */
5544 /* Record this loop in the outer loop's recording */
5547 else if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
|| ins
->handler_idx
== WINED3DSIH_ENDREP
)
5549 if(priv
->target_version
>= NV2
)
5551 /* Nothing to do. The control frame is popped after the HW instr handler */
5555 struct list
*e
= list_head(&priv
->control_frames
);
5556 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5557 list_remove(&control_frame
->entry
);
5559 if(control_frame
->outer_loop
)
5561 unsigned int iteration
;
5565 /* Turn off recording before playback */
5566 priv
->recording
= FALSE
;
5568 /* Move the recorded instructions to a separate list and get them out of the private data
5569 * structure. If there are nested loops, the shader_arb_handle_instruction below will
5570 * be recorded again, thus priv->record might be overwritten
5573 list_move_tail(©
, &priv
->record
);
5574 list_init(&priv
->record
);
5576 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5578 shader_addline(buffer
, "#unrolling loop: %u iterations, aL=%u, inc %d\n",
5579 control_frame
->loop_control
.count
, control_frame
->loop_control
.start
,
5580 control_frame
->loop_control
.step
);
5581 aL
= control_frame
->loop_control
.start
;
5585 shader_addline(buffer
, "#unrolling rep: %u iterations\n", control_frame
->loop_control
.count
);
5588 for (iteration
= 0; iteration
< control_frame
->loop_control
.count
; ++iteration
)
5590 struct recorded_instruction
*rec_ins
;
5591 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5594 shader_addline(buffer
, "#Iteration %u, aL=%d\n", iteration
, aL
);
5598 shader_addline(buffer
, "#Iteration %u\n", iteration
);
5601 LIST_FOR_EACH_ENTRY(rec_ins
, ©
, struct recorded_instruction
, entry
)
5603 shader_arb_handle_instruction(&rec_ins
->ins
);
5606 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5608 aL
+= control_frame
->loop_control
.step
;
5611 shader_addline(buffer
, "#end loop/rep\n");
5613 free_recorded_instruction(©
);
5614 HeapFree(GetProcessHeap(), 0, control_frame
);
5615 return; /* Instruction is handled */
5619 /* This is a nested loop. Proceed to the normal recording function */
5620 HeapFree(GetProcessHeap(), 0, control_frame
);
5627 record_instruction(&priv
->record
, ins
);
5632 if(ins
->handler_idx
== WINED3DSIH_IF
)
5634 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
5635 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5636 control_frame
->type
= IF
;
5638 bool_const
= get_bool_const(ins
, shader
, ins
->src
[0].reg
.idx
[0].offset
);
5639 if (ins
->src
[0].modifiers
== WINED3DSPSM_NOT
)
5640 bool_const
= !bool_const
;
5641 if (!priv
->muted
&& !bool_const
)
5643 shader_addline(buffer
, "#if(FALSE){\n");
5645 control_frame
->muting
= TRUE
;
5647 else shader_addline(buffer
, "#if(TRUE) {\n");
5649 return; /* Instruction is handled */
5651 else if(ins
->handler_idx
== WINED3DSIH_IFC
)
5653 /* IF(bool) and if_cond(a, b) use the same ELSE and ENDIF tokens */
5654 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
5655 control_frame
->type
= IFC
;
5656 control_frame
->no
.ifc
= priv
->num_ifcs
++;
5657 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5659 else if(ins
->handler_idx
== WINED3DSIH_ELSE
)
5661 struct list
*e
= list_head(&priv
->control_frames
);
5662 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5664 if(control_frame
->type
== IF
)
5666 shader_addline(buffer
, "#} else {\n");
5667 if(!priv
->muted
&& !control_frame
->muting
)
5670 control_frame
->muting
= TRUE
;
5672 else if(control_frame
->muting
) priv
->muted
= FALSE
;
5673 return; /* Instruction is handled. */
5675 /* In case of an ifc, generate a HW shader instruction */
5676 if (control_frame
->type
!= IFC
)
5677 ERR("Control frame does not match.\n");
5679 else if(ins
->handler_idx
== WINED3DSIH_ENDIF
)
5681 struct list
*e
= list_head(&priv
->control_frames
);
5682 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5684 if(control_frame
->type
== IF
)
5686 shader_addline(buffer
, "#} endif\n");
5687 if(control_frame
->muting
) priv
->muted
= FALSE
;
5688 list_remove(&control_frame
->entry
);
5689 HeapFree(GetProcessHeap(), 0, control_frame
);
5690 return; /* Instruction is handled */
5692 /* In case of an ifc, generate a HW shader instruction */
5693 if (control_frame
->type
!= IFC
)
5694 ERR("Control frame does not match.\n");
5699 pop_control_frame(ins
);
5703 /* Select handler */
5704 hw_fct
= shader_arb_instruction_handler_table
[ins
->handler_idx
];
5706 /* Unhandled opcode */
5709 FIXME("Backend can't handle opcode %#x\n", ins
->handler_idx
);
5714 pop_control_frame(ins
);
5716 shader_arb_add_instruction_modifiers(ins
);
5719 static BOOL
shader_arb_has_ffp_proj_control(void *shader_priv
)
5721 struct shader_arb_priv
*priv
= shader_priv
;
5723 return priv
->ffp_proj_control
;
5726 const struct wined3d_shader_backend_ops arb_program_shader_backend
=
5728 shader_arb_handle_instruction
,
5731 shader_arb_select_depth_blt
,
5732 shader_arb_deselect_depth_blt
,
5733 shader_arb_update_float_vertex_constants
,
5734 shader_arb_update_float_pixel_constants
,
5735 shader_arb_load_constants
,
5739 shader_arb_allocate_context_data
,
5740 shader_arb_free_context_data
,
5741 shader_arb_get_caps
,
5742 shader_arb_color_fixup_supported
,
5743 shader_arb_has_ffp_proj_control
,
5746 /* ARB_fragment_program fixed function pipeline replacement definitions */
5747 #define ARB_FFP_CONST_TFACTOR 0
5748 #define ARB_FFP_CONST_COLOR_KEY ((ARB_FFP_CONST_TFACTOR) + 1)
5749 #define ARB_FFP_CONST_SPECULAR_ENABLE ((ARB_FFP_CONST_COLOR_KEY) + 1)
5750 #define ARB_FFP_CONST_CONSTANT(i) ((ARB_FFP_CONST_SPECULAR_ENABLE) + 1 + i)
5751 #define ARB_FFP_CONST_BUMPMAT(i) ((ARB_FFP_CONST_CONSTANT(7)) + 1 + i)
5752 #define ARB_FFP_CONST_LUMINANCE(i) ((ARB_FFP_CONST_BUMPMAT(7)) + 1 + i)
5754 struct arbfp_ffp_desc
5756 struct ffp_frag_desc parent
;
5760 /* Context activation is done by the caller. */
5761 static void arbfp_enable(const struct wined3d_gl_info
*gl_info
, BOOL enable
)
5765 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
5766 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
5770 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
5771 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
5775 static void *arbfp_alloc(const struct wined3d_shader_backend_ops
*shader_backend
, void *shader_priv
)
5777 struct shader_arb_priv
*priv
;
5779 /* Share private data between the shader backend and the pipeline
5780 * replacement, if both are the arb implementation. This is needed to
5781 * figure out whether ARBfp should be disabled if no pixel shader is bound
5783 if (shader_backend
== &arb_program_shader_backend
)
5785 else if (!(priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*priv
))))
5788 if (wine_rb_init(&priv
->fragment_shaders
, &wined3d_ffp_frag_program_rb_functions
) == -1)
5790 ERR("Failed to initialize rbtree.\n");
5791 if (priv
!= shader_priv
)
5792 HeapFree(GetProcessHeap(), 0, priv
);
5795 priv
->use_arbfp_fixed_func
= TRUE
;
5800 /* Context activation is done by the caller. */
5801 static void arbfp_free_ffpshader(struct wine_rb_entry
*entry
, void *context
)
5803 const struct wined3d_gl_info
*gl_info
= context
;
5804 struct arbfp_ffp_desc
*entry_arb
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_ffp_desc
, parent
.entry
);
5806 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb
->shader
));
5807 checkGLcall("glDeleteProgramsARB(1, &entry_arb->shader)");
5808 HeapFree(GetProcessHeap(), 0, entry_arb
);
5811 /* Context activation is done by the caller. */
5812 static void arbfp_free(struct wined3d_device
*device
)
5814 struct shader_arb_priv
*priv
= device
->fragment_priv
;
5816 wine_rb_destroy(&priv
->fragment_shaders
, arbfp_free_ffpshader
, &device
->adapter
->gl_info
);
5817 priv
->use_arbfp_fixed_func
= FALSE
;
5819 if (device
->shader_backend
!= &arb_program_shader_backend
)
5821 HeapFree(GetProcessHeap(), 0, device
->fragment_priv
);
5825 static void arbfp_get_caps(const struct wined3d_gl_info
*gl_info
, struct fragment_caps
*caps
)
5827 caps
->wined3d_caps
= WINED3D_FRAGMENT_CAP_PROJ_CONTROL
5828 | WINED3D_FRAGMENT_CAP_SRGB_WRITE
5829 | WINED3D_FRAGMENT_CAP_COLOR_KEY
;
5830 caps
->PrimitiveMiscCaps
= WINED3DPMISCCAPS_TSSARGTEMP
;
5831 caps
->TextureOpCaps
= WINED3DTEXOPCAPS_DISABLE
|
5832 WINED3DTEXOPCAPS_SELECTARG1
|
5833 WINED3DTEXOPCAPS_SELECTARG2
|
5834 WINED3DTEXOPCAPS_MODULATE4X
|
5835 WINED3DTEXOPCAPS_MODULATE2X
|
5836 WINED3DTEXOPCAPS_MODULATE
|
5837 WINED3DTEXOPCAPS_ADDSIGNED2X
|
5838 WINED3DTEXOPCAPS_ADDSIGNED
|
5839 WINED3DTEXOPCAPS_ADD
|
5840 WINED3DTEXOPCAPS_SUBTRACT
|
5841 WINED3DTEXOPCAPS_ADDSMOOTH
|
5842 WINED3DTEXOPCAPS_BLENDCURRENTALPHA
|
5843 WINED3DTEXOPCAPS_BLENDFACTORALPHA
|
5844 WINED3DTEXOPCAPS_BLENDTEXTUREALPHA
|
5845 WINED3DTEXOPCAPS_BLENDDIFFUSEALPHA
|
5846 WINED3DTEXOPCAPS_BLENDTEXTUREALPHAPM
|
5847 WINED3DTEXOPCAPS_MODULATEALPHA_ADDCOLOR
|
5848 WINED3DTEXOPCAPS_MODULATECOLOR_ADDALPHA
|
5849 WINED3DTEXOPCAPS_MODULATEINVCOLOR_ADDALPHA
|
5850 WINED3DTEXOPCAPS_MODULATEINVALPHA_ADDCOLOR
|
5851 WINED3DTEXOPCAPS_DOTPRODUCT3
|
5852 WINED3DTEXOPCAPS_MULTIPLYADD
|
5853 WINED3DTEXOPCAPS_LERP
|
5854 WINED3DTEXOPCAPS_BUMPENVMAP
|
5855 WINED3DTEXOPCAPS_BUMPENVMAPLUMINANCE
;
5857 /* TODO: Implement WINED3DTEXOPCAPS_PREMODULATE */
5859 caps
->MaxTextureBlendStages
= 8;
5860 caps
->MaxSimultaneousTextures
= min(gl_info
->limits
.fragment_samplers
, 8);
5863 static void state_texfactor_arbfp(struct wined3d_context
*context
,
5864 const struct wined3d_state
*state
, DWORD state_id
)
5866 struct wined3d_device
*device
= context
->swapchain
->device
;
5867 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5870 /* Don't load the parameter if we're using an arbfp pixel shader,
5871 * otherwise we'll overwrite application provided constants. */
5872 if (device
->shader_backend
== &arb_program_shader_backend
)
5874 struct shader_arb_priv
*priv
;
5876 if (use_ps(state
)) return;
5878 priv
= device
->shader_priv
;
5879 priv
->pshader_const_dirty
[ARB_FFP_CONST_TFACTOR
] = 1;
5880 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_TFACTOR
+ 1);
5883 D3DCOLORTOGLFLOAT4(state
->render_states
[WINED3D_RS_TEXTUREFACTOR
], col
);
5884 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_TFACTOR
, col
));
5885 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_TFACTOR, col)");
5888 static void state_arb_specularenable(struct wined3d_context
*context
,
5889 const struct wined3d_state
*state
, DWORD state_id
)
5891 struct wined3d_device
*device
= context
->swapchain
->device
;
5892 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5895 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
5896 * application provided constants
5898 if (device
->shader_backend
== &arb_program_shader_backend
)
5900 struct shader_arb_priv
*priv
;
5902 if (use_ps(state
)) return;
5904 priv
= device
->shader_priv
;
5905 priv
->pshader_const_dirty
[ARB_FFP_CONST_SPECULAR_ENABLE
] = 1;
5906 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_SPECULAR_ENABLE
+ 1);
5909 if (state
->render_states
[WINED3D_RS_SPECULARENABLE
])
5911 /* The specular color has no alpha */
5912 col
[0] = 1.0f
; col
[1] = 1.0f
;
5913 col
[2] = 1.0f
; col
[3] = 0.0f
;
5915 col
[0] = 0.0f
; col
[1] = 0.0f
;
5916 col
[2] = 0.0f
; col
[3] = 0.0f
;
5918 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_SPECULAR_ENABLE
, col
));
5919 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_SPECULAR_ENABLE, col)");
5922 static void set_bumpmat_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
5924 DWORD stage
= (state_id
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
5925 struct wined3d_device
*device
= context
->swapchain
->device
;
5926 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5929 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_BUMP_ENV
;
5931 if (device
->shader_backend
== &arb_program_shader_backend
)
5933 struct shader_arb_priv
*priv
= device
->shader_priv
;
5935 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants. */
5939 priv
->pshader_const_dirty
[ARB_FFP_CONST_BUMPMAT(stage
)] = 1;
5940 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_BUMPMAT(stage
) + 1);
5943 mat
[0][0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT00
]);
5944 mat
[0][1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT01
]);
5945 mat
[1][0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT10
]);
5946 mat
[1][1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT11
]);
5948 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_BUMPMAT(stage
), &mat
[0][0]));
5949 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_BUMPMAT(stage), &mat[0][0])");
5952 static void tex_bumpenvlum_arbfp(struct wined3d_context
*context
,
5953 const struct wined3d_state
*state
, DWORD state_id
)
5955 DWORD stage
= (state_id
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
5956 struct wined3d_device
*device
= context
->swapchain
->device
;
5957 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5960 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_BUMP_ENV
;
5962 if (device
->shader_backend
== &arb_program_shader_backend
)
5964 struct shader_arb_priv
*priv
= device
->shader_priv
;
5966 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants. */
5970 priv
->pshader_const_dirty
[ARB_FFP_CONST_LUMINANCE(stage
)] = 1;
5971 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_LUMINANCE(stage
) + 1);
5974 param
[0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_LSCALE
]);
5975 param
[1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_LOFFSET
]);
5979 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_LUMINANCE(stage
), param
));
5980 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_LUMINANCE(stage), param)");
5983 static void alpha_test_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
5985 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5989 TRACE("context %p, state %p, state_id %#x.\n", context
, state
, state_id
);
5991 if (state
->render_states
[WINED3D_RS_ALPHATESTENABLE
])
5993 gl_info
->gl_ops
.gl
.p_glEnable(GL_ALPHA_TEST
);
5994 checkGLcall("glEnable GL_ALPHA_TEST");
5998 gl_info
->gl_ops
.gl
.p_glDisable(GL_ALPHA_TEST
);
5999 checkGLcall("glDisable GL_ALPHA_TEST");
6003 ref
= ((float)state
->render_states
[WINED3D_RS_ALPHAREF
]) / 255.0f
;
6004 glParm
= wined3d_gl_compare_func(state
->render_states
[WINED3D_RS_ALPHAFUNC
]);
6008 gl_info
->gl_ops
.gl
.p_glAlphaFunc(glParm
, ref
);
6009 checkGLcall("glAlphaFunc");
6013 static void color_key_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6015 struct wined3d_device
*device
= context
->swapchain
->device
;
6016 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
6017 struct wined3d_color float_key
;
6018 const struct wined3d_texture
*texture
= state
->textures
[0];
6023 /* Don't load the parameter if we're using an arbfp pixel shader,
6024 * otherwise we'll overwrite application provided constants. */
6025 if (device
->shader_backend
== &arb_program_shader_backend
)
6027 struct shader_arb_priv
*priv
;
6032 priv
= device
->shader_priv
;
6033 priv
->pshader_const_dirty
[ARB_FFP_CONST_COLOR_KEY
] = 1;
6034 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_COLOR_KEY
+ 1);
6037 wined3d_format_convert_color_to_float(texture
->resource
.format
, NULL
,
6038 texture
->async
.src_blt_color_key
.color_space_high_value
, &float_key
);
6039 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_COLOR_KEY
, &float_key
.r
));
6040 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_COLOR_KEY, &float_key.r)");
6043 static const char *get_argreg(struct wined3d_string_buffer
*buffer
, DWORD argnum
, unsigned int stage
, DWORD arg
)
6047 if(arg
== ARG_UNUSED
) return "unused"; /* This is the marker for unused registers */
6049 switch(arg
& WINED3DTA_SELECTMASK
) {
6050 case WINED3DTA_DIFFUSE
:
6051 ret
= "fragment.color.primary"; break;
6053 case WINED3DTA_CURRENT
:
6054 if (!stage
) ret
= "fragment.color.primary";
6058 case WINED3DTA_TEXTURE
:
6060 case 0: ret
= "tex0"; break;
6061 case 1: ret
= "tex1"; break;
6062 case 2: ret
= "tex2"; break;
6063 case 3: ret
= "tex3"; break;
6064 case 4: ret
= "tex4"; break;
6065 case 5: ret
= "tex5"; break;
6066 case 6: ret
= "tex6"; break;
6067 case 7: ret
= "tex7"; break;
6068 default: ret
= "unknown texture";
6072 case WINED3DTA_TFACTOR
:
6073 ret
= "tfactor"; break;
6075 case WINED3DTA_SPECULAR
:
6076 ret
= "fragment.color.secondary"; break;
6078 case WINED3DTA_TEMP
:
6079 ret
= "tempreg"; break;
6081 case WINED3DTA_CONSTANT
:
6082 FIXME("Implement perstage constants\n");
6084 case 0: ret
= "const0"; break;
6085 case 1: ret
= "const1"; break;
6086 case 2: ret
= "const2"; break;
6087 case 3: ret
= "const3"; break;
6088 case 4: ret
= "const4"; break;
6089 case 5: ret
= "const5"; break;
6090 case 6: ret
= "const6"; break;
6091 case 7: ret
= "const7"; break;
6092 default: ret
= "unknown constant";
6100 if(arg
& WINED3DTA_COMPLEMENT
) {
6101 shader_addline(buffer
, "SUB arg%u, const.x, %s;\n", argnum
, ret
);
6102 if(argnum
== 0) ret
= "arg0";
6103 if(argnum
== 1) ret
= "arg1";
6104 if(argnum
== 2) ret
= "arg2";
6106 if(arg
& WINED3DTA_ALPHAREPLICATE
) {
6107 shader_addline(buffer
, "MOV arg%u, %s.w;\n", argnum
, ret
);
6108 if(argnum
== 0) ret
= "arg0";
6109 if(argnum
== 1) ret
= "arg1";
6110 if(argnum
== 2) ret
= "arg2";
6115 static void gen_ffp_instr(struct wined3d_string_buffer
*buffer
, unsigned int stage
, BOOL color
,
6116 BOOL alpha
, DWORD dst
, DWORD op
, DWORD dw_arg0
, DWORD dw_arg1
, DWORD dw_arg2
)
6118 const char *dstmask
, *dstreg
, *arg0
, *arg1
, *arg2
;
6119 unsigned int mul
= 1;
6121 if(color
&& alpha
) dstmask
= "";
6122 else if(color
) dstmask
= ".xyz";
6123 else dstmask
= ".w";
6125 if(dst
== tempreg
) dstreg
= "tempreg";
6126 else dstreg
= "ret";
6128 arg0
= get_argreg(buffer
, 0, stage
, dw_arg0
);
6129 arg1
= get_argreg(buffer
, 1, stage
, dw_arg1
);
6130 arg2
= get_argreg(buffer
, 2, stage
, dw_arg2
);
6134 case WINED3D_TOP_DISABLE
:
6136 shader_addline(buffer
, "MOV %s%s, fragment.color.primary;\n", dstreg
, dstmask
);
6139 case WINED3D_TOP_SELECT_ARG2
:
6142 case WINED3D_TOP_SELECT_ARG1
:
6143 shader_addline(buffer
, "MOV %s%s, %s;\n", dstreg
, dstmask
, arg1
);
6146 case WINED3D_TOP_MODULATE_4X
:
6149 case WINED3D_TOP_MODULATE_2X
:
6152 case WINED3D_TOP_MODULATE
:
6153 shader_addline(buffer
, "MUL %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6156 case WINED3D_TOP_ADD_SIGNED_2X
:
6159 case WINED3D_TOP_ADD_SIGNED
:
6160 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
6163 case WINED3D_TOP_ADD
:
6164 shader_addline(buffer
, "ADD_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6167 case WINED3D_TOP_SUBTRACT
:
6168 shader_addline(buffer
, "SUB_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6171 case WINED3D_TOP_ADD_SMOOTH
:
6172 shader_addline(buffer
, "SUB arg1, const.x, %s;\n", arg1
);
6173 shader_addline(buffer
, "MAD_SAT %s%s, arg1, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6176 case WINED3D_TOP_BLEND_CURRENT_ALPHA
:
6177 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_CURRENT
);
6178 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6180 case WINED3D_TOP_BLEND_FACTOR_ALPHA
:
6181 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TFACTOR
);
6182 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6184 case WINED3D_TOP_BLEND_TEXTURE_ALPHA
:
6185 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
6186 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6188 case WINED3D_TOP_BLEND_DIFFUSE_ALPHA
:
6189 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_DIFFUSE
);
6190 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6193 case WINED3D_TOP_BLEND_TEXTURE_ALPHA_PM
:
6194 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
6195 shader_addline(buffer
, "SUB arg0.w, const.x, %s.w;\n", arg0
);
6196 shader_addline(buffer
, "MAD_SAT %s%s, %s, arg0.w, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6199 /* D3DTOP_PREMODULATE ???? */
6201 case WINED3D_TOP_MODULATE_INVALPHA_ADD_COLOR
:
6202 shader_addline(buffer
, "SUB arg0.w, const.x, %s;\n", arg1
);
6203 shader_addline(buffer
, "MAD_SAT %s%s, arg0.w, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6205 case WINED3D_TOP_MODULATE_ALPHA_ADD_COLOR
:
6206 shader_addline(buffer
, "MAD_SAT %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
6208 case WINED3D_TOP_MODULATE_INVCOLOR_ADD_ALPHA
:
6209 shader_addline(buffer
, "SUB arg0, const.x, %s;\n", arg1
);
6210 shader_addline(buffer
, "MAD_SAT %s%s, arg0, %s, %s.w;\n", dstreg
, dstmask
, arg2
, arg1
);
6212 case WINED3D_TOP_MODULATE_COLOR_ADD_ALPHA
:
6213 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s.w;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
6216 case WINED3D_TOP_DOTPRODUCT3
:
6218 shader_addline(buffer
, "SUB arg1, %s, const.w;\n", arg1
);
6219 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
6220 shader_addline(buffer
, "DP3_SAT %s%s, arg1, arg2;\n", dstreg
, dstmask
);
6223 case WINED3D_TOP_MULTIPLY_ADD
:
6224 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg0
);
6227 case WINED3D_TOP_LERP
:
6228 /* The msdn is not quite right here */
6229 shader_addline(buffer
, "LRP %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6232 case WINED3D_TOP_BUMPENVMAP
:
6233 case WINED3D_TOP_BUMPENVMAP_LUMINANCE
:
6234 /* Those are handled in the first pass of the shader(generation pass 1 and 2) already */
6238 FIXME("Unhandled texture op %08x\n", op
);
6242 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.y;\n", dstreg
, dstmask
, dstreg
);
6244 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.z;\n", dstreg
, dstmask
, dstreg
);
6247 static const char *arbfp_texture_target(enum wined3d_gl_resource_type type
)
6251 case WINED3D_GL_RES_TYPE_TEX_1D
:
6253 case WINED3D_GL_RES_TYPE_TEX_2D
:
6255 case WINED3D_GL_RES_TYPE_TEX_3D
:
6257 case WINED3D_GL_RES_TYPE_TEX_CUBE
:
6259 case WINED3D_GL_RES_TYPE_TEX_RECT
:
6262 return "unexpected_resource_type";
6266 static GLuint
gen_arbfp_ffp_shader(const struct ffp_frag_settings
*settings
, const struct wined3d_gl_info
*gl_info
)
6269 struct wined3d_string_buffer buffer
;
6270 BOOL tex_read
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
6271 BOOL bump_used
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
6272 BOOL luminance_used
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
6273 UINT lowest_disabled_stage
;
6274 const char *textype
;
6276 char colorcor_dst
[8];
6278 DWORD arg0
, arg1
, arg2
;
6279 BOOL tempreg_used
= FALSE
, tfactor_used
= FALSE
;
6281 const char *final_combiner_src
= "ret";
6282 BOOL custom_linear_fog
= FALSE
;
6283 struct color_fixup_masks masks
;
6285 if (!string_buffer_init(&buffer
))
6287 ERR("Failed to initialize shader buffer.\n");
6291 shader_addline(&buffer
, "!!ARBfp1.0\n");
6293 if (settings
->color_key_enabled
)
6295 shader_addline(&buffer
, "PARAM color_key = program.env[%u];\n", ARB_FFP_CONST_COLOR_KEY
);
6299 /* Find out which textures are read */
6300 for (stage
= 0; stage
< MAX_TEXTURES
; ++stage
)
6302 if (settings
->op
[stage
].cop
== WINED3D_TOP_DISABLE
)
6304 arg0
= settings
->op
[stage
].carg0
& WINED3DTA_SELECTMASK
;
6305 arg1
= settings
->op
[stage
].carg1
& WINED3DTA_SELECTMASK
;
6306 arg2
= settings
->op
[stage
].carg2
& WINED3DTA_SELECTMASK
;
6307 if(arg0
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6308 if(arg1
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6309 if(arg2
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6311 if (settings
->op
[stage
].cop
== WINED3D_TOP_BLEND_TEXTURE_ALPHA
)
6312 tex_read
[stage
] = TRUE
;
6313 if (settings
->op
[stage
].cop
== WINED3D_TOP_BLEND_TEXTURE_ALPHA_PM
)
6314 tex_read
[stage
] = TRUE
;
6315 if (settings
->op
[stage
].cop
== WINED3D_TOP_BUMPENVMAP
)
6317 bump_used
[stage
] = TRUE
;
6318 tex_read
[stage
] = TRUE
;
6320 if (settings
->op
[stage
].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
)
6322 bump_used
[stage
] = TRUE
;
6323 tex_read
[stage
] = TRUE
;
6324 luminance_used
[stage
] = TRUE
;
6326 else if (settings
->op
[stage
].cop
== WINED3D_TOP_BLEND_FACTOR_ALPHA
)
6328 tfactor_used
= TRUE
;
6331 if(arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
) {
6332 tfactor_used
= TRUE
;
6335 if(settings
->op
[stage
].dst
== tempreg
) tempreg_used
= TRUE
;
6336 if(arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
) {
6337 tempreg_used
= TRUE
;
6340 if (settings
->op
[stage
].aop
== WINED3D_TOP_DISABLE
)
6342 arg0
= settings
->op
[stage
].aarg0
& WINED3DTA_SELECTMASK
;
6343 arg1
= settings
->op
[stage
].aarg1
& WINED3DTA_SELECTMASK
;
6344 arg2
= settings
->op
[stage
].aarg2
& WINED3DTA_SELECTMASK
;
6345 if(arg0
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6346 if(arg1
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6347 if(arg2
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6349 if(arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
) {
6350 tempreg_used
= TRUE
;
6352 if(arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
) {
6353 tfactor_used
= TRUE
;
6356 lowest_disabled_stage
= stage
;
6358 switch (settings
->fog
)
6360 case WINED3D_FFP_PS_FOG_OFF
: break;
6361 case WINED3D_FFP_PS_FOG_LINEAR
:
6362 if (gl_info
->quirks
& WINED3D_QUIRK_BROKEN_ARB_FOG
)
6364 custom_linear_fog
= TRUE
;
6367 shader_addline(&buffer
, "OPTION ARB_fog_linear;\n");
6370 case WINED3D_FFP_PS_FOG_EXP
: shader_addline(&buffer
, "OPTION ARB_fog_exp;\n"); break;
6371 case WINED3D_FFP_PS_FOG_EXP2
: shader_addline(&buffer
, "OPTION ARB_fog_exp2;\n"); break;
6372 default: FIXME("Unexpected fog setting %d\n", settings
->fog
);
6375 shader_addline(&buffer
, "PARAM const = {1, 2, 4, 0.5};\n");
6376 shader_addline(&buffer
, "TEMP TMP;\n");
6377 shader_addline(&buffer
, "TEMP ret;\n");
6378 if(tempreg_used
|| settings
->sRGB_write
) shader_addline(&buffer
, "TEMP tempreg;\n");
6379 shader_addline(&buffer
, "TEMP arg0;\n");
6380 shader_addline(&buffer
, "TEMP arg1;\n");
6381 shader_addline(&buffer
, "TEMP arg2;\n");
6382 for(stage
= 0; stage
< MAX_TEXTURES
; stage
++) {
6383 if(!tex_read
[stage
]) continue;
6384 shader_addline(&buffer
, "TEMP tex%u;\n", stage
);
6385 if(!bump_used
[stage
]) continue;
6386 shader_addline(&buffer
, "PARAM bumpmat%u = program.env[%u];\n", stage
, ARB_FFP_CONST_BUMPMAT(stage
));
6387 if(!luminance_used
[stage
]) continue;
6388 shader_addline(&buffer
, "PARAM luminance%u = program.env[%u];\n", stage
, ARB_FFP_CONST_LUMINANCE(stage
));
6391 shader_addline(&buffer
, "PARAM tfactor = program.env[%u];\n", ARB_FFP_CONST_TFACTOR
);
6393 shader_addline(&buffer
, "PARAM specular_enable = program.env[%u];\n", ARB_FFP_CONST_SPECULAR_ENABLE
);
6395 if (settings
->sRGB_write
)
6397 shader_addline(&buffer
, "PARAM srgb_consts0 = ");
6398 shader_arb_append_imm_vec4(&buffer
, wined3d_srgb_const0
);
6399 shader_addline(&buffer
, ";\n");
6400 shader_addline(&buffer
, "PARAM srgb_consts1 = ");
6401 shader_arb_append_imm_vec4(&buffer
, wined3d_srgb_const1
);
6402 shader_addline(&buffer
, ";\n");
6405 if (lowest_disabled_stage
< 7 && settings
->emul_clipplanes
)
6406 shader_addline(&buffer
, "KIL fragment.texcoord[7];\n");
6408 if (tempreg_used
|| settings
->sRGB_write
)
6409 shader_addline(&buffer
, "MOV tempreg, 0.0;\n");
6411 /* Generate texture sampling instructions) */
6412 for (stage
= 0; stage
< MAX_TEXTURES
&& settings
->op
[stage
].cop
!= WINED3D_TOP_DISABLE
; ++stage
)
6414 if (!tex_read
[stage
])
6417 textype
= arbfp_texture_target(settings
->op
[stage
].tex_type
);
6419 if(settings
->op
[stage
].projected
== proj_none
) {
6421 } else if(settings
->op
[stage
].projected
== proj_count4
||
6422 settings
->op
[stage
].projected
== proj_count3
) {
6425 FIXME("Unexpected projection mode %d\n", settings
->op
[stage
].projected
);
6430 && (settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP
6431 || settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
))
6433 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, x, z, 0, 0;\n", stage
- 1);
6434 shader_addline(&buffer
, "DP3 ret.x, arg1, tex%u;\n", stage
- 1);
6435 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, y, w, 0, 0;\n", stage
- 1);
6436 shader_addline(&buffer
, "DP3 ret.y, arg1, tex%u;\n", stage
- 1);
6438 /* with projective textures, texbem only divides the static texture coord, not the displacement,
6439 * so multiply the displacement with the dividing parameter before passing it to TXP
6441 if (settings
->op
[stage
].projected
!= proj_none
) {
6442 if(settings
->op
[stage
].projected
== proj_count4
) {
6443 shader_addline(&buffer
, "MOV ret.w, fragment.texcoord[%u].w;\n", stage
);
6444 shader_addline(&buffer
, "MUL ret.xyz, ret, fragment.texcoord[%u].w, fragment.texcoord[%u];\n", stage
, stage
);
6446 shader_addline(&buffer
, "MOV ret.w, fragment.texcoord[%u].z;\n", stage
);
6447 shader_addline(&buffer
, "MAD ret.xyz, ret, fragment.texcoord[%u].z, fragment.texcoord[%u];\n", stage
, stage
);
6450 shader_addline(&buffer
, "ADD ret, ret, fragment.texcoord[%u];\n", stage
);
6453 shader_addline(&buffer
, "%s tex%u, ret, texture[%u], %s;\n",
6454 instr
, stage
, stage
, textype
);
6455 if (settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
)
6457 shader_addline(&buffer
, "MAD_SAT ret.x, tex%u.z, luminance%u.x, luminance%u.y;\n",
6458 stage
- 1, stage
- 1, stage
- 1);
6459 shader_addline(&buffer
, "MUL tex%u, tex%u, ret.x;\n", stage
, stage
);
6461 } else if(settings
->op
[stage
].projected
== proj_count3
) {
6462 shader_addline(&buffer
, "MOV ret, fragment.texcoord[%u];\n", stage
);
6463 shader_addline(&buffer
, "MOV ret.w, ret.z;\n");
6464 shader_addline(&buffer
, "%s tex%u, ret, texture[%u], %s;\n",
6465 instr
, stage
, stage
, textype
);
6467 shader_addline(&buffer
, "%s tex%u, fragment.texcoord[%u], texture[%u], %s;\n",
6468 instr
, stage
, stage
, stage
, textype
);
6471 sprintf(colorcor_dst
, "tex%u", stage
);
6472 masks
= calc_color_correction(settings
->op
[stage
].color_fixup
, WINED3DSP_WRITEMASK_ALL
);
6473 gen_color_correction(&buffer
, colorcor_dst
, colorcor_dst
, "const.x", "const.y",
6474 settings
->op
[stage
].color_fixup
, masks
);
6477 if (settings
->color_key_enabled
)
6479 shader_addline(&buffer
, "SUB TMP, tex0, color_key;\n");
6480 shader_addline(&buffer
, "DP4 TMP.b, TMP, TMP;\n");
6481 shader_addline(&buffer
, "SGE TMP, -TMP.b, 0.0;\n");
6482 shader_addline(&buffer
, "KIL -TMP;\n");
6485 /* Generate the main shader */
6486 for (stage
= 0; stage
< MAX_TEXTURES
; ++stage
)
6488 if (settings
->op
[stage
].cop
== WINED3D_TOP_DISABLE
)
6491 final_combiner_src
= "fragment.color.primary";
6495 if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG1
6496 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG1
)
6497 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
;
6498 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG1
6499 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG2
)
6500 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg2
;
6501 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG2
6502 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG1
)
6503 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg1
;
6504 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG2
6505 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG2
)
6506 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
6508 op_equal
= settings
->op
[stage
].aop
== settings
->op
[stage
].cop
6509 && settings
->op
[stage
].carg0
== settings
->op
[stage
].aarg0
6510 && settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
6511 && settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
6513 if (settings
->op
[stage
].aop
== WINED3D_TOP_DISABLE
)
6515 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].dst
,
6516 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6517 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6519 shader_addline(&buffer
, "MOV ret.w, fragment.color.primary.w;\n");
6523 gen_ffp_instr(&buffer
, stage
, TRUE
, TRUE
, settings
->op
[stage
].dst
,
6524 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6525 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6527 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].dst
,
6528 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6529 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6530 gen_ffp_instr(&buffer
, stage
, FALSE
, TRUE
, settings
->op
[stage
].dst
,
6531 settings
->op
[stage
].aop
, settings
->op
[stage
].aarg0
,
6532 settings
->op
[stage
].aarg1
, settings
->op
[stage
].aarg2
);
6536 if (settings
->sRGB_write
|| custom_linear_fog
)
6538 shader_addline(&buffer
, "MAD ret, fragment.color.secondary, specular_enable, %s;\n", final_combiner_src
);
6539 if (settings
->sRGB_write
)
6540 arbfp_add_sRGB_correction(&buffer
, "ret", "arg0", "arg1", "arg2", "tempreg", FALSE
);
6541 if (custom_linear_fog
)
6542 arbfp_add_linear_fog(&buffer
, "ret", "arg0");
6543 shader_addline(&buffer
, "MOV result.color, ret;\n");
6547 shader_addline(&buffer
, "MAD result.color, fragment.color.secondary, specular_enable, %s;\n",
6548 final_combiner_src
);
6552 shader_addline(&buffer
, "END\n");
6554 /* Generate the shader */
6555 GL_EXTCALL(glGenProgramsARB(1, &ret
));
6556 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, ret
));
6557 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
6559 string_buffer_free(&buffer
);
6563 static void fragment_prog_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6565 const struct wined3d_device
*device
= context
->swapchain
->device
;
6566 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
6567 struct shader_arb_priv
*priv
= device
->fragment_priv
;
6568 BOOL use_pshader
= use_ps(state
);
6569 struct ffp_frag_settings settings
;
6570 const struct arbfp_ffp_desc
*desc
;
6573 TRACE("context %p, state %p, state_id %#x.\n", context
, state
, state_id
);
6575 if (isStateDirty(context
, STATE_RENDER(WINED3D_RS_FOGENABLE
)))
6577 if (!use_pshader
&& device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
)
6579 /* Reload fixed function constants since they collide with the
6580 * pixel shader constants. */
6581 for (i
= 0; i
< MAX_TEXTURES
; ++i
)
6583 set_bumpmat_arbfp(context
, state
, STATE_TEXTURESTAGE(i
, WINED3D_TSS_BUMPENV_MAT00
));
6585 state_texfactor_arbfp(context
, state
, STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
));
6586 state_arb_specularenable(context
, state
, STATE_RENDER(WINED3D_RS_SPECULARENABLE
));
6587 color_key_arbfp(context
, state
, STATE_COLOR_KEY
);
6589 else if (use_pshader
)
6591 context
->shader_update_mask
|= 1 << WINED3D_SHADER_TYPE_PIXEL
;
6598 /* Find or create a shader implementing the fixed function pipeline
6599 * settings, then activate it. */
6600 gen_ffp_frag_op(context
, state
, &settings
, FALSE
);
6601 desc
= (const struct arbfp_ffp_desc
*)find_ffp_frag_shader(&priv
->fragment_shaders
, &settings
);
6603 struct arbfp_ffp_desc
*new_desc
= HeapAlloc(GetProcessHeap(), 0, sizeof(*new_desc
));
6606 ERR("Out of memory\n");
6610 new_desc
->parent
.settings
= settings
;
6611 new_desc
->shader
= gen_arbfp_ffp_shader(&settings
, gl_info
);
6612 add_ffp_frag_shader(&priv
->fragment_shaders
, &new_desc
->parent
);
6613 TRACE("Allocated fixed function replacement shader descriptor %p\n", new_desc
);
6617 /* Now activate the replacement program. GL_FRAGMENT_PROGRAM_ARB is already active (however, note the
6618 * comment above the shader_select call below). If e.g. GLSL is active, the shader_select call will
6621 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, desc
->shader
));
6622 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, desc->shader)");
6623 priv
->current_fprogram_id
= desc
->shader
;
6625 if (device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
)
6627 /* Reload fixed function constants since they collide with the
6628 * pixel shader constants. */
6629 for (i
= 0; i
< MAX_TEXTURES
; ++i
)
6631 set_bumpmat_arbfp(context
, state
, STATE_TEXTURESTAGE(i
, WINED3D_TSS_BUMPENV_MAT00
));
6633 state_texfactor_arbfp(context
, state
, STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
));
6634 state_arb_specularenable(context
, state
, STATE_RENDER(WINED3D_RS_SPECULARENABLE
));
6636 context
->last_was_pshader
= FALSE
;
6638 else if (!context
->last_was_pshader
)
6640 if (device
->shader_backend
== &arb_program_shader_backend
)
6641 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_F
;
6642 context
->last_was_pshader
= TRUE
;
6645 context
->shader_update_mask
|= 1 << WINED3D_SHADER_TYPE_PIXEL
;
6648 /* We can't link the fog states to the fragment state directly since the
6649 * vertex pipeline links them to FOGENABLE. A different linking in different
6650 * pipeline parts can't be expressed in the combined state table, so we need
6651 * to handle that with a forwarding function. The other invisible side effect
6652 * is that changing the fog start and fog end (which links to FOGENABLE in
6653 * vertex) results in the fragment_prog_arbfp function being called because
6654 * FOGENABLE is dirty, which calls this function here. */
6655 static void state_arbfp_fog(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6657 enum fogsource new_source
;
6658 DWORD fogstart
= state
->render_states
[WINED3D_RS_FOGSTART
];
6659 DWORD fogend
= state
->render_states
[WINED3D_RS_FOGEND
];
6661 TRACE("context %p, state %p, state_id %#x.\n", context
, state
, state_id
);
6663 if (!isStateDirty(context
, STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
)))
6664 fragment_prog_arbfp(context
, state
, state_id
);
6666 if (!state
->render_states
[WINED3D_RS_FOGENABLE
])
6669 if (state
->render_states
[WINED3D_RS_FOGTABLEMODE
] == WINED3D_FOG_NONE
)
6673 new_source
= FOGSOURCE_VS
;
6677 if (state
->render_states
[WINED3D_RS_FOGVERTEXMODE
] == WINED3D_FOG_NONE
|| context
->last_was_rhw
)
6678 new_source
= FOGSOURCE_COORD
;
6680 new_source
= FOGSOURCE_FFP
;
6685 new_source
= FOGSOURCE_FFP
;
6688 if (new_source
!= context
->fog_source
|| fogstart
== fogend
)
6690 context
->fog_source
= new_source
;
6691 state_fogstartend(context
, state
, STATE_RENDER(WINED3D_RS_FOGSTART
));
6695 static void textransform(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6697 if (!isStateDirty(context
, STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
)))
6698 fragment_prog_arbfp(context
, state
, state_id
);
6701 static const struct StateEntryTemplate arbfp_fragmentstate_template
[] =
6703 {STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
), { STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
), state_texfactor_arbfp
}, WINED3D_GL_EXT_NONE
},
6704 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6705 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6706 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6707 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6708 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6709 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6710 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6711 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6712 {STATE_TEXTURESTAGE(0, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6713 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6714 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6715 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6716 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6717 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6718 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6719 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6720 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6721 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6722 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6723 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6724 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6725 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6726 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6727 {STATE_TEXTURESTAGE(1, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6728 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6729 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6730 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6731 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6732 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6733 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6734 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6735 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6736 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6737 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6738 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6739 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6740 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6741 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6742 {STATE_TEXTURESTAGE(2, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6743 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6744 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6745 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6746 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6747 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6748 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6749 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6750 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6751 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6752 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6753 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6754 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6755 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6756 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6757 {STATE_TEXTURESTAGE(3, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6758 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6759 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6760 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6761 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6762 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6763 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6764 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6765 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6766 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6767 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6768 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6769 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6770 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6771 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6772 {STATE_TEXTURESTAGE(4, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6773 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6774 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6775 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6776 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6777 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6778 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6779 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6780 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6781 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6782 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6783 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6784 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6785 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6786 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6787 {STATE_TEXTURESTAGE(5, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6788 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6789 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6790 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6791 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6792 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6793 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6794 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6795 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6796 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6797 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6798 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6799 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6800 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6801 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6802 {STATE_TEXTURESTAGE(6, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6803 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6804 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6805 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6806 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6807 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6808 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6809 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6810 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6811 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6812 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6813 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6814 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6815 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6816 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6817 {STATE_TEXTURESTAGE(7, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6818 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6819 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6820 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6821 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6822 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6823 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6824 {STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
6825 {STATE_RENDER(WINED3D_RS_ALPHAFUNC
), { STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6826 {STATE_RENDER(WINED3D_RS_ALPHAREF
), { STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6827 {STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), { STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), alpha_test_arbfp
}, WINED3D_GL_EXT_NONE
},
6828 {STATE_RENDER(WINED3D_RS_COLORKEYENABLE
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6829 {STATE_COLOR_KEY
, { STATE_COLOR_KEY
, color_key_arbfp
}, WINED3D_GL_EXT_NONE
},
6830 {STATE_RENDER(WINED3D_RS_FOGENABLE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), state_arbfp_fog
}, WINED3D_GL_EXT_NONE
},
6831 {STATE_RENDER(WINED3D_RS_FOGTABLEMODE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6832 {STATE_RENDER(WINED3D_RS_FOGVERTEXMODE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6833 {STATE_RENDER(WINED3D_RS_FOGSTART
), { STATE_RENDER(WINED3D_RS_FOGSTART
), state_fogstartend
}, WINED3D_GL_EXT_NONE
},
6834 {STATE_RENDER(WINED3D_RS_FOGEND
), { STATE_RENDER(WINED3D_RS_FOGSTART
), NULL
}, WINED3D_GL_EXT_NONE
},
6835 {STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), { STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), state_srgbwrite
}, ARB_FRAMEBUFFER_SRGB
},
6836 {STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6837 {STATE_RENDER(WINED3D_RS_FOGCOLOR
), { STATE_RENDER(WINED3D_RS_FOGCOLOR
), state_fogcolor
}, WINED3D_GL_EXT_NONE
},
6838 {STATE_RENDER(WINED3D_RS_FOGDENSITY
), { STATE_RENDER(WINED3D_RS_FOGDENSITY
), state_fogdensity
}, WINED3D_GL_EXT_NONE
},
6839 {STATE_TEXTURESTAGE(0,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(0, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6840 {STATE_TEXTURESTAGE(1,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(1, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6841 {STATE_TEXTURESTAGE(2,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(2, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6842 {STATE_TEXTURESTAGE(3,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(3, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6843 {STATE_TEXTURESTAGE(4,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(4, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6844 {STATE_TEXTURESTAGE(5,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(5, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6845 {STATE_TEXTURESTAGE(6,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(6, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6846 {STATE_TEXTURESTAGE(7,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(7, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6847 {STATE_RENDER(WINED3D_RS_SPECULARENABLE
), { STATE_RENDER(WINED3D_RS_SPECULARENABLE
), state_arb_specularenable
}, WINED3D_GL_EXT_NONE
},
6848 {0 /* Terminate */, { 0, 0 }, WINED3D_GL_EXT_NONE
},
6851 static BOOL
arbfp_alloc_context_data(struct wined3d_context
*context
)
6856 static void arbfp_free_context_data(struct wined3d_context
*context
)
6860 const struct fragment_pipeline arbfp_fragment_pipeline
= {
6865 arbfp_alloc_context_data
,
6866 arbfp_free_context_data
,
6867 shader_arb_color_fixup_supported
,
6868 arbfp_fragmentstate_template
,
6871 struct arbfp_blit_type
6873 enum complex_fixup fixup
: 4;
6874 enum wined3d_gl_resource_type res_type
: 3;
6875 DWORD use_color_key
: 1;
6879 struct arbfp_blit_desc
6882 struct arbfp_blit_type type
;
6883 struct wine_rb_entry entry
;
6886 #define ARBFP_BLIT_PARAM_SIZE 0
6887 #define ARBFP_BLIT_PARAM_COLOR_KEY 1
6889 struct arbfp_blit_priv
6891 struct wine_rb_tree shaders
;
6892 GLuint palette_texture
;
6895 static int arbfp_blit_type_compare(const void *key
, const struct wine_rb_entry
*entry
)
6897 const struct arbfp_blit_type
*ka
= key
;
6898 const struct arbfp_blit_type
*kb
= &WINE_RB_ENTRY_VALUE(entry
, const struct arbfp_blit_desc
, entry
)->type
;
6900 return memcmp(ka
, kb
, sizeof(*ka
));
6903 /* Context activation is done by the caller. */
6904 static void arbfp_free_blit_shader(struct wine_rb_entry
*entry
, void *context
)
6906 const struct wined3d_gl_info
*gl_info
= context
;
6907 struct arbfp_blit_desc
*entry_arb
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_blit_desc
, entry
);
6909 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb
->shader
));
6910 checkGLcall("glDeleteProgramsARB(1, &entry_arb->shader)");
6911 HeapFree(GetProcessHeap(), 0, entry_arb
);
6914 static const struct wine_rb_functions wined3d_arbfp_blit_rb_functions
=
6919 arbfp_blit_type_compare
,
6922 static HRESULT
arbfp_blit_alloc(struct wined3d_device
*device
)
6924 struct arbfp_blit_priv
*priv
;
6926 if (!(priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*priv
))))
6927 return E_OUTOFMEMORY
;
6929 if (wine_rb_init(&priv
->shaders
, &wined3d_arbfp_blit_rb_functions
) == -1)
6931 ERR("Failed to initialize rbtree.\n");
6932 HeapFree(GetProcessHeap(), 0, priv
);
6933 return E_OUTOFMEMORY
;
6936 device
->blit_priv
= priv
;
6941 /* Context activation is done by the caller. */
6942 static void arbfp_blit_free(struct wined3d_device
*device
)
6944 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
6945 struct arbfp_blit_priv
*priv
= device
->blit_priv
;
6947 wine_rb_destroy(&priv
->shaders
, arbfp_free_blit_shader
, &device
->adapter
->gl_info
);
6948 checkGLcall("Delete blit programs");
6950 if (priv
->palette_texture
)
6951 gl_info
->gl_ops
.gl
.p_glDeleteTextures(1, &priv
->palette_texture
);
6953 HeapFree(GetProcessHeap(), 0, device
->blit_priv
);
6954 device
->blit_priv
= NULL
;
6957 static BOOL
gen_planar_yuv_read(struct wined3d_string_buffer
*buffer
, const struct arbfp_blit_type
*type
,
6961 const char *tex
, *texinstr
= "TXP";
6963 if (type
->fixup
== COMPLEX_FIXUP_UYVY
)
6974 tex
= arbfp_texture_target(type
->res_type
);
6975 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_RECT
)
6978 /* First we have to read the chroma values. This means we need at least two pixels(no filtering),
6979 * or 4 pixels(with filtering). To get the unmodified chromas, we have to rid ourselves of the
6980 * filtering when we sample the texture.
6982 * These are the rules for reading the chroma:
6988 * So we have to get the sampling x position in non-normalized coordinates in integers
6990 if (type
->res_type
!= WINED3D_GL_RES_TYPE_TEX_RECT
)
6992 shader_addline(buffer
, "MUL texcrd.xy, fragment.texcoord[0], size.x;\n");
6993 shader_addline(buffer
, "MOV texcrd.w, size.x;\n");
6997 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
6999 /* We must not allow filtering between pixel x and x+1, this would mix U and V
7000 * Vertical filtering is ok. However, bear in mind that the pixel center is at
7003 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
7004 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, coef.y;\n");
7006 /* Divide the x coordinate by 0.5 and get the fraction. This gives 0.25 and 0.75 for the
7007 * even and odd pixels respectively
7009 shader_addline(buffer
, "MUL texcrd2, texcrd, coef.y;\n");
7010 shader_addline(buffer
, "FRC texcrd2, texcrd2;\n");
7012 /* Sample Pixel 1 */
7013 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
7015 /* Put the value into either of the chroma values */
7016 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
7017 shader_addline(buffer
, "MUL chroma.x, luminance.%c, temp.x;\n", chroma
);
7018 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
7019 shader_addline(buffer
, "MUL chroma.y, luminance.%c, temp.x;\n", chroma
);
7021 /* Sample pixel 2. If we read an even pixel(SLT above returned 1), sample
7022 * the pixel right to the current one. Otherwise, sample the left pixel.
7023 * Bias and scale the SLT result to -1;1 and add it to the texcrd.x.
7025 shader_addline(buffer
, "MAD temp.x, temp.x, coef.z, -coef.x;\n");
7026 shader_addline(buffer
, "ADD texcrd.x, texcrd, temp.x;\n");
7027 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
7029 /* Put the value into the other chroma */
7030 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
7031 shader_addline(buffer
, "MAD chroma.y, luminance.%c, temp.x, chroma.y;\n", chroma
);
7032 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
7033 shader_addline(buffer
, "MAD chroma.x, luminance.%c, temp.x, chroma.x;\n", chroma
);
7035 /* TODO: If filtering is enabled, sample a 2nd pair of pixels left or right of
7036 * the current one and lerp the two U and V values
7039 /* This gives the correctly filtered luminance value */
7040 shader_addline(buffer
, "TEX luminance, fragment.texcoord[0], texture[0], %s;\n", tex
);
7045 static BOOL
gen_yv12_read(struct wined3d_string_buffer
*buffer
, const struct arbfp_blit_type
*type
,
7049 static const float yv12_coef
[]
7050 = {2.0f
/ 3.0f
, 1.0f
/ 6.0f
, (2.0f
/ 3.0f
) + (1.0f
/ 6.0f
), 1.0f
/ 3.0f
};
7052 tex
= arbfp_texture_target(type
->res_type
);
7054 /* YV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
7055 * V and a (W/2)x(H/2) U plane, each with 8 bit per pixel. So the effective
7056 * bitdepth is 12 bits per pixel. Since the U and V planes have only half the
7057 * pitch of the luminance plane, the packing into the gl texture is a bit
7058 * unfortunate. If the whole texture is interpreted as luminance data it looks
7059 * approximately like this:
7061 * +----------------------------------+----
7073 * +----------------+-----------------+----
7075 * | V even rows | V odd rows |
7077 * +----------------+------------------ -
7079 * | U even rows | U odd rows |
7081 * +----------------+-----------------+----
7085 * So it appears as if there are 4 chroma images, but in fact the odd rows
7086 * in the chroma images are in the same row as the even ones. So its is
7087 * kinda tricky to read
7089 * When reading from rectangle textures, keep in mind that the input y coordinates
7090 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height
7092 shader_addline(buffer
, "PARAM yv12_coef = ");
7093 shader_arb_append_imm_vec4(buffer
, yv12_coef
);
7094 shader_addline(buffer
, ";\n");
7096 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7097 /* the chroma planes have only half the width */
7098 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, coef.y;\n");
7100 /* The first value is between 2/3 and 5/6th of the texture's height, so scale+bias
7101 * the coordinate. Also read the right side of the image when reading odd lines
7103 * Don't forget to clamp the y values in into the range, otherwise we'll get filtering
7106 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7109 shader_addline(buffer
, "RCP chroma.w, size.y;\n");
7111 shader_addline(buffer
, "MUL texcrd2.y, texcrd.y, size.y;\n");
7113 shader_addline(buffer
, "FLR texcrd2.y, texcrd2.y;\n");
7114 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, yv12_coef.y, yv12_coef.x;\n");
7116 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
7117 shader_addline(buffer
, "ADD texcrd2.x, texcrd2.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
7118 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
7119 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
7120 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
7122 /* clamp, keep the half pixel origin in mind */
7123 shader_addline(buffer
, "MAD temp.y, coef.y, chroma.w, yv12_coef.x;\n");
7124 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7125 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.z;\n");
7126 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7130 /* Read from [size - size+size/4] */
7131 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
7132 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, coef.w, size.y;\n");
7134 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
7135 shader_addline(buffer
, "ADD texcrd2.x, texcrd.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
7136 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
7137 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
7138 shader_addline(buffer
, "MUL texcrd2.x, texcrd2.x, size.x;\n");
7139 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
7141 /* Make sure to read exactly from the pixel center */
7142 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
7143 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, coef.y;\n");
7146 shader_addline(buffer
, "MAD temp.y, size.y, coef.w, size.y;\n");
7147 shader_addline(buffer
, "ADD temp.y, temp.y, -coef.y;\n");
7148 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7149 shader_addline(buffer
, "ADD temp.y, size.y, coef.y;\n");
7150 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7152 /* Read the texture, put the result into the output register */
7153 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7154 shader_addline(buffer
, "MOV chroma.x, temp.w;\n");
7156 /* The other chroma value is 1/6th of the texture lower, from 5/6th to 6/6th
7157 * No need to clamp because we're just reusing the already clamped value from above
7159 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7160 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, yv12_coef.y;\n");
7162 shader_addline(buffer
, "MAD texcrd.y, size.y, coef.w, texcrd.y;\n");
7163 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7164 shader_addline(buffer
, "MOV chroma.y, temp.w;\n");
7166 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
7167 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
7168 * values due to filtering
7170 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7171 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7173 /* Multiply the y coordinate by 2/3 and clamp it */
7174 shader_addline(buffer
, "MUL texcrd.y, texcrd.y, yv12_coef.x;\n");
7175 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.x;\n");
7176 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7177 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7181 /* Reading from texture_rectangles is pretty straightforward, just use the unmodified
7182 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
7185 shader_addline(buffer
, "ADD temp.x, size.y, -coef.y;\n");
7186 shader_addline(buffer
, "MIN texcrd.y, texcrd.y, size.x;\n");
7187 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7194 static BOOL
gen_nv12_read(struct wined3d_string_buffer
*buffer
, const struct arbfp_blit_type
*type
,
7198 static const float nv12_coef
[]
7199 = {2.0f
/ 3.0f
, 1.0f
/ 3.0f
, 1.0f
, 1.0f
};
7201 tex
= arbfp_texture_target(type
->res_type
);
7203 /* NV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
7204 * sized plane where each component is an UV pair. So the effective
7205 * bitdepth is 12 bits per pixel If the whole texture is interpreted as luminance
7206 * data it looks approximately like this:
7208 * +----------------------------------+----
7220 * +----------------------------------+----
7221 * |UVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUV|
7222 * |UVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUV|
7228 * +----------------------------------+----
7230 * When reading from rectangle textures, keep in mind that the input y coordinates
7231 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height. */
7233 shader_addline(buffer
, "PARAM nv12_coef = ");
7234 shader_arb_append_imm_vec4(buffer
, nv12_coef
);
7235 shader_addline(buffer
, ";\n");
7237 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7238 /* We only have half the number of chroma pixels. */
7239 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, coef.y;\n");
7241 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7243 shader_addline(buffer
, "RCP chroma.w, size.x;\n");
7244 shader_addline(buffer
, "RCP chroma.z, size.y;\n");
7246 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, nv12_coef.y, nv12_coef.x;\n");
7248 /* We must not allow filtering horizontally, this would mix U and V.
7249 * Vertical filtering is ok. However, bear in mind that the pixel center is at
7250 * 0.5, so add 0.5. */
7252 /* Convert to non-normalized coordinates so we can find the
7253 * individual pixel. */
7254 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, size.x;\n");
7255 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
7256 /* Multiply by 2 since chroma components are stored in UV pixel pairs,
7257 * add 0.5 to hit the center of the pixel. */
7258 shader_addline(buffer
, "MAD texcrd.x, texcrd.x, coef.z, coef.y;\n");
7260 /* Convert back to normalized coordinates. */
7261 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, chroma.w;\n");
7263 /* Clamp, keep the half pixel origin in mind. */
7264 shader_addline(buffer
, "MAD temp.y, coef.y, chroma.z, nv12_coef.x;\n");
7265 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7266 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.z, nv12_coef.z;\n");
7267 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7271 /* Read from [size - size+size/2] */
7272 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, coef.y, size.y;\n");
7274 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
7275 /* Multiply by 2 since chroma components are stored in UV pixel pairs,
7276 * add 0.5 to hit the center of the pixel. */
7277 shader_addline(buffer
, "MAD texcrd.x, texcrd.x, coef.z, coef.y;\n");
7280 shader_addline(buffer
, "MAD temp.y, size.y, coef.y, size.y;\n");
7281 shader_addline(buffer
, "ADD temp.y, temp.y, -coef.y;\n");
7282 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7283 shader_addline(buffer
, "ADD temp.y, size.y, coef.y;\n");
7284 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7286 /* Read the texture, put the result into the output register. */
7287 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7288 shader_addline(buffer
, "MOV chroma.y, temp.w;\n");
7290 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7293 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, chroma.w;\n");
7298 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, coef.x;\n");
7300 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7301 shader_addline(buffer
, "MOV chroma.x, temp.w;\n");
7303 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
7304 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
7305 * values due to filtering. */
7306 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7307 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7309 /* Multiply the y coordinate by 2/3 and clamp it */
7310 shader_addline(buffer
, "MUL texcrd.y, texcrd.y, nv12_coef.x;\n");
7311 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, nv12_coef.x;\n");
7312 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7313 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7317 /* Reading from texture_rectangles is pretty straightforward, just use the unmodified
7318 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
7321 shader_addline(buffer
, "ADD temp.x, size.y, -coef.y;\n");
7322 shader_addline(buffer
, "MIN texcrd.y, texcrd.y, size.x;\n");
7323 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7330 /* Context activation is done by the caller. */
7331 static GLuint
gen_p8_shader(struct arbfp_blit_priv
*priv
,
7332 const struct wined3d_gl_info
*gl_info
, const struct arbfp_blit_type
*type
)
7335 struct wined3d_string_buffer buffer
;
7336 const char *tex_target
= arbfp_texture_target(type
->res_type
);
7338 /* This should not happen because we only use this conversion for
7339 * present blits which don't use color keying. */
7340 if (type
->use_color_key
)
7341 FIXME("Implement P8 color keying.\n");
7344 if (!string_buffer_init(&buffer
))
7346 ERR("Failed to initialize shader buffer.\n");
7350 GL_EXTCALL(glGenProgramsARB(1, &shader
));
7351 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7354 string_buffer_free(&buffer
);
7358 shader_addline(&buffer
, "!!ARBfp1.0\n");
7359 shader_addline(&buffer
, "TEMP index;\n");
7361 /* { 255/256, 0.5/255*255/256, 0, 0 } */
7362 shader_addline(&buffer
, "PARAM constants = { 0.996, 0.00195, 0, 0 };\n");
7364 /* The alpha-component contains the palette index */
7365 shader_addline(&buffer
, "TEX index, fragment.texcoord[0], texture[0], %s;\n", tex_target
);
7367 /* Scale the index by 255/256 and add a bias of '0.5' in order to sample in the middle */
7368 shader_addline(&buffer
, "MAD index.a, index.a, constants.x, constants.y;\n");
7370 /* Use the alpha-component as an index in the palette to get the final color */
7371 shader_addline(&buffer
, "TEX result.color, index.a, texture[1], 1D;\n");
7372 shader_addline(&buffer
, "END\n");
7374 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
7376 string_buffer_free(&buffer
);
7381 /* Context activation is done by the caller. */
7382 static void upload_palette(const struct wined3d_texture
*texture
, struct wined3d_context
*context
)
7384 const struct wined3d_palette
*palette
= texture
->swapchain
? texture
->swapchain
->palette
: NULL
;
7385 struct wined3d_device
*device
= texture
->resource
.device
;
7386 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
7387 struct arbfp_blit_priv
*priv
= device
->blit_priv
;
7389 if (!priv
->palette_texture
)
7390 gl_info
->gl_ops
.gl
.p_glGenTextures(1, &priv
->palette_texture
);
7392 GL_EXTCALL(glActiveTexture(GL_TEXTURE1
));
7393 gl_info
->gl_ops
.gl
.p_glBindTexture(GL_TEXTURE_1D
, priv
->palette_texture
);
7395 gl_info
->gl_ops
.gl
.p_glTexEnvi(GL_TEXTURE_ENV
, GL_TEXTURE_ENV_MODE
, GL_REPLACE
);
7397 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_MIN_FILTER
, GL_NEAREST
);
7398 /* Make sure we have discrete color levels. */
7399 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_MAG_FILTER
, GL_NEAREST
);
7400 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_WRAP_S
, GL_CLAMP_TO_EDGE
);
7401 /* TODO: avoid unneeded uploads in the future by adding some SFLAG_PALETTE_DIRTY mechanism */
7404 gl_info
->gl_ops
.gl
.p_glTexImage1D(GL_TEXTURE_1D
, 0, GL_RGB
, 256, 0, GL_BGRA
,
7405 GL_UNSIGNED_INT_8_8_8_8_REV
, palette
->colors
);
7409 static const DWORD black
;
7410 FIXME("P8 surface loaded without a palette.\n");
7411 gl_info
->gl_ops
.gl
.p_glTexImage1D(GL_TEXTURE_1D
, 0, GL_RGB
, 1, 0, GL_BGRA
,
7412 GL_UNSIGNED_INT_8_8_8_8_REV
, &black
);
7415 /* Switch back to unit 0 in which the 2D texture will be stored. */
7416 context_active_texture(context
, gl_info
, 0);
7419 /* Context activation is done by the caller. */
7420 static GLuint
gen_yuv_shader(struct arbfp_blit_priv
*priv
, const struct wined3d_gl_info
*gl_info
,
7421 const struct arbfp_blit_type
*type
)
7424 struct wined3d_string_buffer buffer
;
7425 char luminance_component
;
7427 if (type
->use_color_key
)
7428 FIXME("Implement YUV color keying.\n");
7431 if (!string_buffer_init(&buffer
))
7433 ERR("Failed to initialize shader buffer.\n");
7437 GL_EXTCALL(glGenProgramsARB(1, &shader
));
7438 checkGLcall("GL_EXTCALL(glGenProgramsARB(1, &shader))");
7439 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7440 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
7443 string_buffer_free(&buffer
);
7447 /* The YUY2 and UYVY formats contain two pixels packed into a 32 bit macropixel,
7448 * giving effectively 16 bit per pixel. The color consists of a luminance(Y) and
7449 * two chroma(U and V) values. Each macropixel has two luminance values, one for
7450 * each single pixel it contains, and one U and one V value shared between both
7453 * The data is loaded into an A8L8 texture. With YUY2, the luminance component
7454 * contains the luminance and alpha the chroma. With UYVY it is vice versa. Thus
7455 * take the format into account when generating the read swizzles
7457 * Reading the Y value is straightforward - just sample the texture. The hardware
7458 * takes care of filtering in the horizontal and vertical direction.
7460 * Reading the U and V values is harder. We have to avoid filtering horizontally,
7461 * because that would mix the U and V values of one pixel or two adjacent pixels.
7462 * Thus floor the texture coordinate and add 0.5 to get an unfiltered read,
7463 * regardless of the filtering setting. Vertical filtering works automatically
7464 * though - the U and V values of two rows are mixed nicely.
7466 * Apart of avoiding filtering issues, the code has to know which value it just
7467 * read, and where it can find the other one. To determine this, it checks if
7468 * it sampled an even or odd pixel, and shifts the 2nd read accordingly.
7470 * Handling horizontal filtering of U and V values requires reading a 2nd pair
7471 * of pixels, extracting U and V and mixing them. This is not implemented yet.
7473 * An alternative implementation idea is to load the texture as A8R8G8B8 texture,
7474 * with width / 2. This way one read gives all 3 values, finding U and V is easy
7475 * in an unfiltered situation. Finding the luminance on the other hand requires
7476 * finding out if it is an odd or even pixel. The real drawback of this approach
7477 * is filtering. This would have to be emulated completely in the shader, reading
7478 * up two 2 packed pixels in up to 2 rows and interpolating both horizontally and
7479 * vertically. Beyond that it would require adjustments to the texture handling
7480 * code to deal with the width scaling
7482 shader_addline(&buffer
, "!!ARBfp1.0\n");
7483 shader_addline(&buffer
, "TEMP luminance;\n");
7484 shader_addline(&buffer
, "TEMP temp;\n");
7485 shader_addline(&buffer
, "TEMP chroma;\n");
7486 shader_addline(&buffer
, "TEMP texcrd;\n");
7487 shader_addline(&buffer
, "TEMP texcrd2;\n");
7488 shader_addline(&buffer
, "PARAM coef = {1.0, 0.5, 2.0, 0.25};\n");
7489 shader_addline(&buffer
, "PARAM yuv_coef = {1.403, 0.344, 0.714, 1.770};\n");
7490 shader_addline(&buffer
, "PARAM size = program.local[%u];\n", ARBFP_BLIT_PARAM_SIZE
);
7492 switch (type
->fixup
)
7494 case COMPLEX_FIXUP_UYVY
:
7495 case COMPLEX_FIXUP_YUY2
:
7496 if (!gen_planar_yuv_read(&buffer
, type
, &luminance_component
))
7498 string_buffer_free(&buffer
);
7503 case COMPLEX_FIXUP_YV12
:
7504 if (!gen_yv12_read(&buffer
, type
, &luminance_component
))
7506 string_buffer_free(&buffer
);
7511 case COMPLEX_FIXUP_NV12
:
7512 if (!gen_nv12_read(&buffer
, type
, &luminance_component
))
7514 string_buffer_free(&buffer
);
7520 FIXME("Unsupported YUV fixup %#x\n", type
->fixup
);
7521 string_buffer_free(&buffer
);
7525 /* Calculate the final result. Formula is taken from
7526 * http://www.fourcc.org/fccyvrgb.php. Note that the chroma
7527 * ranges from -0.5 to 0.5
7529 shader_addline(&buffer
, "SUB chroma.xy, chroma, coef.y;\n");
7531 shader_addline(&buffer
, "MAD result.color.x, chroma.x, yuv_coef.x, luminance.%c;\n", luminance_component
);
7532 shader_addline(&buffer
, "MAD temp.x, -chroma.y, yuv_coef.y, luminance.%c;\n", luminance_component
);
7533 shader_addline(&buffer
, "MAD result.color.y, -chroma.x, yuv_coef.z, temp.x;\n");
7534 shader_addline(&buffer
, "MAD result.color.z, chroma.y, yuv_coef.w, luminance.%c;\n", luminance_component
);
7535 shader_addline(&buffer
, "END\n");
7537 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
7539 string_buffer_free(&buffer
);
7544 /* Context activation is done by the caller. */
7545 static GLuint
arbfp_gen_plain_shader(struct arbfp_blit_priv
*priv
,
7546 const struct wined3d_gl_info
*gl_info
, const struct arbfp_blit_type
*type
)
7549 struct wined3d_string_buffer buffer
;
7550 const char *tex_target
= arbfp_texture_target(type
->res_type
);
7553 if (!string_buffer_init(&buffer
))
7555 ERR("Failed to initialize shader buffer.\n");
7559 GL_EXTCALL(glGenProgramsARB(1, &shader
));
7562 string_buffer_free(&buffer
);
7565 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7567 shader_addline(&buffer
, "!!ARBfp1.0\n");
7569 if (type
->use_color_key
)
7571 shader_addline(&buffer
, "TEMP color;\n");
7572 shader_addline(&buffer
, "TEMP compare;\n");
7573 shader_addline(&buffer
, "PARAM color_key = program.local[%u];\n", ARBFP_BLIT_PARAM_COLOR_KEY
);
7574 shader_addline(&buffer
, "TEX color, fragment.texcoord[0], texture[0], %s;\n", tex_target
);
7575 shader_addline(&buffer
, "SUB compare, color, color_key;\n");
7576 shader_addline(&buffer
, "DP4 compare.r, compare, compare;\n");
7577 shader_addline(&buffer
, "SGE compare, -compare.r, 0.0;\n");
7578 shader_addline(&buffer
, "KIL -compare;\n");
7579 shader_addline(&buffer
, "MOV result.color, color;\n");
7583 shader_addline(&buffer
, "TEX result.color, fragment.texcoord[0], texture[0], %s;\n", tex_target
);
7586 shader_addline(&buffer
, "END\n");
7588 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
7590 string_buffer_free(&buffer
);
7595 /* Context activation is done by the caller. */
7596 static HRESULT
arbfp_blit_set(void *blit_priv
, struct wined3d_context
*context
, const struct wined3d_surface
*surface
,
7597 const struct wined3d_color_key
*color_key
)
7600 float size
[4] = {(float) surface
->pow2Width
, (float) surface
->pow2Height
, 1.0f
, 1.0f
};
7601 struct arbfp_blit_priv
*priv
= blit_priv
;
7602 enum complex_fixup fixup
;
7603 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
7604 GLenum gl_texture_type
= surface
->container
->target
;
7605 struct wine_rb_entry
*entry
;
7606 struct arbfp_blit_type type
;
7607 struct arbfp_blit_desc
*desc
;
7608 struct wined3d_color float_color_key
;
7610 if (is_complex_fixup(surface
->resource
.format
->color_fixup
))
7611 fixup
= get_complex_fixup(surface
->resource
.format
->color_fixup
);
7613 fixup
= COMPLEX_FIXUP_NONE
;
7615 switch (gl_texture_type
)
7618 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_1D
;
7622 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_2D
;
7626 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_3D
;
7629 case GL_TEXTURE_CUBE_MAP_ARB
:
7630 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_CUBE
;
7633 case GL_TEXTURE_RECTANGLE_ARB
:
7634 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_RECT
;
7638 ERR("Unexpected GL texture type %x.\n", gl_texture_type
);
7639 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_2D
;
7642 type
.use_color_key
= !!color_key
;
7645 entry
= wine_rb_get(&priv
->shaders
, &type
);
7648 desc
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_blit_desc
, entry
);
7649 shader
= desc
->shader
;
7655 case COMPLEX_FIXUP_NONE
:
7656 if (!is_identity_fixup(surface
->resource
.format
->color_fixup
))
7657 FIXME("Implement support for sign or swizzle fixups.\n");
7658 shader
= arbfp_gen_plain_shader(priv
, gl_info
, &type
);
7661 case COMPLEX_FIXUP_P8
:
7662 shader
= gen_p8_shader(priv
, gl_info
, &type
);
7665 case COMPLEX_FIXUP_YUY2
:
7666 case COMPLEX_FIXUP_UYVY
:
7667 case COMPLEX_FIXUP_YV12
:
7668 case COMPLEX_FIXUP_NV12
:
7669 shader
= gen_yuv_shader(priv
, gl_info
, &type
);
7675 FIXME("Unsupported complex fixup %#x, not setting a shader\n", fixup
);
7679 desc
= HeapAlloc(GetProcessHeap(), 0, sizeof(*desc
));
7684 desc
->shader
= shader
;
7685 if (wine_rb_put(&priv
->shaders
, &desc
->type
, &desc
->entry
) == -1)
7688 ERR("Out of memory\n");
7689 GL_EXTCALL(glDeleteProgramsARB(1, &shader
));
7690 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader))");
7691 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, 0));
7692 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, 0)");
7693 HeapFree(GetProcessHeap(), 0, desc
);
7694 return E_OUTOFMEMORY
;
7698 if (fixup
== COMPLEX_FIXUP_P8
)
7699 upload_palette(surface
->container
, context
);
7701 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
7702 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
7703 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7704 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
7705 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARBFP_BLIT_PARAM_SIZE
, size
));
7706 checkGLcall("glProgramLocalParameter4fvARB");
7707 if (type
.use_color_key
)
7709 wined3d_format_convert_color_to_float(surface
->resource
.format
, NULL
,
7710 surface
->container
->async
.src_blt_color_key
.color_space_high_value
, &float_color_key
);
7711 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
7712 ARBFP_BLIT_PARAM_COLOR_KEY
, &float_color_key
.r
));
7713 checkGLcall("glProgramLocalParameter4fvARB");
7719 /* Context activation is done by the caller. */
7720 static void arbfp_blit_unset(const struct wined3d_gl_info
*gl_info
)
7722 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
7723 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
7726 static BOOL
arbfp_blit_supported(const struct wined3d_gl_info
*gl_info
,
7727 const struct wined3d_d3d_info
*d3d_info
, enum wined3d_blit_op blit_op
,
7728 const RECT
*src_rect
, DWORD src_usage
, enum wined3d_pool src_pool
, const struct wined3d_format
*src_format
,
7729 const RECT
*dst_rect
, DWORD dst_usage
, enum wined3d_pool dst_pool
, const struct wined3d_format
*dst_format
)
7731 enum complex_fixup src_fixup
;
7733 if (!gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
7738 case WINED3D_BLIT_OP_COLOR_BLIT_CKEY
:
7739 if (!d3d_info
->shader_color_key
)
7741 /* The conversion modifies the alpha channel so the color key might no longer match. */
7742 TRACE("Color keying not supported with converted textures.\n");
7745 case WINED3D_BLIT_OP_COLOR_BLIT
:
7749 TRACE("Unsupported blit_op=%d\n", blit_op
);
7753 if (src_pool
== WINED3D_POOL_SYSTEM_MEM
|| dst_pool
== WINED3D_POOL_SYSTEM_MEM
)
7756 src_fixup
= get_complex_fixup(src_format
->color_fixup
);
7757 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
7759 TRACE("Checking support for fixup:\n");
7760 dump_color_fixup_desc(src_format
->color_fixup
);
7763 if (!is_identity_fixup(dst_format
->color_fixup
))
7765 TRACE("Destination fixups are not supported\n");
7769 if (is_identity_fixup(src_format
->color_fixup
))
7775 /* We only support YUV conversions. */
7776 if (!is_complex_fixup(src_format
->color_fixup
))
7778 TRACE("[FAILED]\n");
7784 case COMPLEX_FIXUP_YUY2
:
7785 case COMPLEX_FIXUP_UYVY
:
7786 case COMPLEX_FIXUP_YV12
:
7787 case COMPLEX_FIXUP_NV12
:
7788 case COMPLEX_FIXUP_P8
:
7793 FIXME("Unsupported YUV fixup %#x\n", src_fixup
);
7794 TRACE("[FAILED]\n");
7799 static void arbfp_blit_surface(struct wined3d_device
*device
, DWORD filter
,
7800 struct wined3d_surface
*src_surface
, const RECT
*src_rect_in
,
7801 struct wined3d_surface
*dst_surface
, const RECT
*dst_rect_in
,
7802 const struct wined3d_color_key
*color_key
)
7804 struct wined3d_context
*context
;
7805 RECT src_rect
= *src_rect_in
;
7806 RECT dst_rect
= *dst_rect_in
;
7808 /* Activate the destination context, set it up for blitting */
7809 context
= context_acquire(device
, dst_surface
);
7811 /* Now load the surface */
7812 if (wined3d_settings
.offscreen_rendering_mode
!= ORM_FBO
7813 && (src_surface
->locations
& (WINED3D_LOCATION_TEXTURE_RGB
| WINED3D_LOCATION_DRAWABLE
))
7814 == WINED3D_LOCATION_DRAWABLE
7815 && !wined3d_resource_is_offscreen(&src_surface
->container
->resource
))
7817 /* Without FBO blits transferring from the drawable to the texture is
7818 * expensive, because we have to flip the data in sysmem. Since we can
7819 * flip in the blitter, we don't actually need that flip anyway. So we
7820 * use the surface's texture as scratch texture, and flip the source
7821 * rectangle instead. */
7822 surface_load_fb_texture(src_surface
, FALSE
);
7824 src_rect
.top
= src_surface
->resource
.height
- src_rect
.top
;
7825 src_rect
.bottom
= src_surface
->resource
.height
- src_rect
.bottom
;
7828 wined3d_texture_load(src_surface
->container
, context
, FALSE
);
7830 context_apply_blit_state(context
, device
);
7832 if (!wined3d_resource_is_offscreen(&dst_surface
->container
->resource
))
7833 surface_translate_drawable_coords(dst_surface
, context
->win_handle
, &dst_rect
);
7835 arbfp_blit_set(device
->blit_priv
, context
, src_surface
, color_key
);
7837 /* Draw a textured quad */
7838 draw_textured_quad(src_surface
, context
, &src_rect
, &dst_rect
, filter
);
7840 /* Leave the opengl state valid for blitting */
7841 arbfp_blit_unset(context
->gl_info
);
7843 if (wined3d_settings
.strict_draw_ordering
7844 || (dst_surface
->container
->swapchain
7845 && (dst_surface
->container
->swapchain
->front_buffer
== dst_surface
->container
)))
7846 context
->gl_info
->gl_ops
.gl
.p_glFlush(); /* Flush to ensure ordering across contexts. */
7848 context_release(context
);
7850 surface_validate_location(dst_surface
, dst_surface
->container
->resource
.draw_binding
);
7851 surface_invalidate_location(dst_surface
, ~dst_surface
->container
->resource
.draw_binding
);
7854 static HRESULT
arbfp_blit_color_fill(struct wined3d_device
*device
, struct wined3d_surface
*dst_surface
,
7855 const RECT
*dst_rect
, const struct wined3d_color
*color
)
7857 FIXME("Color filling not implemented by arbfp_blit\n");
7858 return WINED3DERR_INVALIDCALL
;
7861 static HRESULT
arbfp_blit_depth_fill(struct wined3d_device
*device
,
7862 struct wined3d_surface
*surface
, const RECT
*rect
, float depth
)
7864 FIXME("Depth filling not implemented by arbfp_blit.\n");
7865 return WINED3DERR_INVALIDCALL
;
7868 const struct blit_shader arbfp_blit
= {
7873 arbfp_blit_supported
,
7874 arbfp_blit_color_fill
,
7875 arbfp_blit_depth_fill
,