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 static void gen_color_correction(struct wined3d_string_buffer
*buffer
, const char *reg
,
1275 DWORD dst_mask
, const char *one
, const char *two
, struct color_fixup_desc fixup
)
1279 if (is_complex_fixup(fixup
))
1281 enum complex_fixup complex_fixup
= get_complex_fixup(fixup
);
1282 FIXME("Complex fixup (%#x) not supported\n", complex_fixup
);
1287 if (fixup
.x_source
!= CHANNEL_SOURCE_X
) mask
|= WINED3DSP_WRITEMASK_0
;
1288 if (fixup
.y_source
!= CHANNEL_SOURCE_Y
) mask
|= WINED3DSP_WRITEMASK_1
;
1289 if (fixup
.z_source
!= CHANNEL_SOURCE_Z
) mask
|= WINED3DSP_WRITEMASK_2
;
1290 if (fixup
.w_source
!= CHANNEL_SOURCE_W
) mask
|= WINED3DSP_WRITEMASK_3
;
1295 shader_addline(buffer
, "SWZ %s, %s, %s, %s, %s, %s;\n", reg
, reg
,
1296 shader_arb_get_fixup_swizzle(fixup
.x_source
), shader_arb_get_fixup_swizzle(fixup
.y_source
),
1297 shader_arb_get_fixup_swizzle(fixup
.z_source
), shader_arb_get_fixup_swizzle(fixup
.w_source
));
1301 if (fixup
.x_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_0
;
1302 if (fixup
.y_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_1
;
1303 if (fixup
.z_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_2
;
1304 if (fixup
.w_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_3
;
1310 char *ptr
= reg_mask
;
1312 if (mask
!= WINED3DSP_WRITEMASK_ALL
)
1315 if (mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
1316 if (mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
1317 if (mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
1318 if (mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
1322 shader_addline(buffer
, "MAD %s%s, %s, %s, -%s;\n", reg
, reg_mask
, reg
, two
, one
);
1326 static const char *shader_arb_get_modifier(const struct wined3d_shader_instruction
*ins
)
1329 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1330 if (!ins
->dst_count
) return "";
1332 mod
= ins
->dst
[0].modifiers
;
1334 /* Silently ignore PARTIALPRECISION if its not supported */
1335 if(priv
->target_version
== ARB
) mod
&= ~WINED3DSPDM_PARTIALPRECISION
;
1337 if(mod
& WINED3DSPDM_MSAMPCENTROID
)
1339 FIXME("Unhandled modifier WINED3DSPDM_MSAMPCENTROID\n");
1340 mod
&= ~WINED3DSPDM_MSAMPCENTROID
;
1345 case WINED3DSPDM_SATURATE
| WINED3DSPDM_PARTIALPRECISION
:
1348 case WINED3DSPDM_SATURATE
:
1351 case WINED3DSPDM_PARTIALPRECISION
:
1358 FIXME("Unknown modifiers 0x%08x\n", mod
);
1363 #define TEX_PROJ 0x1
1364 #define TEX_BIAS 0x2
1366 #define TEX_DERIV 0x10
1368 static void shader_hw_sample(const struct wined3d_shader_instruction
*ins
, DWORD sampler_idx
,
1369 const char *dst_str
, const char *coord_reg
, WORD flags
, const char *dsx
, const char *dsy
)
1371 enum wined3d_shader_resource_type resource_type
= ins
->ctx
->reg_maps
->resource_info
[sampler_idx
].type
;
1372 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1373 const char *tex_type
;
1374 BOOL np2_fixup
= FALSE
;
1375 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1377 BOOL pshader
= shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
1378 const struct wined3d_shader
*shader
;
1379 const struct wined3d_device
*device
;
1380 const struct wined3d_gl_info
*gl_info
;
1382 /* D3D vertex shader sampler IDs are vertex samplers(0-3), not global d3d samplers */
1383 if(!pshader
) sampler_idx
+= MAX_FRAGMENT_SAMPLERS
;
1385 switch (resource_type
)
1387 case WINED3D_SHADER_RESOURCE_TEXTURE_1D
:
1391 case WINED3D_SHADER_RESOURCE_TEXTURE_2D
:
1392 shader
= ins
->ctx
->shader
;
1393 device
= shader
->device
;
1394 gl_info
= &device
->adapter
->gl_info
;
1396 if (pshader
&& priv
->cur_ps_args
->super
.np2_fixup
& (1 << sampler_idx
)
1397 && gl_info
->supported
[ARB_TEXTURE_RECTANGLE
])
1401 if (shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
))
1403 if (priv
->cur_np2fixup_info
->super
.active
& (1 << sampler_idx
))
1405 if (flags
) FIXME("Only ordinary sampling from NP2 textures is supported.\n");
1406 else np2_fixup
= TRUE
;
1411 case WINED3D_SHADER_RESOURCE_TEXTURE_3D
:
1415 case WINED3D_SHADER_RESOURCE_TEXTURE_CUBE
:
1420 ERR("Unexpected resource type %#x.\n", resource_type
);
1424 /* TEX, TXL, TXD and TXP do not support the "H" modifier,
1425 * so don't use shader_arb_get_modifier
1427 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
) mod
= "_SAT";
1430 /* Fragment samplers always have indentity mapping */
1431 if(sampler_idx
>= MAX_FRAGMENT_SAMPLERS
)
1433 sampler_idx
= priv
->cur_vs_args
->vertex
.samplers
[sampler_idx
- MAX_FRAGMENT_SAMPLERS
];
1436 if (flags
& TEX_DERIV
)
1438 if(flags
& TEX_PROJ
) FIXME("Projected texture sampling with custom derivatives\n");
1439 if(flags
& TEX_BIAS
) FIXME("Biased texture sampling with custom derivatives\n");
1440 shader_addline(buffer
, "TXD%s %s, %s, %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
,
1441 dsx
, dsy
,sampler_idx
, tex_type
);
1443 else if(flags
& TEX_LOD
)
1445 if(flags
& TEX_PROJ
) FIXME("Projected texture sampling with explicit lod\n");
1446 if(flags
& TEX_BIAS
) FIXME("Biased texture sampling with explicit lod\n");
1447 shader_addline(buffer
, "TXL%s %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
,
1448 sampler_idx
, tex_type
);
1450 else if (flags
& TEX_BIAS
)
1452 /* Shouldn't be possible, but let's check for it */
1453 if(flags
& TEX_PROJ
) FIXME("Biased and Projected texture sampling\n");
1454 /* TXB takes the 4th component of the source vector automatically, as d3d. Nothing more to do */
1455 shader_addline(buffer
, "TXB%s %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
, sampler_idx
, tex_type
);
1457 else if (flags
& TEX_PROJ
)
1459 shader_addline(buffer
, "TXP%s %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
, sampler_idx
, tex_type
);
1465 const unsigned char idx
= priv
->cur_np2fixup_info
->super
.idx
[sampler_idx
];
1466 shader_addline(buffer
, "MUL TA, np2fixup[%u].%s, %s;\n", idx
>> 1,
1467 (idx
% 2) ? "zwxy" : "xyzw", coord_reg
);
1469 shader_addline(buffer
, "TEX%s %s, TA, texture[%u], %s;\n", mod
, dst_str
, sampler_idx
, tex_type
);
1472 shader_addline(buffer
, "TEX%s %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
, sampler_idx
, tex_type
);
1477 gen_color_correction(buffer
, dst_str
, ins
->dst
[0].write_mask
,
1478 arb_get_helper_value(WINED3D_SHADER_TYPE_PIXEL
, ARB_ONE
),
1479 arb_get_helper_value(WINED3D_SHADER_TYPE_PIXEL
, ARB_TWO
),
1480 priv
->cur_ps_args
->super
.color_fixup
[sampler_idx
]);
1484 static void shader_arb_get_src_param(const struct wined3d_shader_instruction
*ins
,
1485 const struct wined3d_shader_src_param
*src
, unsigned int tmpreg
, char *outregstr
)
1487 /* Generate a line that does the input modifier computation and return the input register to use */
1488 BOOL is_color
= FALSE
, insert_line
;
1491 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1492 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1493 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
1494 const char *two
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_TWO
);
1496 /* Assume a new line will be added */
1499 /* Get register name */
1500 shader_arb_get_register_name(ins
, &src
->reg
, regstr
, &is_color
);
1501 shader_arb_get_swizzle(src
, is_color
, swzstr
);
1503 switch (src
->modifiers
)
1505 case WINED3DSPSM_NONE
:
1506 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
1507 insert_line
= FALSE
;
1509 case WINED3DSPSM_NEG
:
1510 sprintf(outregstr
, "-%s%s", regstr
, swzstr
);
1511 insert_line
= FALSE
;
1513 case WINED3DSPSM_BIAS
:
1514 shader_addline(buffer
, "ADD T%c, %s, -coefdiv.x;\n", 'A' + tmpreg
, regstr
);
1516 case WINED3DSPSM_BIASNEG
:
1517 shader_addline(buffer
, "ADD T%c, -%s, coefdiv.x;\n", 'A' + tmpreg
, regstr
);
1519 case WINED3DSPSM_SIGN
:
1520 shader_addline(buffer
, "MAD T%c, %s, %s, -%s;\n", 'A' + tmpreg
, regstr
, two
, one
);
1522 case WINED3DSPSM_SIGNNEG
:
1523 shader_addline(buffer
, "MAD T%c, %s, -%s, %s;\n", 'A' + tmpreg
, regstr
, two
, one
);
1525 case WINED3DSPSM_COMP
:
1526 shader_addline(buffer
, "SUB T%c, %s, %s;\n", 'A' + tmpreg
, one
, regstr
);
1528 case WINED3DSPSM_X2
:
1529 shader_addline(buffer
, "ADD T%c, %s, %s;\n", 'A' + tmpreg
, regstr
, regstr
);
1531 case WINED3DSPSM_X2NEG
:
1532 shader_addline(buffer
, "ADD T%c, -%s, -%s;\n", 'A' + tmpreg
, regstr
, regstr
);
1534 case WINED3DSPSM_DZ
:
1535 shader_addline(buffer
, "RCP T%c, %s.z;\n", 'A' + tmpreg
, regstr
);
1536 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
1538 case WINED3DSPSM_DW
:
1539 shader_addline(buffer
, "RCP T%c, %s.w;\n", 'A' + tmpreg
, regstr
);
1540 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
1542 case WINED3DSPSM_ABS
:
1543 if(ctx
->target_version
>= NV2
) {
1544 sprintf(outregstr
, "|%s%s|", regstr
, swzstr
);
1545 insert_line
= FALSE
;
1547 shader_addline(buffer
, "ABS T%c, %s;\n", 'A' + tmpreg
, regstr
);
1550 case WINED3DSPSM_ABSNEG
:
1551 if(ctx
->target_version
>= NV2
) {
1552 sprintf(outregstr
, "-|%s%s|", regstr
, swzstr
);
1554 shader_addline(buffer
, "ABS T%c, %s;\n", 'A' + tmpreg
, regstr
);
1555 sprintf(outregstr
, "-T%c%s", 'A' + tmpreg
, swzstr
);
1557 insert_line
= FALSE
;
1560 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
1561 insert_line
= FALSE
;
1564 /* Return modified or original register, with swizzle */
1566 sprintf(outregstr
, "T%c%s", 'A' + tmpreg
, swzstr
);
1569 static void pshader_hw_bem(const struct wined3d_shader_instruction
*ins
)
1571 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1572 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1573 DWORD sampler_code
= dst
->reg
.idx
[0].offset
;
1575 char src_name
[2][50];
1577 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1579 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed
1581 * Keep in mind that src_name[1] can be "TB" and src_name[0] can be "TA" because modifiers like _x2 are valid
1582 * with bem. So delay loading the first parameter until after the perturbation calculation which needs two
1585 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1586 shader_addline(buffer
, "SWZ TA, bumpenvmat%d, x, z, 0, 0;\n", sampler_code
);
1587 shader_addline(buffer
, "DP3 TC.r, TA, %s;\n", src_name
[1]);
1588 shader_addline(buffer
, "SWZ TA, bumpenvmat%d, y, w, 0, 0;\n", sampler_code
);
1589 shader_addline(buffer
, "DP3 TC.g, TA, %s;\n", src_name
[1]);
1591 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1592 shader_addline(buffer
, "ADD %s, %s, TC;\n", dst_name
, src_name
[0]);
1595 static DWORD
negate_modifiers(DWORD mod
, char *extra_char
)
1600 case WINED3DSPSM_NONE
: return WINED3DSPSM_NEG
;
1601 case WINED3DSPSM_NEG
: return WINED3DSPSM_NONE
;
1602 case WINED3DSPSM_BIAS
: return WINED3DSPSM_BIASNEG
;
1603 case WINED3DSPSM_BIASNEG
: return WINED3DSPSM_BIAS
;
1604 case WINED3DSPSM_SIGN
: return WINED3DSPSM_SIGNNEG
;
1605 case WINED3DSPSM_SIGNNEG
: return WINED3DSPSM_SIGN
;
1606 case WINED3DSPSM_COMP
: *extra_char
= '-'; return WINED3DSPSM_COMP
;
1607 case WINED3DSPSM_X2
: return WINED3DSPSM_X2NEG
;
1608 case WINED3DSPSM_X2NEG
: return WINED3DSPSM_X2
;
1609 case WINED3DSPSM_DZ
: *extra_char
= '-'; return WINED3DSPSM_DZ
;
1610 case WINED3DSPSM_DW
: *extra_char
= '-'; return WINED3DSPSM_DW
;
1611 case WINED3DSPSM_ABS
: return WINED3DSPSM_ABSNEG
;
1612 case WINED3DSPSM_ABSNEG
: return WINED3DSPSM_ABS
;
1614 FIXME("Unknown modifier %u\n", mod
);
1618 static void pshader_hw_cnd(const struct wined3d_shader_instruction
*ins
)
1620 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1621 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1623 char src_name
[3][50];
1624 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1625 ins
->ctx
->reg_maps
->shader_version
.minor
);
1627 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1628 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1630 if (shader_version
<= WINED3D_SHADER_VERSION(1, 3) && ins
->coissue
1631 && ins
->dst
->write_mask
!= WINED3DSP_WRITEMASK_3
)
1633 shader_addline(buffer
, "MOV%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
[1]);
1637 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
1640 /* src0 may have a negate srcmod set, so we can't blindly add "-" to the name */
1641 src0_copy
.modifiers
= negate_modifiers(src0_copy
.modifiers
, &extra_neg
);
1643 shader_arb_get_src_param(ins
, &src0_copy
, 0, src_name
[0]);
1644 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1645 shader_addline(buffer
, "ADD TA, %c%s, coefdiv.x;\n", extra_neg
, src_name
[0]);
1646 shader_addline(buffer
, "CMP%s %s, TA, %s, %s;\n", shader_arb_get_modifier(ins
),
1647 dst_name
, src_name
[1], src_name
[2]);
1651 static void pshader_hw_cmp(const struct wined3d_shader_instruction
*ins
)
1653 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1654 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1656 char src_name
[3][50];
1658 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1660 /* Generate input register names (with modifiers) */
1661 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1662 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1663 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1665 shader_addline(buffer
, "CMP%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1666 dst_name
, src_name
[0], src_name
[2], src_name
[1]);
1669 /** Process the WINED3DSIO_DP2ADD instruction in ARB.
1670 * dst = dot2(src0, src1) + src2 */
1671 static void pshader_hw_dp2add(const struct wined3d_shader_instruction
*ins
)
1673 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1674 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1676 char src_name
[3][50];
1677 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1679 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1680 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1681 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1683 if(ctx
->target_version
>= NV3
)
1685 /* GL_NV_fragment_program2 has a 1:1 matching instruction */
1686 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1687 shader_addline(buffer
, "DP2A%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1688 dst_name
, src_name
[0], src_name
[1], src_name
[2]);
1690 else if(ctx
->target_version
>= NV2
)
1692 /* dst.x = src2.?, src0.x, src1.x + src0.y * src1.y
1693 * dst.y = src2.?, src0.x, src1.z + src0.y * src1.w
1694 * dst.z = src2.?, src0.x, src1.x + src0.y * src1.y
1695 * dst.z = src2.?, src0.x, src1.z + src0.y * src1.w
1697 * Make sure that src1.zw = src1.xy, then we get a classic dp2add
1699 * .xyxy and other swizzles that we could get with this are not valid in
1700 * plain ARBfp, but luckily the NV extension grammar lifts this limitation.
1702 struct wined3d_shader_src_param tmp_param
= ins
->src
[1];
1703 DWORD swizzle
= tmp_param
.swizzle
& 0xf; /* Selects .xy */
1704 tmp_param
.swizzle
= swizzle
| (swizzle
<< 4); /* Creates .xyxy */
1706 shader_arb_get_src_param(ins
, &tmp_param
, 1, src_name
[1]);
1708 shader_addline(buffer
, "X2D%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1709 dst_name
, src_name
[2], src_name
[0], src_name
[1]);
1713 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1714 /* 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]
1715 * src_name[0] can be TA, but TA is a private temp for modifiers, so it is save to overwrite
1717 shader_addline(buffer
, "MOV TA, %s;\n", src_name
[0]);
1718 shader_addline(buffer
, "MOV TA.z, 0.0;\n");
1719 shader_addline(buffer
, "DP3 TA, TA, %s;\n", src_name
[1]);
1720 shader_addline(buffer
, "ADD%s %s, TA, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
[2]);
1724 /* Map the opcode 1-to-1 to the GL code */
1725 static void shader_hw_map2gl(const struct wined3d_shader_instruction
*ins
)
1727 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1728 const char *instruction
;
1729 char arguments
[256], dst_str
[50];
1731 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1733 switch (ins
->handler_idx
)
1735 case WINED3DSIH_ABS
: instruction
= "ABS"; break;
1736 case WINED3DSIH_ADD
: instruction
= "ADD"; break;
1737 case WINED3DSIH_CRS
: instruction
= "XPD"; break;
1738 case WINED3DSIH_DP3
: instruction
= "DP3"; break;
1739 case WINED3DSIH_DP4
: instruction
= "DP4"; break;
1740 case WINED3DSIH_DST
: instruction
= "DST"; break;
1741 case WINED3DSIH_FRC
: instruction
= "FRC"; break;
1742 case WINED3DSIH_LIT
: instruction
= "LIT"; break;
1743 case WINED3DSIH_LRP
: instruction
= "LRP"; break;
1744 case WINED3DSIH_MAD
: instruction
= "MAD"; break;
1745 case WINED3DSIH_MAX
: instruction
= "MAX"; break;
1746 case WINED3DSIH_MIN
: instruction
= "MIN"; break;
1747 case WINED3DSIH_MOV
: instruction
= "MOV"; break;
1748 case WINED3DSIH_MUL
: instruction
= "MUL"; break;
1749 case WINED3DSIH_SGE
: instruction
= "SGE"; break;
1750 case WINED3DSIH_SLT
: instruction
= "SLT"; break;
1751 case WINED3DSIH_SUB
: instruction
= "SUB"; break;
1752 case WINED3DSIH_MOVA
:instruction
= "ARR"; break;
1753 case WINED3DSIH_DSX
: instruction
= "DDX"; break;
1754 default: instruction
= "";
1755 FIXME("Unhandled opcode %#x\n", ins
->handler_idx
);
1759 /* Note that shader_arb_add_dst_param() adds spaces. */
1760 arguments
[0] = '\0';
1761 shader_arb_get_dst_param(ins
, dst
, dst_str
);
1762 for (i
= 0; i
< ins
->src_count
; ++i
)
1765 strcat(arguments
, ", ");
1766 shader_arb_get_src_param(ins
, &ins
->src
[i
], i
, operand
);
1767 strcat(arguments
, operand
);
1769 shader_addline(buffer
, "%s%s %s%s;\n", instruction
, shader_arb_get_modifier(ins
), dst_str
, arguments
);
1772 static void shader_hw_nop(const struct wined3d_shader_instruction
*ins
) {}
1774 static DWORD
shader_arb_select_component(DWORD swizzle
, DWORD component
)
1776 return ((swizzle
>> 2 * component
) & 0x3) * 0x55;
1779 static void shader_hw_mov(const struct wined3d_shader_instruction
*ins
)
1781 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
1782 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
1783 BOOL pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
1784 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1785 const char *zero
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ZERO
);
1786 const char *one
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ONE
);
1787 const char *two
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_TWO
);
1789 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1790 char src0_param
[256];
1792 if (ins
->handler_idx
== WINED3DSIH_MOVA
)
1794 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1796 const char *offset
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_VS_REL_OFFSET
);
1798 if(ctx
->target_version
>= NV2
) {
1799 shader_hw_map2gl(ins
);
1802 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_param
);
1803 shader_arb_get_write_mask(ins
, &ins
->dst
[0], write_mask
);
1805 /* This implements the mova formula used in GLSL. The first two instructions
1806 * prepare the sign() part. Note that it is fine to have my_sign(0.0) = 1.0
1810 * A0.x = arl(floor(abs(0.0) + 0.5) * 1.0) = floor(0.5) = 0.0 since arl does a floor
1812 * The ARL is performed when A0 is used - the requested component is read from A0_SHADOW into
1813 * A0.x. We can use the overwritten component of A0_shadow as temporary storage for the sign.
1815 shader_addline(buffer
, "SGE A0_SHADOW%s, %s, %s;\n", write_mask
, src0_param
, zero
);
1816 shader_addline(buffer
, "MAD A0_SHADOW%s, A0_SHADOW, %s, -%s;\n", write_mask
, two
, one
);
1818 shader_addline(buffer
, "ABS TA%s, %s;\n", write_mask
, src0_param
);
1819 shader_addline(buffer
, "ADD TA%s, TA, rel_addr_const.x;\n", write_mask
);
1820 shader_addline(buffer
, "FLR TA%s, TA;\n", write_mask
);
1821 if (shader_data
->rel_offset
)
1823 shader_addline(buffer
, "ADD TA%s, TA, %s;\n", write_mask
, offset
);
1825 shader_addline(buffer
, "MUL A0_SHADOW%s, TA, A0_SHADOW;\n", write_mask
);
1827 ((struct shader_arb_ctx_priv
*)ins
->ctx
->backend_data
)->addr_reg
[0] = '\0';
1829 else if (reg_maps
->shader_version
.major
== 1
1830 && !shader_is_pshader_version(reg_maps
->shader_version
.type
)
1831 && ins
->dst
[0].reg
.type
== WINED3DSPR_ADDR
)
1833 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1834 src0_param
[0] = '\0';
1836 if (shader_data
->rel_offset
&& ctx
->target_version
== ARB
)
1838 const char *offset
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_VS_REL_OFFSET
);
1839 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_param
);
1840 shader_addline(buffer
, "ADD TA.x, %s, %s;\n", src0_param
, offset
);
1841 shader_addline(buffer
, "ARL A0.x, TA.x;\n");
1845 /* Apple's ARB_vertex_program implementation does not accept an ARL source argument
1846 * with more than one component. Thus replicate the first source argument over all
1847 * 4 components. For example, .xyzw -> .x (or better: .xxxx), .zwxy -> .z, etc) */
1848 struct wined3d_shader_src_param tmp_src
= ins
->src
[0];
1849 tmp_src
.swizzle
= shader_arb_select_component(tmp_src
.swizzle
, 0);
1850 shader_arb_get_src_param(ins
, &tmp_src
, 0, src0_param
);
1851 shader_addline(buffer
, "ARL A0.x, %s;\n", src0_param
);
1854 else if (ins
->dst
[0].reg
.type
== WINED3DSPR_COLOROUT
&& !ins
->dst
[0].reg
.idx
[0].offset
&& pshader
)
1856 if (ctx
->ps_post_process
&& shader
->u
.ps
.color0_mov
)
1858 shader_addline(buffer
, "#mov handled in srgb write or fog code\n");
1861 shader_hw_map2gl(ins
);
1865 shader_hw_map2gl(ins
);
1869 static void pshader_hw_texkill(const struct wined3d_shader_instruction
*ins
)
1871 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1872 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1875 /* No swizzles are allowed in d3d's texkill. PS 1.x ignores the 4th component as documented,
1876 * but >= 2.0 honors it (undocumented, but tested by the d3d9 testsuite)
1878 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1880 if (ins
->ctx
->reg_maps
->shader_version
.major
>= 2)
1882 const char *kilsrc
= "TA";
1885 shader_arb_get_register_name(ins
, &dst
->reg
, reg_dest
, &is_color
);
1886 if(dst
->write_mask
== WINED3DSP_WRITEMASK_ALL
)
1892 /* Sigh. KIL doesn't support swizzles/writemasks. KIL passes a writemask, but ".xy" for example
1893 * is not valid as a swizzle in ARB (needs ".xyyy"). Use SWZ to load the register properly, and set
1894 * masked out components to 0(won't kill)
1896 char x
= '0', y
= '0', z
= '0', w
= '0';
1897 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
) x
= 'x';
1898 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
) y
= 'y';
1899 if(dst
->write_mask
& WINED3DSP_WRITEMASK_2
) z
= 'z';
1900 if(dst
->write_mask
& WINED3DSP_WRITEMASK_3
) w
= 'w';
1901 shader_addline(buffer
, "SWZ TA, %s, %c, %c, %c, %c;\n", reg_dest
, x
, y
, z
, w
);
1903 shader_addline(buffer
, "KIL %s;\n", kilsrc
);
1907 /* ARB fp doesn't like swizzles on the parameter of the KIL instruction. To mask the 4th component,
1908 * copy the register into our general purpose TMP variable, overwrite .w and pass TMP to KIL
1910 * ps_1_3 shaders use the texcoord incarnation of the Tx register. ps_1_4 shaders can use the same,
1911 * or pass in any temporary register(in shader phase 2)
1913 if (ins
->ctx
->reg_maps
->shader_version
.minor
<= 3)
1914 sprintf(reg_dest
, "fragment.texcoord[%u]", dst
->reg
.idx
[0].offset
);
1916 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1917 shader_addline(buffer
, "SWZ TA, %s, x, y, z, 1;\n", reg_dest
);
1918 shader_addline(buffer
, "KIL TA;\n");
1922 static void pshader_hw_tex(const struct wined3d_shader_instruction
*ins
)
1924 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1925 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1926 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1927 ins
->ctx
->reg_maps
->shader_version
.minor
);
1928 struct wined3d_shader_src_param src
;
1932 DWORD reg_sampler_code
;
1934 BOOL swizzle_coord
= FALSE
;
1936 /* All versions have a destination register */
1937 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1939 /* 1.0-1.4: Use destination register number as texture code.
1940 2.0+: Use provided sampler number as texure code. */
1941 if (shader_version
< WINED3D_SHADER_VERSION(2,0))
1942 reg_sampler_code
= dst
->reg
.idx
[0].offset
;
1944 reg_sampler_code
= ins
->src
[1].reg
.idx
[0].offset
;
1946 /* 1.0-1.3: Use the texcoord varying.
1947 1.4+: Use provided coordinate source register. */
1948 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
1949 sprintf(reg_coord
, "fragment.texcoord[%u]", reg_sampler_code
);
1951 /* TEX is the only instruction that can handle DW and DZ natively */
1953 if(src
.modifiers
== WINED3DSPSM_DW
) src
.modifiers
= WINED3DSPSM_NONE
;
1954 if(src
.modifiers
== WINED3DSPSM_DZ
) src
.modifiers
= WINED3DSPSM_NONE
;
1955 shader_arb_get_src_param(ins
, &src
, 0, reg_coord
);
1959 * 1.1, 1.2, 1.3: Use WINED3D_TSS_TEXTURETRANSFORMFLAGS
1960 * 1.4: Use WINED3DSPSM_DZ or WINED3DSPSM_DW on src[0]
1961 * 2.0+: Use WINED3DSI_TEXLD_PROJECT on the opcode
1963 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
1966 if (reg_sampler_code
< MAX_TEXTURES
)
1967 flags
= priv
->cur_ps_args
->super
.tex_transform
>> reg_sampler_code
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
;
1968 if (flags
& WINED3D_PSARGS_PROJECTED
)
1970 myflags
|= TEX_PROJ
;
1971 if ((flags
& ~WINED3D_PSARGS_PROJECTED
) == WINED3D_TTFF_COUNT3
)
1972 swizzle_coord
= TRUE
;
1975 else if (shader_version
< WINED3D_SHADER_VERSION(2,0))
1977 enum wined3d_shader_src_modifier src_mod
= ins
->src
[0].modifiers
;
1978 if (src_mod
== WINED3DSPSM_DZ
)
1980 swizzle_coord
= TRUE
;
1981 myflags
|= TEX_PROJ
;
1982 } else if(src_mod
== WINED3DSPSM_DW
) {
1983 myflags
|= TEX_PROJ
;
1986 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) myflags
|= TEX_PROJ
;
1987 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) myflags
|= TEX_BIAS
;
1992 /* TXP cannot handle DZ natively, so move the z coordinate to .w.
1993 * reg_coord is a read-only varying register, so we need a temp reg */
1994 shader_addline(ins
->ctx
->buffer
, "SWZ TA, %s, x, y, z, z;\n", reg_coord
);
1995 strcpy(reg_coord
, "TA");
1998 shader_hw_sample(ins
, reg_sampler_code
, reg_dest
, reg_coord
, myflags
, NULL
, NULL
);
2001 static void pshader_hw_texcoord(const struct wined3d_shader_instruction
*ins
)
2003 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2004 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2005 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
2006 ins
->ctx
->reg_maps
->shader_version
.minor
);
2009 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
2011 DWORD reg
= dst
->reg
.idx
[0].offset
;
2013 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2014 shader_addline(buffer
, "MOV_SAT %s, fragment.texcoord[%u];\n", dst_str
, reg
);
2018 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_src
);
2019 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2020 shader_addline(buffer
, "MOV %s, %s;\n", dst_str
, reg_src
);
2024 static void pshader_hw_texreg2ar(const struct wined3d_shader_instruction
*ins
)
2026 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2029 DWORD reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2033 /* Note that texreg2ar treats Tx as a temporary register, not as a varying */
2034 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2035 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2036 /* Move .x first in case src_str is "TA" */
2037 shader_addline(buffer
, "MOV TA.y, %s.x;\n", src_str
);
2038 shader_addline(buffer
, "MOV TA.x, %s.w;\n", src_str
);
2039 if (reg1
< MAX_TEXTURES
)
2041 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2042 flags
= priv
->cur_ps_args
->super
.tex_transform
>> reg1
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
;
2044 shader_hw_sample(ins
, reg1
, dst_str
, "TA", flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2047 static void pshader_hw_texreg2gb(const struct wined3d_shader_instruction
*ins
)
2049 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2051 DWORD reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2055 /* Note that texreg2gb treats Tx as a temporary register, not as a varying */
2056 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2057 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2058 shader_addline(buffer
, "MOV TA.x, %s.y;\n", src_str
);
2059 shader_addline(buffer
, "MOV TA.y, %s.z;\n", src_str
);
2060 shader_hw_sample(ins
, reg1
, dst_str
, "TA", 0, NULL
, NULL
);
2063 static void pshader_hw_texreg2rgb(const struct wined3d_shader_instruction
*ins
)
2065 DWORD reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2069 /* Note that texreg2rg treats Tx as a temporary register, not as a varying */
2070 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2071 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2072 shader_hw_sample(ins
, reg1
, dst_str
, src_str
, 0, NULL
, NULL
);
2075 static void pshader_hw_texbem(const struct wined3d_shader_instruction
*ins
)
2077 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2078 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2079 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2080 char reg_coord
[40], dst_reg
[50], src_reg
[50];
2081 DWORD reg_dest_code
;
2083 /* All versions have a destination register. The Tx where the texture coordinates come
2084 * from is the varying incarnation of the texture register
2086 reg_dest_code
= dst
->reg
.idx
[0].offset
;
2087 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_reg
);
2088 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_reg
);
2089 sprintf(reg_coord
, "fragment.texcoord[%u]", reg_dest_code
);
2091 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed
2092 * The Tx in which the perturbation map is stored is the tempreg incarnation of the texture register
2094 * GL_NV_fragment_program_option could handle this in one instruction via X2D:
2095 * X2D TA.xy, fragment.texcoord, T%u, bumpenvmat%u.xzyw
2097 * However, the NV extensions are never enabled for <= 2.0 shaders because of the performance penalty that
2098 * comes with it, and texbem is an 1.x only instruction. No 1.x instruction forces us to enable the NV
2101 shader_addline(buffer
, "SWZ TB, bumpenvmat%d, x, z, 0, 0;\n", reg_dest_code
);
2102 shader_addline(buffer
, "DP3 TA.x, TB, %s;\n", src_reg
);
2103 shader_addline(buffer
, "SWZ TB, bumpenvmat%d, y, w, 0, 0;\n", reg_dest_code
);
2104 shader_addline(buffer
, "DP3 TA.y, TB, %s;\n", src_reg
);
2106 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2107 * so we can't let the GL handle this.
2109 if ((priv
->cur_ps_args
->super
.tex_transform
>> reg_dest_code
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
)
2110 & WINED3D_PSARGS_PROJECTED
)
2112 shader_addline(buffer
, "RCP TB.w, %s.w;\n", reg_coord
);
2113 shader_addline(buffer
, "MUL TB.xy, %s, TB.w;\n", reg_coord
);
2114 shader_addline(buffer
, "ADD TA.xy, TA, TB;\n");
2116 shader_addline(buffer
, "ADD TA.xy, TA, %s;\n", reg_coord
);
2119 shader_hw_sample(ins
, reg_dest_code
, dst_reg
, "TA", 0, NULL
, NULL
);
2121 if (ins
->handler_idx
== WINED3DSIH_TEXBEML
)
2123 /* No src swizzles are allowed, so this is ok */
2124 shader_addline(buffer
, "MAD TA, %s.z, luminance%d.x, luminance%d.y;\n",
2125 src_reg
, reg_dest_code
, reg_dest_code
);
2126 shader_addline(buffer
, "MUL %s, %s, TA;\n", dst_reg
, dst_reg
);
2130 static void pshader_hw_texm3x2pad(const struct wined3d_shader_instruction
*ins
)
2132 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2133 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2134 char src0_name
[50], dst_name
[50];
2136 struct wined3d_shader_register tmp_reg
= ins
->dst
[0].reg
;
2138 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2139 /* The next instruction will be a texm3x2tex or texm3x2depth that writes to the uninitialized
2140 * T<reg+1> register. Use this register to store the calculated vector
2142 tmp_reg
.idx
[0].offset
= reg
+ 1;
2143 shader_arb_get_register_name(ins
, &tmp_reg
, dst_name
, &is_color
);
2144 shader_addline(buffer
, "DP3 %s.x, fragment.texcoord[%u], %s;\n", dst_name
, reg
, src0_name
);
2147 static void pshader_hw_texm3x2tex(const struct wined3d_shader_instruction
*ins
)
2149 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2151 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2152 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2158 /* We know that we're writing to the uninitialized T<reg> register, so use it for temporary storage */
2159 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2161 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2162 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2163 shader_addline(buffer
, "DP3 %s.y, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2164 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2165 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2168 static void pshader_hw_texm3x3pad(const struct wined3d_shader_instruction
*ins
)
2170 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2171 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2172 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2173 char src0_name
[50], dst_name
[50];
2174 struct wined3d_shader_register tmp_reg
= ins
->dst
[0].reg
;
2177 /* There are always 2 texm3x3pad instructions followed by one texm3x3[tex,vspec, ...] instruction, with
2178 * incrementing ins->dst[0].register_idx numbers. So the pad instruction already knows the final destination
2179 * register, and this register is uninitialized(otherwise the assembler complains that it is 'redeclared')
2181 tmp_reg
.idx
[0].offset
= reg
+ 2 - tex_mx
->current_row
;
2182 shader_arb_get_register_name(ins
, &tmp_reg
, dst_name
, &is_color
);
2184 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2185 shader_addline(buffer
, "DP3 %s.%c, fragment.texcoord[%u], %s;\n",
2186 dst_name
, 'x' + tex_mx
->current_row
, reg
, src0_name
);
2187 tex_mx
->texcoord_w
[tex_mx
->current_row
++] = reg
;
2190 static void pshader_hw_texm3x3tex(const struct wined3d_shader_instruction
*ins
)
2192 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2193 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2195 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2196 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2198 char src0_name
[50], dst_name
[50];
2201 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2202 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2203 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_name
, reg
, src0_name
);
2205 /* Sample the texture using the calculated coordinates */
2206 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2207 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2208 shader_hw_sample(ins
, reg
, dst_str
, dst_name
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2209 tex_mx
->current_row
= 0;
2212 static void pshader_hw_texm3x3vspec(const struct wined3d_shader_instruction
*ins
)
2214 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2215 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2217 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2218 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2224 /* Get the dst reg without writemask strings. We know this register is uninitialized, so we can use all
2225 * components for temporary data storage
2227 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2228 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2229 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2231 /* Construct the eye-ray vector from w coordinates */
2232 shader_addline(buffer
, "MOV TB.x, fragment.texcoord[%u].w;\n", tex_mx
->texcoord_w
[0]);
2233 shader_addline(buffer
, "MOV TB.y, fragment.texcoord[%u].w;\n", tex_mx
->texcoord_w
[1]);
2234 shader_addline(buffer
, "MOV TB.z, fragment.texcoord[%u].w;\n", reg
);
2236 /* Calculate reflection vector
2238 shader_addline(buffer
, "DP3 %s.w, %s, TB;\n", dst_reg
, dst_reg
);
2239 /* The .w is ignored when sampling, so I can use TB.w to calculate dot(N, N) */
2240 shader_addline(buffer
, "DP3 TB.w, %s, %s;\n", dst_reg
, dst_reg
);
2241 shader_addline(buffer
, "RCP TB.w, TB.w;\n");
2242 shader_addline(buffer
, "MUL %s.w, %s.w, TB.w;\n", dst_reg
, dst_reg
);
2243 shader_addline(buffer
, "MUL %s, %s.w, %s;\n", dst_reg
, dst_reg
, dst_reg
);
2244 shader_addline(buffer
, "MAD %s, coefmul.x, %s, -TB;\n", dst_reg
, dst_reg
);
2246 /* Sample the texture using the calculated coordinates */
2247 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2248 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2249 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2250 tex_mx
->current_row
= 0;
2253 static void pshader_hw_texm3x3spec(const struct wined3d_shader_instruction
*ins
)
2255 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2256 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2258 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2259 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2266 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2267 shader_arb_get_src_param(ins
, &ins
->src
[0], 1, src1_name
);
2268 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2269 /* Note: dst_reg.xy is input here, generated by two texm3x3pad instructions */
2270 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2272 /* Calculate reflection vector.
2275 * dst_reg.xyz = 2 * --------- * N - E
2278 * Which normalizes the normal vector
2280 shader_addline(buffer
, "DP3 %s.w, %s, %s;\n", dst_reg
, dst_reg
, src1_name
);
2281 shader_addline(buffer
, "DP3 TC.w, %s, %s;\n", dst_reg
, dst_reg
);
2282 shader_addline(buffer
, "RCP TC.w, TC.w;\n");
2283 shader_addline(buffer
, "MUL %s.w, %s.w, TC.w;\n", dst_reg
, dst_reg
);
2284 shader_addline(buffer
, "MUL %s, %s.w, %s;\n", dst_reg
, dst_reg
, dst_reg
);
2285 shader_addline(buffer
, "MAD %s, coefmul.x, %s, -%s;\n", dst_reg
, dst_reg
, src1_name
);
2287 /* Sample the texture using the calculated coordinates */
2288 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2289 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2290 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2291 tex_mx
->current_row
= 0;
2294 static void pshader_hw_texdepth(const struct wined3d_shader_instruction
*ins
)
2296 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2297 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2299 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2300 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2302 /* texdepth has an implicit destination, the fragment depth value. It's only parameter,
2303 * which is essentially an input, is the destination register because it is the first
2304 * parameter. According to the msdn, this must be register r5, but let's keep it more flexible
2305 * here(writemasks/swizzles are not valid on texdepth)
2307 shader_arb_get_dst_param(ins
, dst
, dst_name
);
2309 /* According to the msdn, the source register(must be r5) is unusable after
2310 * the texdepth instruction, so we're free to modify it
2312 shader_addline(buffer
, "MIN %s.y, %s.y, %s;\n", dst_name
, dst_name
, one
);
2314 /* How to deal with the special case dst_name.g == 0? if r != 0, then
2315 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
2316 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
2318 shader_addline(buffer
, "RCP %s.y, %s.y;\n", dst_name
, dst_name
);
2319 shader_addline(buffer
, "MUL TA.x, %s.x, %s.y;\n", dst_name
, dst_name
);
2320 shader_addline(buffer
, "MIN TA.x, TA.x, %s;\n", one
);
2321 shader_addline(buffer
, "MAX result.depth, TA.x, %s;\n", zero
);
2324 /** Process the WINED3DSIO_TEXDP3TEX instruction in ARB:
2325 * Take a 3-component dot product of the TexCoord[dstreg] and src,
2326 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
2327 static void pshader_hw_texdp3tex(const struct wined3d_shader_instruction
*ins
)
2329 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2330 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
[0].offset
;
2334 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2335 shader_addline(buffer
, "MOV TB, 0.0;\n");
2336 shader_addline(buffer
, "DP3 TB.x, fragment.texcoord[%u], %s;\n", sampler_idx
, src0
);
2338 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2339 shader_hw_sample(ins
, sampler_idx
, dst_str
, "TB", 0 /* Only one coord, can't be projected */, NULL
, NULL
);
2342 /** Process the WINED3DSIO_TEXDP3 instruction in ARB:
2343 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
2344 static void pshader_hw_texdp3(const struct wined3d_shader_instruction
*ins
)
2346 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2349 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2351 /* Handle output register */
2352 shader_arb_get_dst_param(ins
, dst
, dst_str
);
2353 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2354 shader_addline(buffer
, "DP3 %s, fragment.texcoord[%u], %s;\n", dst_str
, dst
->reg
.idx
[0].offset
, src0
);
2357 /** Process the WINED3DSIO_TEXM3X3 instruction in ARB
2358 * Perform the 3rd row of a 3x3 matrix multiply */
2359 static void pshader_hw_texm3x3(const struct wined3d_shader_instruction
*ins
)
2361 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2362 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2363 char dst_str
[50], dst_name
[50];
2367 shader_arb_get_dst_param(ins
, dst
, dst_str
);
2368 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2369 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2370 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_name
, dst
->reg
.idx
[0].offset
, src0
);
2371 shader_addline(buffer
, "MOV %s, %s;\n", dst_str
, dst_name
);
2374 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in ARB:
2375 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
2376 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
2377 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
2379 static void pshader_hw_texm3x2depth(const struct wined3d_shader_instruction
*ins
)
2381 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2382 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2383 char src0
[50], dst_name
[50];
2385 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2386 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2388 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2389 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2390 shader_addline(buffer
, "DP3 %s.y, fragment.texcoord[%u], %s;\n", dst_name
, dst
->reg
.idx
[0].offset
, src0
);
2392 /* How to deal with the special case dst_name.g == 0? if r != 0, then
2393 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
2394 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
2396 shader_addline(buffer
, "RCP %s.y, %s.y;\n", dst_name
, dst_name
);
2397 shader_addline(buffer
, "MUL %s.x, %s.x, %s.y;\n", dst_name
, dst_name
, dst_name
);
2398 shader_addline(buffer
, "MIN %s.x, %s.x, %s;\n", dst_name
, dst_name
, one
);
2399 shader_addline(buffer
, "MAX result.depth, %s.x, %s;\n", dst_name
, zero
);
2402 /** Handles transforming all WINED3DSIO_M?x? opcodes for
2403 Vertex/Pixel shaders to ARB_vertex_program codes */
2404 static void shader_hw_mnxn(const struct wined3d_shader_instruction
*ins
)
2407 int nComponents
= 0;
2408 struct wined3d_shader_dst_param tmp_dst
= {{0}};
2409 struct wined3d_shader_src_param tmp_src
[2] = {{{0}}};
2410 struct wined3d_shader_instruction tmp_ins
;
2412 memset(&tmp_ins
, 0, sizeof(tmp_ins
));
2414 /* Set constants for the temporary argument */
2415 tmp_ins
.ctx
= ins
->ctx
;
2416 tmp_ins
.dst_count
= 1;
2417 tmp_ins
.dst
= &tmp_dst
;
2418 tmp_ins
.src_count
= 2;
2419 tmp_ins
.src
= tmp_src
;
2421 switch(ins
->handler_idx
)
2423 case WINED3DSIH_M4x4
:
2425 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2427 case WINED3DSIH_M4x3
:
2429 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2431 case WINED3DSIH_M3x4
:
2433 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2435 case WINED3DSIH_M3x3
:
2437 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2439 case WINED3DSIH_M3x2
:
2441 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2444 FIXME("Unhandled opcode %#x\n", ins
->handler_idx
);
2448 tmp_dst
= ins
->dst
[0];
2449 tmp_src
[0] = ins
->src
[0];
2450 tmp_src
[1] = ins
->src
[1];
2451 for (i
= 0; i
< nComponents
; ++i
)
2453 tmp_dst
.write_mask
= WINED3DSP_WRITEMASK_0
<< i
;
2454 shader_hw_map2gl(&tmp_ins
);
2455 ++tmp_src
[1].reg
.idx
[0].offset
;
2459 static DWORD
abs_modifier(DWORD mod
, BOOL
*need_abs
)
2465 case WINED3DSPSM_NONE
: return WINED3DSPSM_ABS
;
2466 case WINED3DSPSM_NEG
: return WINED3DSPSM_ABS
;
2467 case WINED3DSPSM_BIAS
: *need_abs
= TRUE
; return WINED3DSPSM_BIAS
;
2468 case WINED3DSPSM_BIASNEG
: *need_abs
= TRUE
; return WINED3DSPSM_BIASNEG
;
2469 case WINED3DSPSM_SIGN
: *need_abs
= TRUE
; return WINED3DSPSM_SIGN
;
2470 case WINED3DSPSM_SIGNNEG
: *need_abs
= TRUE
; return WINED3DSPSM_SIGNNEG
;
2471 case WINED3DSPSM_COMP
: *need_abs
= TRUE
; return WINED3DSPSM_COMP
;
2472 case WINED3DSPSM_X2
: *need_abs
= TRUE
; return WINED3DSPSM_X2
;
2473 case WINED3DSPSM_X2NEG
: *need_abs
= TRUE
; return WINED3DSPSM_X2NEG
;
2474 case WINED3DSPSM_DZ
: *need_abs
= TRUE
; return WINED3DSPSM_DZ
;
2475 case WINED3DSPSM_DW
: *need_abs
= TRUE
; return WINED3DSPSM_DW
;
2476 case WINED3DSPSM_ABS
: return WINED3DSPSM_ABS
;
2477 case WINED3DSPSM_ABSNEG
: return WINED3DSPSM_ABS
;
2479 FIXME("Unknown modifier %u\n", mod
);
2483 static void shader_hw_scalar_op(const struct wined3d_shader_instruction
*ins
)
2485 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2486 const char *instruction
;
2487 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
2488 BOOL need_abs
= FALSE
;
2493 switch(ins
->handler_idx
)
2495 case WINED3DSIH_RSQ
: instruction
= "RSQ"; break;
2496 case WINED3DSIH_RCP
: instruction
= "RCP"; break;
2497 case WINED3DSIH_EXPP
:
2498 if (ins
->ctx
->reg_maps
->shader_version
.major
< 2)
2500 instruction
= "EXP";
2504 case WINED3DSIH_EXP
:
2505 instruction
= "EX2";
2507 case WINED3DSIH_LOG
:
2508 case WINED3DSIH_LOGP
:
2509 /* The precision requirements suggest that LOGP matches ARBvp's LOG
2510 * instruction, but notice that the output of those instructions is
2512 src0_copy
.modifiers
= abs_modifier(src0_copy
.modifiers
, &need_abs
);
2513 instruction
= "LG2";
2515 default: instruction
= "";
2516 FIXME("Unhandled opcode %#x\n", ins
->handler_idx
);
2520 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
2522 src0_copy
.swizzle
= shader_arb_select_component(src0_copy
.swizzle
, 3);
2524 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
); /* Destination */
2525 shader_arb_get_src_param(ins
, &src0_copy
, 0, src
);
2529 shader_addline(buffer
, "ABS TA.w, %s;\n", src
);
2530 shader_addline(buffer
, "%s%s %s, TA.w;\n", instruction
, shader_arb_get_modifier(ins
), dst
);
2534 shader_addline(buffer
, "%s%s %s, %s;\n", instruction
, shader_arb_get_modifier(ins
), dst
, src
);
2539 static void shader_hw_nrm(const struct wined3d_shader_instruction
*ins
)
2541 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2544 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2545 BOOL pshader
= shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2546 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2548 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2549 shader_arb_get_src_param(ins
, &ins
->src
[0], 1 /* Use TB */, src_name
);
2551 /* In D3D, NRM of a vector with length zero returns zero. Catch this situation, as
2552 * otherwise NRM or RSQ would return NaN */
2553 if(pshader
&& priv
->target_version
>= NV3
)
2555 /* GL_NV_fragment_program2's NRM needs protection against length zero vectors too
2557 * TODO: Find out if DP3+NRM+MOV is really faster than DP3+RSQ+MUL
2559 shader_addline(buffer
, "DP3C TA, %s, %s;\n", src_name
, src_name
);
2560 shader_addline(buffer
, "NRM%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
);
2561 shader_addline(buffer
, "MOV %s (EQ), %s;\n", dst_name
, zero
);
2563 else if(priv
->target_version
>= NV2
)
2565 shader_addline(buffer
, "DP3C TA.x, %s, %s;\n", src_name
, src_name
);
2566 shader_addline(buffer
, "RSQ TA.x (NE), TA.x;\n");
2567 shader_addline(buffer
, "MUL%s %s, %s, TA.x;\n", shader_arb_get_modifier(ins
), dst_name
,
2572 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2574 shader_addline(buffer
, "DP3 TA.x, %s, %s;\n", src_name
, src_name
);
2575 /* Pass any non-zero value to RSQ if the input vector has a length of zero. The
2576 * RSQ result doesn't matter, as long as multiplying it by 0 returns 0.
2578 shader_addline(buffer
, "SGE TA.y, -TA.x, %s;\n", zero
);
2579 shader_addline(buffer
, "MAD TA.x, %s, TA.y, TA.x;\n", one
);
2581 shader_addline(buffer
, "RSQ TA.x, TA.x;\n");
2582 /* dst.w = src[0].w * 1 / (src.x^2 + src.y^2 + src.z^2)^(1/2) according to msdn*/
2583 shader_addline(buffer
, "MUL%s %s, %s, TA.x;\n", shader_arb_get_modifier(ins
), dst_name
,
2588 static void shader_hw_lrp(const struct wined3d_shader_instruction
*ins
)
2590 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2592 char src_name
[3][50];
2594 /* ARB_fragment_program has a convenient LRP instruction */
2595 if(shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)) {
2596 shader_hw_map2gl(ins
);
2600 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2601 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
2602 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
2603 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
2605 shader_addline(buffer
, "SUB TA, %s, %s;\n", src_name
[1], src_name
[2]);
2606 shader_addline(buffer
, "MAD%s %s, %s, TA, %s;\n", shader_arb_get_modifier(ins
),
2607 dst_name
, src_name
[0], src_name
[2]);
2610 static void shader_hw_sincos(const struct wined3d_shader_instruction
*ins
)
2612 /* This instruction exists in ARB, but the d3d instruction takes two extra parameters which
2613 * must contain fixed constants. So we need a separate function to filter those constants and
2616 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2617 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2618 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2620 char src_name0
[50], src_name1
[50], src_name2
[50];
2623 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
2624 if(shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)) {
2625 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2626 /* No modifiers are supported on SCS */
2627 shader_addline(buffer
, "SCS %s, %s;\n", dst_name
, src_name0
);
2629 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
)
2631 shader_arb_get_register_name(ins
, &dst
->reg
, src_name0
, &is_color
);
2632 shader_addline(buffer
, "MOV_SAT %s, %s;\n", dst_name
, src_name0
);
2634 } else if(priv
->target_version
>= NV2
) {
2635 shader_arb_get_register_name(ins
, &dst
->reg
, dst_name
, &is_color
);
2637 /* Sincos writemask must be .x, .y or .xy */
2638 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
)
2639 shader_addline(buffer
, "COS%s %s.x, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name0
);
2640 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
)
2641 shader_addline(buffer
, "SIN%s %s.y, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name0
);
2643 /* Approximate sine and cosine with a taylor series, as per math textbook. The application passes 8
2644 * helper constants(D3DSINCOSCONST1 and D3DSINCOSCONST2) in src1 and src2.
2646 * sin(x) = x - x^3/3! + x^5/5! - x^7/7! + ...
2647 * cos(x) = 1 - x^2/2! + x^4/4! - x^6/6! + ...
2649 * The constants we get are:
2651 * +1 +1, -1 -1 +1 +1 -1 -1
2652 * ---- , ---- , ---- , ----- , ----- , ----- , ------
2653 * 1!*2 2!*4 3!*8 4!*16 5!*32 6!*64 7!*128
2655 * If used with x^2, x^3, x^4 etc they calculate sin(x/2) and cos(x/2):
2659 * (x/2)^4 = x^4 / 16
2660 * (x/2)^5 = x^5 / 32
2663 * To get the final result:
2664 * sin(x) = 2 * sin(x/2) * cos(x/2)
2665 * cos(x) = cos(x/2)^2 - sin(x/2)^2
2666 * (from sin(x+y) and cos(x+y) rules)
2668 * As per MSDN, dst.z is undefined after the operation, and so is
2669 * dst.x and dst.y if they're masked out by the writemask. Ie
2670 * sincos dst.y, src1, c0, c1
2671 * returns the sine in dst.y. dst.x and dst.z are undefined, dst.w is not touched. The assembler
2672 * vsa.exe also stops with an error if the dest register is the same register as the source
2673 * register. This means we can use dest.xyz as temporary storage. The assembler vsa.exe output also
2674 * indicates that sincos consumes 8 instruction slots in vs_2_0(and, strangely, in vs_3_0).
2676 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
2677 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name2
);
2678 shader_arb_get_register_name(ins
, &dst
->reg
, dst_name
, &is_color
);
2680 shader_addline(buffer
, "MUL %s.x, %s, %s;\n", dst_name
, src_name0
, src_name0
); /* x ^ 2 */
2681 shader_addline(buffer
, "MUL TA.y, %s.x, %s;\n", dst_name
, src_name0
); /* x ^ 3 */
2682 shader_addline(buffer
, "MUL %s.y, TA.y, %s;\n", dst_name
, src_name0
); /* x ^ 4 */
2683 shader_addline(buffer
, "MUL TA.z, %s.y, %s;\n", dst_name
, src_name0
); /* x ^ 5 */
2684 shader_addline(buffer
, "MUL %s.z, TA.z, %s;\n", dst_name
, src_name0
); /* x ^ 6 */
2685 shader_addline(buffer
, "MUL TA.w, %s.z, %s;\n", dst_name
, src_name0
); /* x ^ 7 */
2689 * Unfortunately we don't get the constants in a DP4-capable form. Is there a way to
2690 * properly merge that with MULs in the code above?
2691 * The swizzles .yz and xw however fit into the .yzxw swizzle added to ps_2_0. Maybe
2692 * we can merge the sine and cosine MAD rows to calculate them together.
2694 shader_addline(buffer
, "MUL TA.x, %s, %s.w;\n", src_name0
, src_name2
); /* x^1, +1/(1!*2) */
2695 shader_addline(buffer
, "MAD TA.x, TA.y, %s.x, TA.x;\n", src_name2
); /* -1/(3!*8) */
2696 shader_addline(buffer
, "MAD TA.x, TA.z, %s.w, TA.x;\n", src_name1
); /* +1/(5!*32) */
2697 shader_addline(buffer
, "MAD TA.x, TA.w, %s.x, TA.x;\n", src_name1
); /* -1/(7!*128) */
2700 shader_addline(buffer
, "MAD TA.y, %s.x, %s.y, %s.z;\n", dst_name
, src_name2
, src_name2
); /* -1/(2!*4), +1.0 */
2701 shader_addline(buffer
, "MAD TA.y, %s.y, %s.z, TA.y;\n", dst_name
, src_name1
); /* +1/(4!*16) */
2702 shader_addline(buffer
, "MAD TA.y, %s.z, %s.y, TA.y;\n", dst_name
, src_name1
); /* -1/(6!*64) */
2704 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
) {
2706 shader_addline(buffer
, "MUL TA.z, TA.y, TA.y;\n");
2707 shader_addline(buffer
, "MAD %s.x, -TA.x, TA.x, TA.z;\n", dst_name
);
2709 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
) {
2711 shader_addline(buffer
, "MUL %s.y, TA.x, TA.y;\n", dst_name
);
2712 shader_addline(buffer
, "ADD %s.y, %s.y, %s.y;\n", dst_name
, dst_name
, dst_name
);
2717 static void shader_hw_sgn(const struct wined3d_shader_instruction
*ins
)
2719 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2722 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
2724 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2725 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
);
2727 /* SGN is only valid in vertex shaders */
2728 if(ctx
->target_version
>= NV2
) {
2729 shader_addline(buffer
, "SSG%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
);
2733 /* If SRC > 0.0, -SRC < SRC = TRUE, otherwise false.
2734 * if SRC < 0.0, SRC < -SRC = TRUE. If neither is true, src = 0.0
2736 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
) {
2737 shader_addline(buffer
, "SLT %s, -%s, %s;\n", dst_name
, src_name
, src_name
);
2739 /* src contains TA? Write to the dest first. This won't overwrite our destination.
2740 * Then use TA, and calculate the final result
2742 * Not reading from TA? Store the first result in TA to avoid overwriting the
2743 * destination if src reg = dst reg
2745 if(strstr(src_name
, "TA"))
2747 shader_addline(buffer
, "SLT %s, %s, -%s;\n", dst_name
, src_name
, src_name
);
2748 shader_addline(buffer
, "SLT TA, -%s, %s;\n", src_name
, src_name
);
2749 shader_addline(buffer
, "ADD %s, %s, -TA;\n", dst_name
, dst_name
);
2753 shader_addline(buffer
, "SLT TA, -%s, %s;\n", src_name
, src_name
);
2754 shader_addline(buffer
, "SLT %s, %s, -%s;\n", dst_name
, src_name
, src_name
);
2755 shader_addline(buffer
, "ADD %s, TA, -%s;\n", dst_name
, dst_name
);
2760 static void shader_hw_dsy(const struct wined3d_shader_instruction
*ins
)
2762 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2768 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
);
2769 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src
);
2770 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2772 shader_addline(buffer
, "DDY %s, %s;\n", dst
, src
);
2773 shader_addline(buffer
, "MUL%s %s, %s, ycorrection.y;\n", shader_arb_get_modifier(ins
), dst
, dst_name
);
2776 static void shader_hw_pow(const struct wined3d_shader_instruction
*ins
)
2778 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2779 char src0
[50], src1
[50], dst
[50];
2780 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
2781 BOOL need_abs
= FALSE
;
2782 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2783 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2785 /* POW operates on the absolute value of the input */
2786 src0_copy
.modifiers
= abs_modifier(src0_copy
.modifiers
, &need_abs
);
2788 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
);
2789 shader_arb_get_src_param(ins
, &src0_copy
, 0, src0
);
2790 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src1
);
2793 shader_addline(buffer
, "ABS TA.x, %s;\n", src0
);
2795 shader_addline(buffer
, "MOV TA.x, %s;\n", src0
);
2797 if (priv
->target_version
>= NV2
)
2799 shader_addline(buffer
, "MOVC TA.y, %s;\n", src1
);
2800 shader_addline(buffer
, "POW%s %s, TA.x, TA.y;\n", shader_arb_get_modifier(ins
), dst
);
2801 shader_addline(buffer
, "MOV %s (EQ.y), %s;\n", dst
, one
);
2805 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2806 const char *flt_eps
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_EPS
);
2808 shader_addline(buffer
, "ABS TA.y, %s;\n", src1
);
2809 shader_addline(buffer
, "SGE TA.y, -TA.y, %s;\n", zero
);
2810 /* Possibly add flt_eps to avoid getting float special values */
2811 shader_addline(buffer
, "MAD TA.z, TA.y, %s, %s;\n", flt_eps
, src1
);
2812 shader_addline(buffer
, "POW%s TA.x, TA.x, TA.z;\n", shader_arb_get_modifier(ins
));
2813 shader_addline(buffer
, "MAD TA.x, -TA.x, TA.y, TA.x;\n");
2814 shader_addline(buffer
, "MAD %s, TA.y, %s, TA.x;\n", dst
, one
);
2818 static void shader_hw_loop(const struct wined3d_shader_instruction
*ins
)
2820 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2822 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2825 shader_arb_get_src_param(ins
, &ins
->src
[1], 0, src_name
);
2829 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2830 struct list
*e
= list_head(&priv
->control_frames
);
2831 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2833 if(priv
->loop_depth
> 1) shader_addline(buffer
, "PUSHA aL;\n");
2834 /* The constant loader makes sure to load -1 into iX.w */
2835 shader_addline(buffer
, "ARLC aL, %s.xywz;\n", src_name
);
2836 shader_addline(buffer
, "BRA loop_%u_end (LE.x);\n", control_frame
->no
.loop
);
2837 shader_addline(buffer
, "loop_%u_start:\n", control_frame
->no
.loop
);
2841 shader_addline(buffer
, "LOOP %s;\n", src_name
);
2845 static void shader_hw_rep(const struct wined3d_shader_instruction
*ins
)
2847 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2849 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2851 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
);
2853 /* The constant loader makes sure to load -1 into iX.w */
2856 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2857 struct list
*e
= list_head(&priv
->control_frames
);
2858 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2860 if(priv
->loop_depth
> 1) shader_addline(buffer
, "PUSHA aL;\n");
2862 shader_addline(buffer
, "ARLC aL, %s.xywz;\n", src_name
);
2863 shader_addline(buffer
, "BRA loop_%u_end (LE.x);\n", control_frame
->no
.loop
);
2864 shader_addline(buffer
, "loop_%u_start:\n", control_frame
->no
.loop
);
2868 shader_addline(buffer
, "REP %s;\n", src_name
);
2872 static void shader_hw_endloop(const struct wined3d_shader_instruction
*ins
)
2874 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2875 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2879 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2880 struct list
*e
= list_head(&priv
->control_frames
);
2881 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2883 shader_addline(buffer
, "ARAC aL.xy, aL;\n");
2884 shader_addline(buffer
, "BRA loop_%u_start (GT.x);\n", control_frame
->no
.loop
);
2885 shader_addline(buffer
, "loop_%u_end:\n", control_frame
->no
.loop
);
2887 if(priv
->loop_depth
> 1) shader_addline(buffer
, "POPA aL;\n");
2891 shader_addline(buffer
, "ENDLOOP;\n");
2895 static void shader_hw_endrep(const struct wined3d_shader_instruction
*ins
)
2897 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2898 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2902 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2903 struct list
*e
= list_head(&priv
->control_frames
);
2904 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2906 shader_addline(buffer
, "ARAC aL.xy, aL;\n");
2907 shader_addline(buffer
, "BRA loop_%u_start (GT.x);\n", control_frame
->no
.loop
);
2908 shader_addline(buffer
, "loop_%u_end:\n", control_frame
->no
.loop
);
2910 if(priv
->loop_depth
> 1) shader_addline(buffer
, "POPA aL;\n");
2914 shader_addline(buffer
, "ENDREP;\n");
2918 static const struct control_frame
*find_last_loop(const struct shader_arb_ctx_priv
*priv
)
2920 struct control_frame
*control_frame
;
2922 LIST_FOR_EACH_ENTRY(control_frame
, &priv
->control_frames
, struct control_frame
, entry
)
2924 if(control_frame
->type
== LOOP
|| control_frame
->type
== REP
) return control_frame
;
2926 ERR("Could not find loop for break\n");
2930 static void shader_hw_break(const struct wined3d_shader_instruction
*ins
)
2932 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2933 const struct control_frame
*control_frame
= find_last_loop(ins
->ctx
->backend_data
);
2934 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2938 shader_addline(buffer
, "BRA loop_%u_end;\n", control_frame
->no
.loop
);
2942 shader_addline(buffer
, "BRK;\n");
2946 static const char *get_compare(enum wined3d_shader_rel_op op
)
2950 case WINED3D_SHADER_REL_OP_GT
: return "GT";
2951 case WINED3D_SHADER_REL_OP_EQ
: return "EQ";
2952 case WINED3D_SHADER_REL_OP_GE
: return "GE";
2953 case WINED3D_SHADER_REL_OP_LT
: return "LT";
2954 case WINED3D_SHADER_REL_OP_NE
: return "NE";
2955 case WINED3D_SHADER_REL_OP_LE
: return "LE";
2957 FIXME("Unrecognized operator %#x.\n", op
);
2962 static enum wined3d_shader_rel_op
invert_compare(enum wined3d_shader_rel_op op
)
2966 case WINED3D_SHADER_REL_OP_GT
: return WINED3D_SHADER_REL_OP_LE
;
2967 case WINED3D_SHADER_REL_OP_EQ
: return WINED3D_SHADER_REL_OP_NE
;
2968 case WINED3D_SHADER_REL_OP_GE
: return WINED3D_SHADER_REL_OP_LT
;
2969 case WINED3D_SHADER_REL_OP_LT
: return WINED3D_SHADER_REL_OP_GE
;
2970 case WINED3D_SHADER_REL_OP_NE
: return WINED3D_SHADER_REL_OP_EQ
;
2971 case WINED3D_SHADER_REL_OP_LE
: return WINED3D_SHADER_REL_OP_GT
;
2973 FIXME("Unrecognized operator %#x.\n", op
);
2978 static void shader_hw_breakc(const struct wined3d_shader_instruction
*ins
)
2980 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2981 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2982 const struct control_frame
*control_frame
= find_last_loop(ins
->ctx
->backend_data
);
2985 const char *comp
= get_compare(ins
->flags
);
2987 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
2988 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
2992 /* SUBC CC, src0, src1" works only in pixel shaders, so use TA to throw
2993 * away the subtraction result
2995 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
2996 shader_addline(buffer
, "BRA loop_%u_end (%s.x);\n", control_frame
->no
.loop
, comp
);
3000 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3001 shader_addline(buffer
, "BRK (%s.x);\n", comp
);
3005 static void shader_hw_ifc(const struct wined3d_shader_instruction
*ins
)
3007 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3008 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3009 struct list
*e
= list_head(&priv
->control_frames
);
3010 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3014 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3016 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
3017 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
3021 /* Invert the flag. We jump to the else label if the condition is NOT true */
3022 comp
= get_compare(invert_compare(ins
->flags
));
3023 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3024 shader_addline(buffer
, "BRA ifc_%u_else (%s.x);\n", control_frame
->no
.ifc
, comp
);
3028 comp
= get_compare(ins
->flags
);
3029 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3030 shader_addline(buffer
, "IF %s.x;\n", comp
);
3034 static void shader_hw_else(const struct wined3d_shader_instruction
*ins
)
3036 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3037 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3038 struct list
*e
= list_head(&priv
->control_frames
);
3039 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3040 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3044 shader_addline(buffer
, "BRA ifc_%u_endif;\n", control_frame
->no
.ifc
);
3045 shader_addline(buffer
, "ifc_%u_else:\n", control_frame
->no
.ifc
);
3046 control_frame
->had_else
= TRUE
;
3050 shader_addline(buffer
, "ELSE;\n");
3054 static void shader_hw_endif(const struct wined3d_shader_instruction
*ins
)
3056 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3057 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3058 struct list
*e
= list_head(&priv
->control_frames
);
3059 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3060 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3064 if(control_frame
->had_else
)
3066 shader_addline(buffer
, "ifc_%u_endif:\n", control_frame
->no
.ifc
);
3070 shader_addline(buffer
, "#No else branch. else is endif\n");
3071 shader_addline(buffer
, "ifc_%u_else:\n", control_frame
->no
.ifc
);
3076 shader_addline(buffer
, "ENDIF;\n");
3080 static void shader_hw_texldd(const struct wined3d_shader_instruction
*ins
)
3082 DWORD sampler_idx
= ins
->src
[1].reg
.idx
[0].offset
;
3084 char reg_src
[3][40];
3085 WORD flags
= TEX_DERIV
;
3087 shader_arb_get_dst_param(ins
, &ins
->dst
[0], reg_dest
);
3088 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_src
[0]);
3089 shader_arb_get_src_param(ins
, &ins
->src
[2], 1, reg_src
[1]);
3090 shader_arb_get_src_param(ins
, &ins
->src
[3], 2, reg_src
[2]);
3092 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) flags
|= TEX_PROJ
;
3093 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) flags
|= TEX_BIAS
;
3095 shader_hw_sample(ins
, sampler_idx
, reg_dest
, reg_src
[0], flags
, reg_src
[1], reg_src
[2]);
3098 static void shader_hw_texldl(const struct wined3d_shader_instruction
*ins
)
3100 DWORD sampler_idx
= ins
->src
[1].reg
.idx
[0].offset
;
3103 WORD flags
= TEX_LOD
;
3105 shader_arb_get_dst_param(ins
, &ins
->dst
[0], reg_dest
);
3106 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_coord
);
3108 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) flags
|= TEX_PROJ
;
3109 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) flags
|= TEX_BIAS
;
3111 shader_hw_sample(ins
, sampler_idx
, reg_dest
, reg_coord
, flags
, NULL
, NULL
);
3114 static void shader_hw_label(const struct wined3d_shader_instruction
*ins
)
3116 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3117 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3119 priv
->in_main_func
= FALSE
;
3120 /* Call instructions activate the NV extensions, not labels and rets. If there is an uncalled
3121 * subroutine, don't generate a label that will make GL complain
3123 if(priv
->target_version
== ARB
) return;
3125 shader_addline(buffer
, "l%u:\n", ins
->src
[0].reg
.idx
[0].offset
);
3128 static void vshader_add_footer(struct shader_arb_ctx_priv
*priv_ctx
,
3129 const struct arb_vshader_private
*shader_data
, const struct arb_vs_compile_args
*args
,
3130 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
,
3131 struct wined3d_string_buffer
*buffer
)
3135 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
3136 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
3137 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
3138 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
3140 if (args
->super
.fog_src
== VS_FOG_Z
)
3142 shader_addline(buffer
, "MOV result.fogcoord, TMP_OUT.z;\n");
3148 /* posFixup.x is always 1.0, so we can safely use it */
3149 shader_addline(buffer
, "ADD result.fogcoord, posFixup.x, -posFixup.x;\n");
3153 /* Clamp fogcoord */
3154 const char *zero
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ZERO
);
3155 const char *one
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ONE
);
3157 shader_addline(buffer
, "MIN TMP_FOGCOORD.x, TMP_FOGCOORD.x, %s;\n", one
);
3158 shader_addline(buffer
, "MAX result.fogcoord.x, TMP_FOGCOORD.x, %s;\n", zero
);
3162 /* Clipplanes are always stored without y inversion */
3163 if (use_nv_clip(gl_info
) && priv_ctx
->target_version
>= NV2
)
3165 if (args
->super
.clip_enabled
)
3167 for (i
= 0; i
< priv_ctx
->vs_clipplanes
; i
++)
3169 shader_addline(buffer
, "DP4 result.clip[%u].x, TMP_OUT, state.clip[%u].plane;\n", i
, i
);
3173 else if (args
->clip
.boolclip
.clip_texcoord
)
3175 unsigned int cur_clip
= 0;
3176 char component
[4] = {'x', 'y', 'z', 'w'};
3177 const char *zero
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_ZERO
);
3179 for (i
= 0; i
< gl_info
->limits
.clipplanes
; ++i
)
3181 if (args
->clip
.boolclip
.clipplane_mask
& (1 << i
))
3183 shader_addline(buffer
, "DP4 TA.%c, TMP_OUT, state.clip[%u].plane;\n",
3184 component
[cur_clip
++], i
);
3190 shader_addline(buffer
, "MOV TA, %s;\n", zero
);
3193 shader_addline(buffer
, "MOV TA.yzw, %s;\n", zero
);
3196 shader_addline(buffer
, "MOV TA.zw, %s;\n", zero
);
3199 shader_addline(buffer
, "MOV TA.w, %s;\n", zero
);
3202 shader_addline(buffer
, "MOV result.texcoord[%u], TA;\n",
3203 args
->clip
.boolclip
.clip_texcoord
- 1);
3206 /* Write the final position.
3208 * OpenGL coordinates specify the center of the pixel while d3d coords specify
3209 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
3210 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
3211 * contains 1.0 to allow a mad, but arb vs swizzles are too restricted for that.
3213 shader_addline(buffer
, "MUL TA, posFixup, TMP_OUT.w;\n");
3214 shader_addline(buffer
, "ADD TMP_OUT.x, TMP_OUT.x, TA.z;\n");
3215 shader_addline(buffer
, "MAD TMP_OUT.y, TMP_OUT.y, posFixup.y, TA.w;\n");
3217 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
3218 * and the glsl equivalent
3220 if (need_helper_const(shader_data
, reg_maps
, gl_info
))
3222 const char *two
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_TWO
);
3223 shader_addline(buffer
, "MAD TMP_OUT.z, TMP_OUT.z, %s, -TMP_OUT.w;\n", two
);
3227 shader_addline(buffer
, "ADD TMP_OUT.z, TMP_OUT.z, TMP_OUT.z;\n");
3228 shader_addline(buffer
, "ADD TMP_OUT.z, TMP_OUT.z, -TMP_OUT.w;\n");
3231 shader_addline(buffer
, "MOV result.position, TMP_OUT;\n");
3233 priv_ctx
->footer_written
= TRUE
;
3236 static void shader_hw_ret(const struct wined3d_shader_instruction
*ins
)
3238 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3239 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3240 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
3241 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3243 if(priv
->target_version
== ARB
) return;
3247 if (priv
->in_main_func
) vshader_add_footer(priv
, shader
->backend_data
,
3248 priv
->cur_vs_args
, ins
->ctx
->reg_maps
, ins
->ctx
->gl_info
, buffer
);
3251 shader_addline(buffer
, "RET;\n");
3254 static void shader_hw_call(const struct wined3d_shader_instruction
*ins
)
3256 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3257 shader_addline(buffer
, "CAL l%u;\n", ins
->src
[0].reg
.idx
[0].offset
);
3260 static BOOL
shader_arb_compile(const struct wined3d_gl_info
*gl_info
, GLenum target
, const char *src
)
3262 const char *ptr
, *line
;
3265 if (TRACE_ON(d3d_shader
))
3268 while ((line
= get_line(&ptr
))) TRACE_(d3d_shader
)(" %.*s", (int)(ptr
- line
), line
);
3271 GL_EXTCALL(glProgramStringARB(target
, GL_PROGRAM_FORMAT_ASCII_ARB
, strlen(src
), src
));
3272 checkGLcall("glProgramStringARB()");
3274 if (FIXME_ON(d3d_shader
))
3276 gl_info
->gl_ops
.gl
.p_glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
3279 FIXME_(d3d_shader
)("Program error at position %d: %s\n\n", pos
,
3280 debugstr_a((const char *)gl_info
->gl_ops
.gl
.p_glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
3282 while ((line
= get_line(&ptr
))) FIXME_(d3d_shader
)(" %.*s", (int)(ptr
- line
), line
);
3283 FIXME_(d3d_shader
)("\n");
3289 if (WARN_ON(d3d_perf
))
3291 GL_EXTCALL(glGetProgramivARB(target
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
3292 checkGLcall("glGetProgramivARB()");
3294 WARN_(d3d_perf
)("Program exceeds native resource limits.\n");
3300 /* Context activation is done by the caller. */
3301 static GLuint
create_arb_blt_vertex_program(const struct wined3d_gl_info
*gl_info
)
3303 GLuint program_id
= 0;
3305 static const char blt_vprogram
[] =
3307 "PARAM c[1] = { { 1, 0.5 } };\n"
3308 "MOV result.position, vertex.position;\n"
3309 "MOV result.color, c[0].x;\n"
3310 "MOV result.texcoord[0], vertex.texcoord[0];\n"
3313 GL_EXTCALL(glGenProgramsARB(1, &program_id
));
3314 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, program_id
));
3315 shader_arb_compile(gl_info
, GL_VERTEX_PROGRAM_ARB
, blt_vprogram
);
3320 /* Context activation is done by the caller. */
3321 static GLuint
create_arb_blt_fragment_program(const struct wined3d_gl_info
*gl_info
,
3322 enum wined3d_gl_resource_type tex_type
, BOOL masked
)
3324 GLuint program_id
= 0;
3325 const char *fprogram
;
3327 static const char * const blt_fprograms_full
[WINED3D_GL_RES_TYPE_COUNT
] =
3329 /* WINED3D_GL_RES_TYPE_TEX_1D */
3331 /* WINED3D_GL_RES_TYPE_TEX_2D */
3334 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
3335 "MOV result.depth.z, R0.x;\n"
3337 /* WINED3D_GL_RES_TYPE_TEX_3D */
3339 /* WINED3D_GL_RES_TYPE_TEX_CUBE */
3342 "TEX R0.x, fragment.texcoord[0], texture[0], CUBE;\n"
3343 "MOV result.depth.z, R0.x;\n"
3345 /* WINED3D_GL_RES_TYPE_TEX_RECT */
3348 "TEX R0.x, fragment.texcoord[0], texture[0], RECT;\n"
3349 "MOV result.depth.z, R0.x;\n"
3353 static const char * const blt_fprograms_masked
[WINED3D_GL_RES_TYPE_COUNT
] =
3355 /* WINED3D_GL_RES_TYPE_TEX_1D */
3357 /* WINED3D_GL_RES_TYPE_TEX_2D */
3359 "PARAM mask = program.local[0];\n"
3361 "SLT R0.xy, fragment.position, mask.zwzw;\n"
3362 "MUL R0.x, R0.x, R0.y;\n"
3364 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
3365 "MOV result.depth.z, R0.x;\n"
3367 /* WINED3D_GL_RES_TYPE_TEX_3D */
3369 /* WINED3D_GL_RES_TYPE_TEX_CUBE */
3371 "PARAM mask = program.local[0];\n"
3373 "SLT R0.xy, fragment.position, mask.zwzw;\n"
3374 "MUL R0.x, R0.x, R0.y;\n"
3376 "TEX R0.x, fragment.texcoord[0], texture[0], CUBE;\n"
3377 "MOV result.depth.z, R0.x;\n"
3379 /* WINED3D_GL_RES_TYPE_TEX_RECT */
3381 "PARAM mask = program.local[0];\n"
3383 "SLT R0.xy, fragment.position, mask.zwzw;\n"
3384 "MUL R0.x, R0.x, R0.y;\n"
3386 "TEX R0.x, fragment.texcoord[0], texture[0], RECT;\n"
3387 "MOV result.depth.z, R0.x;\n"
3391 fprogram
= masked
? blt_fprograms_masked
[tex_type
] : blt_fprograms_full
[tex_type
];
3394 FIXME("tex_type %#x not supported, falling back to 2D\n", tex_type
);
3395 tex_type
= WINED3D_GL_RES_TYPE_TEX_2D
;
3396 fprogram
= masked
? blt_fprograms_masked
[tex_type
] : blt_fprograms_full
[tex_type
];
3399 GL_EXTCALL(glGenProgramsARB(1, &program_id
));
3400 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, program_id
));
3401 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, fprogram
);
3406 static void arbfp_add_sRGB_correction(struct wined3d_string_buffer
*buffer
, const char *fragcolor
,
3407 const char *tmp1
, const char *tmp2
, const char *tmp3
, const char *tmp4
, BOOL condcode
)
3409 /* Perform sRGB write correction. See GLX_EXT_framebuffer_sRGB */
3413 /* Sigh. MOVC CC doesn't work, so use one of the temps as dummy dest */
3414 shader_addline(buffer
, "SUBC %s, %s.x, srgb_consts1.x;\n", tmp1
, fragcolor
);
3415 /* Calculate the > 0.0031308 case */
3416 shader_addline(buffer
, "POW %s.x (GE), %s.x, srgb_consts0.x;\n", fragcolor
, fragcolor
);
3417 shader_addline(buffer
, "POW %s.y (GE), %s.y, srgb_consts0.x;\n", fragcolor
, fragcolor
);
3418 shader_addline(buffer
, "POW %s.z (GE), %s.z, srgb_consts0.x;\n", fragcolor
, fragcolor
);
3419 shader_addline(buffer
, "MUL %s.xyz (GE), %s, srgb_consts0.y;\n", fragcolor
, fragcolor
);
3420 shader_addline(buffer
, "SUB %s.xyz (GE), %s, srgb_consts0.z;\n", fragcolor
, fragcolor
);
3421 /* Calculate the < case */
3422 shader_addline(buffer
, "MUL %s.xyz (LT), srgb_consts0.w, %s;\n", fragcolor
, fragcolor
);
3426 /* Calculate the > 0.0031308 case */
3427 shader_addline(buffer
, "POW %s.x, %s.x, srgb_consts0.x;\n", tmp1
, fragcolor
);
3428 shader_addline(buffer
, "POW %s.y, %s.y, srgb_consts0.x;\n", tmp1
, fragcolor
);
3429 shader_addline(buffer
, "POW %s.z, %s.z, srgb_consts0.x;\n", tmp1
, fragcolor
);
3430 shader_addline(buffer
, "MUL %s, %s, srgb_consts0.y;\n", tmp1
, tmp1
);
3431 shader_addline(buffer
, "SUB %s, %s, srgb_consts0.z;\n", tmp1
, tmp1
);
3432 /* Calculate the < case */
3433 shader_addline(buffer
, "MUL %s, srgb_consts0.w, %s;\n", tmp2
, fragcolor
);
3434 /* Get 1.0 / 0.0 masks for > 0.0031308 and < 0.0031308 */
3435 shader_addline(buffer
, "SLT %s, srgb_consts1.x, %s;\n", tmp3
, fragcolor
);
3436 shader_addline(buffer
, "SGE %s, srgb_consts1.x, %s;\n", tmp4
, fragcolor
);
3437 /* Store the components > 0.0031308 in the destination */
3438 shader_addline(buffer
, "MUL %s.xyz, %s, %s;\n", fragcolor
, tmp1
, tmp3
);
3439 /* Add the components that are < 0.0031308 */
3440 shader_addline(buffer
, "MAD %s.xyz, %s, %s, %s;\n", fragcolor
, tmp2
, tmp4
, fragcolor
);
3441 /* Move everything into result.color at once. Nvidia hardware cannot handle partial
3442 * result.color writes(.rgb first, then .a), or handle overwriting already written
3443 * components. The assembler uses a temporary register in this case, which is usually
3444 * not allocated from one of our registers that were used earlier.
3447 /* [0.0;1.0] clamping. Not needed, this is done implicitly */
3450 static const DWORD
*find_loop_control_values(const struct wined3d_shader
*shader
, DWORD idx
)
3452 const struct wined3d_shader_lconst
*constant
;
3454 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsI
, struct wined3d_shader_lconst
, entry
)
3456 if (constant
->idx
== idx
)
3458 return constant
->value
;
3464 static void init_ps_input(const struct wined3d_shader
*shader
,
3465 const struct arb_ps_compile_args
*args
, struct shader_arb_ctx_priv
*priv
)
3467 static const char * const texcoords
[8] =
3469 "fragment.texcoord[0]", "fragment.texcoord[1]", "fragment.texcoord[2]", "fragment.texcoord[3]",
3470 "fragment.texcoord[4]", "fragment.texcoord[5]", "fragment.texcoord[6]", "fragment.texcoord[7]"
3473 const struct wined3d_shader_signature_element
*input
;
3474 const char *semantic_name
;
3477 switch(args
->super
.vp_mode
)
3479 case pretransformed
:
3481 /* The pixelshader has to collect the varyings on its own. In any case properly load
3482 * color0 and color1. In the case of pretransformed vertices also load texcoords. Set
3483 * other attribs to 0.0.
3485 * For fixedfunction this behavior is correct, according to the tests. For pretransformed
3486 * we'd either need a replacement shader that can load other attribs like BINORMAL, or
3487 * load the texcoord attrib pointers to match the pixel shader signature
3489 for (i
= 0; i
< shader
->input_signature
.element_count
; ++i
)
3491 input
= &shader
->input_signature
.elements
[i
];
3492 if (!(semantic_name
= input
->semantic_name
))
3494 semantic_idx
= input
->semantic_idx
;
3496 if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_COLOR
))
3499 priv
->ps_input
[input
->register_idx
] = "fragment.color.primary";
3500 else if (semantic_idx
== 1)
3501 priv
->ps_input
[input
->register_idx
] = "fragment.color.secondary";
3503 priv
->ps_input
[input
->register_idx
] = "0.0";
3505 else if (args
->super
.vp_mode
== fixedfunction
)
3507 priv
->ps_input
[input
->register_idx
] = "0.0";
3509 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
3511 if (semantic_idx
< 8)
3512 priv
->ps_input
[input
->register_idx
] = texcoords
[semantic_idx
];
3514 priv
->ps_input
[input
->register_idx
] = "0.0";
3516 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_FOG
))
3519 priv
->ps_input
[input
->register_idx
] = "fragment.fogcoord";
3521 priv
->ps_input
[input
->register_idx
] = "0.0";
3525 priv
->ps_input
[input
->register_idx
] = "0.0";
3528 TRACE("v%u, semantic %s%u is %s\n", input
->register_idx
,
3529 semantic_name
, semantic_idx
, priv
->ps_input
[input
->register_idx
]);
3534 /* That one is easy. The vertex shaders provide v0-v7 in fragment.texcoord and v8 and v9 in
3537 for(i
= 0; i
< 8; i
++)
3539 priv
->ps_input
[i
] = texcoords
[i
];
3541 priv
->ps_input
[8] = "fragment.color.primary";
3542 priv
->ps_input
[9] = "fragment.color.secondary";
3547 static void arbfp_add_linear_fog(struct wined3d_string_buffer
*buffer
,
3548 const char *fragcolor
, const char *tmp
)
3550 shader_addline(buffer
, "SUB %s.x, state.fog.params.z, fragment.fogcoord.x;\n", tmp
);
3551 shader_addline(buffer
, "MUL_SAT %s.x, %s.x, state.fog.params.w;\n", tmp
, tmp
);
3552 shader_addline(buffer
, "LRP %s.rgb, %s.x, %s, state.fog.color;\n", fragcolor
, tmp
, fragcolor
);
3555 /* Context activation is done by the caller. */
3556 static GLuint
shader_arb_generate_pshader(const struct wined3d_shader
*shader
,
3557 const struct wined3d_gl_info
*gl_info
, struct wined3d_string_buffer
*buffer
,
3558 const struct arb_ps_compile_args
*args
, struct arb_ps_compiled_shader
*compiled
)
3560 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
3561 const DWORD
*function
= shader
->function
;
3564 DWORD next_local
= 0;
3565 struct shader_arb_ctx_priv priv_ctx
;
3566 BOOL dcl_td
= FALSE
;
3567 BOOL want_nv_prog
= FALSE
;
3568 struct arb_pshader_private
*shader_priv
= shader
->backend_data
;
3570 BOOL custom_linear_fog
= FALSE
;
3574 unsigned int i
, found
= 0;
3576 for (i
= 0, map
= reg_maps
->temporary
; map
; map
>>= 1, ++i
)
3579 || (shader
->u
.ps
.color0_mov
&& i
== shader
->u
.ps
.color0_reg
)
3580 || (reg_maps
->shader_version
.major
< 2 && !i
))
3583 sprintf(srgbtmp
[found
], "R%u", i
);
3585 if (found
== 4) break;
3590 sprintf(srgbtmp
[0], "TA");
3591 sprintf(srgbtmp
[1], "TB");
3592 sprintf(srgbtmp
[2], "TC");
3593 sprintf(srgbtmp
[3], "TD");
3597 sprintf(srgbtmp
[1], "TA");
3598 sprintf(srgbtmp
[2], "TB");
3599 sprintf(srgbtmp
[3], "TC");
3602 sprintf(srgbtmp
[2], "TA");
3603 sprintf(srgbtmp
[3], "TB");
3606 sprintf(srgbtmp
[3], "TA");
3612 /* Create the hw ARB shader */
3613 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
3614 priv_ctx
.cur_ps_args
= args
;
3615 priv_ctx
.compiled_fprog
= compiled
;
3616 priv_ctx
.cur_np2fixup_info
= &compiled
->np2fixup_info
;
3617 init_ps_input(shader
, args
, &priv_ctx
);
3618 list_init(&priv_ctx
.control_frames
);
3619 priv_ctx
.ps_post_process
= args
->super
.srgb_correction
;
3621 /* Avoid enabling NV_fragment_program* if we do not need it.
3623 * Enabling GL_NV_fragment_program_option causes the driver to occupy a temporary register,
3624 * and it slows down the shader execution noticeably(about 5%). Usually our instruction emulation
3625 * is faster than what we gain from using higher native instructions. There are some things though
3626 * that cannot be emulated. In that case enable the extensions.
3627 * If the extension is enabled, instruction handlers that support both ways will use it.
3629 * Testing shows no performance difference between OPTION NV_fragment_program2 and NV_fragment_program.
3630 * So enable the best we can get.
3632 if(reg_maps
->usesdsx
|| reg_maps
->usesdsy
|| reg_maps
->loop_depth
> 0 || reg_maps
->usestexldd
||
3633 reg_maps
->usestexldl
|| reg_maps
->usesfacing
|| reg_maps
->usesifc
|| reg_maps
->usescall
)
3635 want_nv_prog
= TRUE
;
3638 shader_addline(buffer
, "!!ARBfp1.0\n");
3639 if (want_nv_prog
&& gl_info
->supported
[NV_FRAGMENT_PROGRAM2
])
3641 shader_addline(buffer
, "OPTION NV_fragment_program2;\n");
3642 priv_ctx
.target_version
= NV3
;
3644 else if (want_nv_prog
&& gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
])
3646 shader_addline(buffer
, "OPTION NV_fragment_program;\n");
3647 priv_ctx
.target_version
= NV2
;
3651 /* This is an error - either we're advertising the wrong shader version, or aren't enforcing some
3654 ERR("The shader requires instructions that are not available in plain GL_ARB_fragment_program\n");
3657 priv_ctx
.target_version
= ARB
;
3660 if (reg_maps
->rt_mask
> 1)
3662 shader_addline(buffer
, "OPTION ARB_draw_buffers;\n");
3665 if (reg_maps
->shader_version
.major
< 3)
3667 switch (args
->super
.fog
)
3669 case WINED3D_FFP_PS_FOG_OFF
:
3671 case WINED3D_FFP_PS_FOG_LINEAR
:
3672 if (gl_info
->quirks
& WINED3D_QUIRK_BROKEN_ARB_FOG
)
3674 custom_linear_fog
= TRUE
;
3675 priv_ctx
.ps_post_process
= TRUE
;
3678 shader_addline(buffer
, "OPTION ARB_fog_linear;\n");
3680 case WINED3D_FFP_PS_FOG_EXP
:
3681 shader_addline(buffer
, "OPTION ARB_fog_exp;\n");
3683 case WINED3D_FFP_PS_FOG_EXP2
:
3684 shader_addline(buffer
, "OPTION ARB_fog_exp2;\n");
3689 /* For now always declare the temps. At least the Nvidia assembler optimizes completely
3690 * unused temps away(but occupies them for the whole shader if they're used once). Always
3691 * declaring them avoids tricky bookkeeping work
3693 shader_addline(buffer
, "TEMP TA;\n"); /* Used for modifiers */
3694 shader_addline(buffer
, "TEMP TB;\n"); /* Used for modifiers */
3695 shader_addline(buffer
, "TEMP TC;\n"); /* Used for modifiers */
3696 if(dcl_td
) shader_addline(buffer
, "TEMP TD;\n"); /* Used for sRGB writing */
3697 shader_addline(buffer
, "PARAM coefdiv = { 0.5, 0.25, 0.125, 0.0625 };\n");
3698 shader_addline(buffer
, "PARAM coefmul = { 2, 4, 8, 16 };\n");
3699 wined3d_ftoa(eps
, ftoa_tmp
);
3700 shader_addline(buffer
, "PARAM ps_helper_const = { 0.0, 1.0, %s, 0.0 };\n", ftoa_tmp
);
3702 if (reg_maps
->shader_version
.major
< 2)
3704 strcpy(fragcolor
, "R0");
3708 if (priv_ctx
.ps_post_process
)
3710 if (shader
->u
.ps
.color0_mov
)
3712 sprintf(fragcolor
, "R%u", shader
->u
.ps
.color0_reg
);
3716 shader_addline(buffer
, "TEMP TMP_COLOR;\n");
3717 strcpy(fragcolor
, "TMP_COLOR");
3720 strcpy(fragcolor
, "result.color");
3724 if (args
->super
.srgb_correction
)
3726 shader_addline(buffer
, "PARAM srgb_consts0 = ");
3727 shader_arb_append_imm_vec4(buffer
, wined3d_srgb_const0
);
3728 shader_addline(buffer
, ";\n");
3729 shader_addline(buffer
, "PARAM srgb_consts1 = ");
3730 shader_arb_append_imm_vec4(buffer
, wined3d_srgb_const1
);
3731 shader_addline(buffer
, ";\n");
3734 /* Base Declarations */
3735 shader_generate_arb_declarations(shader
, reg_maps
, buffer
, gl_info
, NULL
, &priv_ctx
);
3737 for (i
= 0, map
= reg_maps
->bumpmat
; map
; map
>>= 1, ++i
)
3739 unsigned char bump_const
;
3741 if (!(map
& 1)) continue;
3743 bump_const
= compiled
->numbumpenvmatconsts
;
3744 compiled
->bumpenvmatconst
[bump_const
].const_num
= WINED3D_CONST_NUM_UNUSED
;
3745 compiled
->bumpenvmatconst
[bump_const
].texunit
= i
;
3746 compiled
->luminanceconst
[bump_const
].const_num
= WINED3D_CONST_NUM_UNUSED
;
3747 compiled
->luminanceconst
[bump_const
].texunit
= i
;
3749 /* We can fit the constants into the constant limit for sure because texbem, texbeml, bem and beml are only supported
3750 * in 1.x shaders, and GL_ARB_fragment_program has a constant limit of 24 constants. So in the worst case we're loading
3751 * 8 shader constants, 8 bump matrices and 8 luminance parameters and are perfectly fine. (No NP2 fixup on bumpmapped
3752 * textures due to conditional NP2 restrictions)
3754 * Use local constants to load the bump env parameters, not program.env. This avoids collisions with d3d constants of
3755 * shaders in newer shader models. Since the bump env parameters have to share their space with NP2 fixup constants,
3756 * their location is shader dependent anyway and they cannot be loaded globally.
3758 compiled
->bumpenvmatconst
[bump_const
].const_num
= next_local
++;
3759 shader_addline(buffer
, "PARAM bumpenvmat%d = program.local[%d];\n",
3760 i
, compiled
->bumpenvmatconst
[bump_const
].const_num
);
3761 compiled
->numbumpenvmatconsts
= bump_const
+ 1;
3763 if (!(reg_maps
->luminanceparams
& (1 << i
))) continue;
3765 compiled
->luminanceconst
[bump_const
].const_num
= next_local
++;
3766 shader_addline(buffer
, "PARAM luminance%d = program.local[%d];\n",
3767 i
, compiled
->luminanceconst
[bump_const
].const_num
);
3770 for(i
= 0; i
< MAX_CONST_I
; i
++)
3772 compiled
->int_consts
[i
] = WINED3D_CONST_NUM_UNUSED
;
3773 if (reg_maps
->integer_constants
& (1 << i
) && priv_ctx
.target_version
>= NV2
)
3775 const DWORD
*control_values
= find_loop_control_values(shader
, i
);
3779 shader_addline(buffer
, "PARAM I%u = {%u, %u, %u, -1};\n", i
,
3780 control_values
[0], control_values
[1], control_values
[2]);
3784 compiled
->int_consts
[i
] = next_local
;
3785 compiled
->num_int_consts
++;
3786 shader_addline(buffer
, "PARAM I%u = program.local[%u];\n", i
, next_local
++);
3791 if(reg_maps
->vpos
|| reg_maps
->usesdsy
)
3793 compiled
->ycorrection
= next_local
;
3794 shader_addline(buffer
, "PARAM ycorrection = program.local[%u];\n", next_local
++);
3798 shader_addline(buffer
, "TEMP vpos;\n");
3799 /* ycorrection.x: Backbuffer height(onscreen) or 0(offscreen).
3800 * ycorrection.y: -1.0(onscreen), 1.0(offscreen)
3801 * ycorrection.z: 1.0
3802 * ycorrection.w: 0.0
3804 shader_addline(buffer
, "MAD vpos, fragment.position, ycorrection.zyww, ycorrection.wxww;\n");
3805 shader_addline(buffer
, "FLR vpos.xy, vpos;\n");
3810 compiled
->ycorrection
= WINED3D_CONST_NUM_UNUSED
;
3813 /* Load constants to fixup NP2 texcoords if there are still free constants left:
3814 * Constants (texture dimensions) for the NP2 fixup are loaded as local program parameters. This will consume
3815 * at most 8 (MAX_FRAGMENT_SAMPLERS / 2) parameters, which is highly unlikely, since the application had to
3816 * use 16 NP2 textures at the same time. In case that we run out of constants the fixup is simply not
3817 * applied / activated. This will probably result in wrong rendering of the texture, but will save us from
3818 * shader compilation errors and the subsequent errors when drawing with this shader. */
3819 if (priv_ctx
.cur_ps_args
->super
.np2_fixup
) {
3820 unsigned char cur_fixup_sampler
= 0;
3822 struct arb_ps_np2fixup_info
* const fixup
= priv_ctx
.cur_np2fixup_info
;
3823 const WORD map
= priv_ctx
.cur_ps_args
->super
.np2_fixup
;
3824 const UINT max_lconsts
= gl_info
->limits
.arb_ps_local_constants
;
3826 fixup
->offset
= next_local
;
3827 fixup
->super
.active
= 0;
3829 for (i
= 0; i
< MAX_FRAGMENT_SAMPLERS
; ++i
) {
3830 if (!(map
& (1 << i
))) continue;
3832 if (fixup
->offset
+ (cur_fixup_sampler
>> 1) < max_lconsts
) {
3833 fixup
->super
.active
|= (1 << i
);
3834 fixup
->super
.idx
[i
] = cur_fixup_sampler
++;
3836 FIXME("No free constant found to load NP2 fixup data into shader. "
3837 "Sampling from this texture will probably look wrong.\n");
3842 fixup
->super
.num_consts
= (cur_fixup_sampler
+ 1) >> 1;
3843 if (fixup
->super
.num_consts
) {
3844 shader_addline(buffer
, "PARAM np2fixup[%u] = { program.env[%u..%u] };\n",
3845 fixup
->super
.num_consts
, fixup
->offset
, fixup
->super
.num_consts
+ fixup
->offset
- 1);
3849 if (shader_priv
->clipplane_emulation
!= ~0U && args
->clip
)
3851 shader_addline(buffer
, "KIL fragment.texcoord[%u];\n", shader_priv
->clipplane_emulation
);
3854 /* Base Shader Body */
3855 shader_generate_main(shader
, buffer
, reg_maps
, function
, &priv_ctx
);
3857 if(args
->super
.srgb_correction
) {
3858 arbfp_add_sRGB_correction(buffer
, fragcolor
, srgbtmp
[0], srgbtmp
[1], srgbtmp
[2], srgbtmp
[3],
3859 priv_ctx
.target_version
>= NV2
);
3862 if (custom_linear_fog
)
3863 arbfp_add_linear_fog(buffer
, fragcolor
, "TA");
3865 if(strcmp(fragcolor
, "result.color")) {
3866 shader_addline(buffer
, "MOV result.color, %s;\n", fragcolor
);
3868 shader_addline(buffer
, "END\n");
3870 /* TODO: change to resource.glObjectHandle or something like that */
3871 GL_EXTCALL(glGenProgramsARB(1, &retval
));
3873 TRACE("Creating a hw pixel shader, prg=%d\n", retval
);
3874 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, retval
));
3876 TRACE("Created hw pixel shader, prg=%d\n", retval
);
3877 if (!shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
->buffer
))
3883 static int compare_sig(const struct wined3d_shader_signature
*sig1
, const struct wined3d_shader_signature
*sig2
)
3888 if (sig1
->element_count
!= sig2
->element_count
)
3889 return sig1
->element_count
< sig2
->element_count
? -1 : 1;
3891 for (i
= 0; i
< sig1
->element_count
; ++i
)
3893 const struct wined3d_shader_signature_element
*e1
, *e2
;
3895 e1
= &sig1
->elements
[i
];
3896 e2
= &sig2
->elements
[i
];
3898 if (!e1
->semantic_name
|| !e2
->semantic_name
)
3900 /* Compare pointers, not contents. One string is NULL (element
3901 * does not exist), the other one is not NULL. */
3902 if (e1
->semantic_name
!= e2
->semantic_name
)
3903 return e1
->semantic_name
< e2
->semantic_name
? -1 : 1;
3907 if ((ret
= strcmp(e1
->semantic_name
, e2
->semantic_name
)))
3909 if (e1
->semantic_idx
!= e2
->semantic_idx
)
3910 return e1
->semantic_idx
< e2
->semantic_idx
? -1 : 1;
3911 if (e1
->sysval_semantic
!= e2
->sysval_semantic
)
3912 return e1
->sysval_semantic
< e2
->sysval_semantic
? -1 : 1;
3913 if (e1
->component_type
!= e2
->component_type
)
3914 return e1
->component_type
< e2
->component_type
? -1 : 1;
3915 if (e1
->register_idx
!= e2
->register_idx
)
3916 return e1
->register_idx
< e2
->register_idx
? -1 : 1;
3917 if (e1
->mask
!= e2
->mask
)
3918 return e1
->mask
< e2
->mask
? -1 : 1;
3923 static void clone_sig(struct wined3d_shader_signature
*new, const struct wined3d_shader_signature
*sig
)
3928 new->element_count
= sig
->element_count
;
3929 new->elements
= HeapAlloc(GetProcessHeap(), 0, sizeof(*new->elements
) * new->element_count
);
3930 for (i
= 0; i
< sig
->element_count
; ++i
)
3932 new->elements
[i
] = sig
->elements
[i
];
3934 if (!new->elements
[i
].semantic_name
)
3937 /* Clone the semantic string */
3938 name
= HeapAlloc(GetProcessHeap(), 0, strlen(sig
->elements
[i
].semantic_name
) + 1);
3939 strcpy(name
, sig
->elements
[i
].semantic_name
);
3940 new->elements
[i
].semantic_name
= name
;
3944 static DWORD
find_input_signature(struct shader_arb_priv
*priv
, const struct wined3d_shader_signature
*sig
)
3946 struct wine_rb_entry
*entry
= wine_rb_get(&priv
->signature_tree
, sig
);
3947 struct ps_signature
*found_sig
;
3951 found_sig
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
3952 TRACE("Found existing signature %u\n", found_sig
->idx
);
3953 return found_sig
->idx
;
3955 found_sig
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*found_sig
));
3956 clone_sig(&found_sig
->sig
, sig
);
3957 found_sig
->idx
= priv
->ps_sig_number
++;
3958 TRACE("New signature stored and assigned number %u\n", found_sig
->idx
);
3959 if(wine_rb_put(&priv
->signature_tree
, sig
, &found_sig
->entry
) == -1)
3961 ERR("Failed to insert program entry.\n");
3963 return found_sig
->idx
;
3966 static void init_output_registers(const struct wined3d_shader
*shader
,
3967 const struct wined3d_shader_signature
*ps_input_sig
,
3968 struct shader_arb_ctx_priv
*priv_ctx
, struct arb_vs_compiled_shader
*compiled
)
3971 static const char * const texcoords
[8] =
3973 "result.texcoord[0]", "result.texcoord[1]", "result.texcoord[2]", "result.texcoord[3]",
3974 "result.texcoord[4]", "result.texcoord[5]", "result.texcoord[6]", "result.texcoord[7]"
3976 /* Write generic input varyings 0 to 7 to result.texcoord[], varying 8 to result.color.primary
3977 * and varying 9 to result.color.secondary
3979 static const char * const decl_idx_to_string
[MAX_REG_INPUT
] =
3981 "result.texcoord[0]", "result.texcoord[1]", "result.texcoord[2]", "result.texcoord[3]",
3982 "result.texcoord[4]", "result.texcoord[5]", "result.texcoord[6]", "result.texcoord[7]",
3983 "result.color.primary", "result.color.secondary"
3988 TRACE("Pixel shader uses builtin varyings\n");
3989 /* Map builtins to builtins */
3990 for(i
= 0; i
< 8; i
++)
3992 priv_ctx
->texcrd_output
[i
] = texcoords
[i
];
3994 priv_ctx
->color_output
[0] = "result.color.primary";
3995 priv_ctx
->color_output
[1] = "result.color.secondary";
3996 priv_ctx
->fog_output
= "TMP_FOGCOORD";
3998 /* Map declared regs to builtins. Use "TA" to /dev/null unread output */
3999 for (i
= 0; i
< shader
->output_signature
.element_count
; ++i
)
4001 const struct wined3d_shader_signature_element
*output
= &shader
->output_signature
.elements
[i
];
4003 if (!output
->semantic_name
)
4006 if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_POSITION
))
4008 TRACE("o%u is TMP_OUT\n", output
->register_idx
);
4009 if (!output
->semantic_idx
)
4010 priv_ctx
->vs_output
[output
->register_idx
] = "TMP_OUT";
4012 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4014 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_PSIZE
))
4016 TRACE("o%u is result.pointsize\n", output
->register_idx
);
4017 if (!output
->semantic_idx
)
4018 priv_ctx
->vs_output
[output
->register_idx
] = "result.pointsize";
4020 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4022 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_COLOR
))
4024 TRACE("o%u is result.color.?, idx %u\n", output
->register_idx
, output
->semantic_idx
);
4025 if (!output
->semantic_idx
)
4026 priv_ctx
->vs_output
[output
->register_idx
] = "result.color.primary";
4027 else if (output
->semantic_idx
== 1)
4028 priv_ctx
->vs_output
[output
->register_idx
] = "result.color.secondary";
4029 else priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4031 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
4033 TRACE("o%u is result.texcoord[%u]\n", output
->register_idx
, output
->semantic_idx
);
4034 if (output
->semantic_idx
>= 8)
4035 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4037 priv_ctx
->vs_output
[output
->register_idx
] = texcoords
[output
->semantic_idx
];
4039 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_FOG
))
4041 TRACE("o%u is result.fogcoord\n", output
->register_idx
);
4042 if (output
->semantic_idx
> 0)
4043 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4045 priv_ctx
->vs_output
[output
->register_idx
] = "result.fogcoord";
4049 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4055 TRACE("Pixel shader uses declared varyings\n");
4057 /* Map builtin to declared. /dev/null the results by default to the TA temp reg */
4058 for(i
= 0; i
< 8; i
++)
4060 priv_ctx
->texcrd_output
[i
] = "TA";
4062 priv_ctx
->color_output
[0] = "TA";
4063 priv_ctx
->color_output
[1] = "TA";
4064 priv_ctx
->fog_output
= "TA";
4066 for (i
= 0; i
< ps_input_sig
->element_count
; ++i
)
4068 const struct wined3d_shader_signature_element
*input
= &ps_input_sig
->elements
[i
];
4070 if (!input
->semantic_name
)
4073 /* If a declared input register is not written by builtin arguments, don't write to it.
4074 * GL_NV_vertex_program makes sure the input defaults to 0.0, which is correct with D3D
4076 * Don't care about POSITION and PSIZE here - this is a builtin vertex shader, position goes
4077 * to TMP_OUT in any case
4079 if (shader_match_semantic(input
->semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
4081 if (input
->semantic_idx
< 8)
4082 priv_ctx
->texcrd_output
[input
->semantic_idx
] = decl_idx_to_string
[input
->register_idx
];
4084 else if (shader_match_semantic(input
->semantic_name
, WINED3D_DECL_USAGE_COLOR
))
4086 if (input
->semantic_idx
< 2)
4087 priv_ctx
->color_output
[input
->semantic_idx
] = decl_idx_to_string
[input
->register_idx
];
4089 else if (shader_match_semantic(input
->semantic_name
, WINED3D_DECL_USAGE_FOG
))
4091 if (!input
->semantic_idx
)
4092 priv_ctx
->fog_output
= decl_idx_to_string
[input
->register_idx
];
4099 if (!strcmp(decl_idx_to_string
[input
->register_idx
], "result.color.primary")
4100 || !strcmp(decl_idx_to_string
[input
->register_idx
], "result.color.secondary"))
4102 compiled
->need_color_unclamp
= TRUE
;
4106 /* Map declared to declared */
4107 for (i
= 0; i
< shader
->output_signature
.element_count
; ++i
)
4109 const struct wined3d_shader_signature_element
*output
= &shader
->output_signature
.elements
[i
];
4111 /* Write unread output to TA to throw them away */
4112 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4114 if (!output
->semantic_name
)
4117 if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_POSITION
) && !output
->semantic_idx
)
4119 priv_ctx
->vs_output
[output
->register_idx
] = "TMP_OUT";
4122 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_PSIZE
) && !output
->semantic_idx
)
4124 priv_ctx
->vs_output
[output
->register_idx
] = "result.pointsize";
4128 for (j
= 0; j
< ps_input_sig
->element_count
; ++j
)
4130 const struct wined3d_shader_signature_element
*input
= &ps_input_sig
->elements
[j
];
4132 if (!input
->semantic_name
)
4135 if (!strcmp(input
->semantic_name
, output
->semantic_name
)
4136 && input
->semantic_idx
== output
->semantic_idx
)
4138 priv_ctx
->vs_output
[output
->register_idx
] = decl_idx_to_string
[input
->register_idx
];
4140 if (!strcmp(priv_ctx
->vs_output
[output
->register_idx
], "result.color.primary")
4141 || !strcmp(priv_ctx
->vs_output
[output
->register_idx
], "result.color.secondary"))
4143 compiled
->need_color_unclamp
= TRUE
;
4150 /* Context activation is done by the caller. */
4151 static GLuint
shader_arb_generate_vshader(const struct wined3d_shader
*shader
,
4152 const struct wined3d_gl_info
*gl_info
, struct wined3d_string_buffer
*buffer
,
4153 const struct arb_vs_compile_args
*args
, struct arb_vs_compiled_shader
*compiled
,
4154 const struct wined3d_shader_signature
*ps_input_sig
)
4156 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
4157 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
4158 struct shader_arb_priv
*priv
= shader
->device
->shader_priv
;
4159 const DWORD
*function
= shader
->function
;
4161 DWORD next_local
= 0;
4162 struct shader_arb_ctx_priv priv_ctx
;
4165 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
4166 priv_ctx
.cur_vs_args
= args
;
4167 list_init(&priv_ctx
.control_frames
);
4168 init_output_registers(shader
, ps_input_sig
, &priv_ctx
, compiled
);
4170 /* Create the hw ARB shader */
4171 shader_addline(buffer
, "!!ARBvp1.0\n");
4173 /* Always enable the NV extension if available. Unlike fragment shaders, there is no
4174 * mesurable performance penalty, and we can always make use of it for clipplanes.
4176 if (gl_info
->supported
[NV_VERTEX_PROGRAM3
])
4178 shader_addline(buffer
, "OPTION NV_vertex_program3;\n");
4179 priv_ctx
.target_version
= NV3
;
4180 shader_addline(buffer
, "ADDRESS aL;\n");
4182 else if (gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
])
4184 shader_addline(buffer
, "OPTION NV_vertex_program2;\n");
4185 priv_ctx
.target_version
= NV2
;
4186 shader_addline(buffer
, "ADDRESS aL;\n");
4188 priv_ctx
.target_version
= ARB
;
4191 shader_addline(buffer
, "TEMP TMP_OUT;\n");
4193 shader_addline(buffer
, "TEMP TMP_FOGCOORD;\n");
4194 if (need_helper_const(shader_data
, reg_maps
, gl_info
))
4197 wined3d_ftoa(eps
, ftoa_tmp
);
4198 shader_addline(buffer
, "PARAM helper_const = { 0.0, 1.0, 2.0, %s};\n", ftoa_tmp
);
4200 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
))
4202 shader_addline(buffer
, "PARAM rel_addr_const = { 0.5, %d.0, 0.0, 0.0 };\n", shader_data
->rel_offset
);
4203 shader_addline(buffer
, "TEMP A0_SHADOW;\n");
4206 shader_addline(buffer
, "TEMP TA;\n");
4207 shader_addline(buffer
, "TEMP TB;\n");
4209 /* Base Declarations */
4210 shader_generate_arb_declarations(shader
, reg_maps
, buffer
, gl_info
,
4211 &priv_ctx
.vs_clipplanes
, &priv_ctx
);
4213 for(i
= 0; i
< MAX_CONST_I
; i
++)
4215 compiled
->int_consts
[i
] = WINED3D_CONST_NUM_UNUSED
;
4216 if(reg_maps
->integer_constants
& (1 << i
) && priv_ctx
.target_version
>= NV2
)
4218 const DWORD
*control_values
= find_loop_control_values(shader
, i
);
4222 shader_addline(buffer
, "PARAM I%u = {%u, %u, %u, -1};\n", i
,
4223 control_values
[0], control_values
[1], control_values
[2]);
4227 compiled
->int_consts
[i
] = next_local
;
4228 compiled
->num_int_consts
++;
4229 shader_addline(buffer
, "PARAM I%u = program.local[%u];\n", i
, next_local
++);
4234 /* We need a constant to fixup the final position */
4235 shader_addline(buffer
, "PARAM posFixup = program.local[%u];\n", next_local
);
4236 compiled
->pos_fixup
= next_local
++;
4238 /* Initialize output parameters. GL_ARB_vertex_program does not require special initialization values
4239 * for output parameters. D3D in theory does not do that either, but some applications depend on a
4240 * proper initialization of the secondary color, and programs using the fixed function pipeline without
4241 * a replacement shader depend on the texcoord.w being set properly.
4243 * GL_NV_vertex_program defines that all output values are initialized to {0.0, 0.0, 0.0, 1.0}. This
4244 * assertion is in effect even when using GL_ARB_vertex_program without any NV specific additions. So
4245 * skip this if NV_vertex_program is supported. Otherwise, initialize the secondary color. For the tex-
4246 * coords, we have a flag in the opengl caps. Many cards do not require the texcoord being set, and
4247 * this can eat a number of instructions, so skip it unless this cap is set as well
4249 if (!gl_info
->supported
[NV_VERTEX_PROGRAM
])
4251 const char *color_init
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_0001
);
4252 shader_addline(buffer
, "MOV result.color.secondary, %s;\n", color_init
);
4254 if (gl_info
->quirks
& WINED3D_QUIRK_SET_TEXCOORD_W
&& !priv
->ffp_proj_control
)
4257 const char *one
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_ONE
);
4258 for(i
= 0; i
< MAX_REG_TEXCRD
; i
++)
4260 if (reg_maps
->texcoord_mask
[i
] && reg_maps
->texcoord_mask
[i
] != WINED3DSP_WRITEMASK_ALL
)
4261 shader_addline(buffer
, "MOV result.texcoord[%u].w, %s\n", i
, one
);
4266 /* The shader starts with the main function */
4267 priv_ctx
.in_main_func
= TRUE
;
4268 /* Base Shader Body */
4269 shader_generate_main(shader
, buffer
, reg_maps
, function
, &priv_ctx
);
4271 if (!priv_ctx
.footer_written
) vshader_add_footer(&priv_ctx
,
4272 shader_data
, args
, reg_maps
, gl_info
, buffer
);
4274 shader_addline(buffer
, "END\n");
4276 /* TODO: change to resource.glObjectHandle or something like that */
4277 GL_EXTCALL(glGenProgramsARB(1, &ret
));
4279 TRACE("Creating a hw vertex shader, prg=%d\n", ret
);
4280 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, ret
));
4282 TRACE("Created hw vertex shader, prg=%d\n", ret
);
4283 if (!shader_arb_compile(gl_info
, GL_VERTEX_PROGRAM_ARB
, buffer
->buffer
))
4289 /* Context activation is done by the caller. */
4290 static struct arb_ps_compiled_shader
*find_arb_pshader(struct wined3d_shader
*shader
,
4291 const struct arb_ps_compile_args
*args
)
4293 struct wined3d_device
*device
= shader
->device
;
4294 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
4295 const struct wined3d_d3d_info
*d3d_info
= &device
->adapter
->d3d_info
;
4298 struct arb_ps_compiled_shader
*new_array
;
4299 struct wined3d_string_buffer buffer
;
4300 struct arb_pshader_private
*shader_data
;
4303 if (!shader
->backend_data
)
4305 struct shader_arb_priv
*priv
= device
->shader_priv
;
4307 shader
->backend_data
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
4308 shader_data
= shader
->backend_data
;
4309 shader_data
->clamp_consts
= shader
->reg_maps
.shader_version
.major
== 1;
4311 if (shader
->reg_maps
.shader_version
.major
< 3)
4312 shader_data
->input_signature_idx
= ~0U;
4314 shader_data
->input_signature_idx
= find_input_signature(priv
, &shader
->input_signature
);
4316 TRACE("Shader got assigned input signature index %u\n", shader_data
->input_signature_idx
);
4318 if (!d3d_info
->vs_clipping
)
4319 shader_data
->clipplane_emulation
= shader_find_free_input_register(&shader
->reg_maps
,
4320 d3d_info
->limits
.ffp_blend_stages
- 1);
4322 shader_data
->clipplane_emulation
= ~0U;
4324 shader_data
= shader
->backend_data
;
4326 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4327 * so a linear search is more performant than a hashmap or a binary search
4328 * (cache coherency etc)
4330 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4332 if (!memcmp(&shader_data
->gl_shaders
[i
].args
, args
, sizeof(*args
)))
4333 return &shader_data
->gl_shaders
[i
];
4336 TRACE("No matching GL shader found, compiling a new shader\n");
4337 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4338 if (shader_data
->num_gl_shaders
)
4340 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4341 new_array
= HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, shader_data
->gl_shaders
,
4342 new_size
* sizeof(*shader_data
->gl_shaders
));
4344 new_array
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
->gl_shaders
));
4349 ERR("Out of memory\n");
4352 shader_data
->gl_shaders
= new_array
;
4353 shader_data
->shader_array_size
= new_size
;
4356 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4358 pixelshader_update_resource_types(shader
, args
->super
.tex_types
);
4360 if (!string_buffer_init(&buffer
))
4362 ERR("Failed to initialize shader buffer.\n");
4366 ret
= shader_arb_generate_pshader(shader
, gl_info
, &buffer
, args
,
4367 &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
]);
4368 string_buffer_free(&buffer
);
4369 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].prgId
= ret
;
4371 return &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++];
4374 static inline BOOL
vs_args_equal(const struct arb_vs_compile_args
*stored
, const struct arb_vs_compile_args
*new,
4375 const DWORD use_map
, BOOL skip_int
) {
4376 if((stored
->super
.swizzle_map
& use_map
) != new->super
.swizzle_map
) return FALSE
;
4377 if(stored
->super
.clip_enabled
!= new->super
.clip_enabled
) return FALSE
;
4378 if(stored
->super
.fog_src
!= new->super
.fog_src
) return FALSE
;
4379 if(stored
->clip
.boolclip_compare
!= new->clip
.boolclip_compare
) return FALSE
;
4380 if(stored
->ps_signature
!= new->ps_signature
) return FALSE
;
4381 if(stored
->vertex
.samplers_compare
!= new->vertex
.samplers_compare
) return FALSE
;
4382 if(skip_int
) return TRUE
;
4384 return !memcmp(stored
->loop_ctrl
, new->loop_ctrl
, sizeof(stored
->loop_ctrl
));
4387 static struct arb_vs_compiled_shader
*find_arb_vshader(struct wined3d_shader
*shader
,
4388 const struct wined3d_gl_info
*gl_info
, DWORD use_map
, const struct arb_vs_compile_args
*args
,
4389 const struct wined3d_shader_signature
*ps_input_sig
)
4393 struct arb_vs_compiled_shader
*new_array
;
4394 struct wined3d_string_buffer buffer
;
4395 struct arb_vshader_private
*shader_data
;
4398 if (!shader
->backend_data
)
4400 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
4402 shader
->backend_data
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
4403 shader_data
= shader
->backend_data
;
4405 if ((gl_info
->quirks
& WINED3D_QUIRK_ARB_VS_OFFSET_LIMIT
)
4406 && reg_maps
->min_rel_offset
<= reg_maps
->max_rel_offset
)
4408 if (reg_maps
->max_rel_offset
- reg_maps
->min_rel_offset
> 127)
4410 FIXME("The difference between the minimum and maximum relative offset is > 127.\n");
4411 FIXME("Which this OpenGL implementation does not support. Try using GLSL.\n");
4412 FIXME("Min: %u, Max: %u.\n", reg_maps
->min_rel_offset
, reg_maps
->max_rel_offset
);
4414 else if (reg_maps
->max_rel_offset
- reg_maps
->min_rel_offset
> 63)
4415 shader_data
->rel_offset
= reg_maps
->min_rel_offset
+ 63;
4416 else if (reg_maps
->max_rel_offset
> 63)
4417 shader_data
->rel_offset
= reg_maps
->min_rel_offset
;
4420 shader_data
= shader
->backend_data
;
4422 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4423 * so a linear search is more performant than a hashmap or a binary search
4424 * (cache coherency etc)
4426 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
4427 if (vs_args_equal(&shader_data
->gl_shaders
[i
].args
, args
,
4428 use_map
, gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]))
4430 return &shader_data
->gl_shaders
[i
];
4434 TRACE("No matching GL shader found, compiling a new shader\n");
4436 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4437 if (shader_data
->num_gl_shaders
)
4439 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4440 new_array
= HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, shader_data
->gl_shaders
,
4441 new_size
* sizeof(*shader_data
->gl_shaders
));
4443 new_array
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
->gl_shaders
));
4448 ERR("Out of memory\n");
4451 shader_data
->gl_shaders
= new_array
;
4452 shader_data
->shader_array_size
= new_size
;
4455 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4457 if (!string_buffer_init(&buffer
))
4459 ERR("Failed to initialize shader buffer.\n");
4463 ret
= shader_arb_generate_vshader(shader
, gl_info
, &buffer
, args
,
4464 &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
],
4466 string_buffer_free(&buffer
);
4467 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].prgId
= ret
;
4469 return &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++];
4472 static void find_arb_ps_compile_args(const struct wined3d_state
*state
,
4473 const struct wined3d_context
*context
, const struct wined3d_shader
*shader
,
4474 struct arb_ps_compile_args
*args
)
4476 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4477 const struct wined3d_d3d_info
*d3d_info
= context
->d3d_info
;
4481 find_ps_compile_args(state
, shader
, context
->stream_info
.position_transformed
, &args
->super
, gl_info
);
4483 /* This forces all local boolean constants to 1 to make them stateblock independent */
4484 args
->bools
= shader
->reg_maps
.local_bool_consts
;
4486 for(i
= 0; i
< MAX_CONST_B
; i
++)
4488 if (state
->ps_consts_b
[i
])
4489 args
->bools
|= ( 1 << i
);
4492 /* Only enable the clip plane emulation KIL if at least one clipplane is enabled. The KIL instruction
4493 * is quite expensive because it forces the driver to disable early Z discards. It is cheaper to
4494 * duplicate the shader than have a no-op KIL instruction in every shader
4496 if (!d3d_info
->vs_clipping
&& use_vs(state
)
4497 && state
->render_states
[WINED3D_RS_CLIPPING
]
4498 && state
->render_states
[WINED3D_RS_CLIPPLANEENABLE
])
4503 /* Skip if unused or local, or supported natively */
4504 int_skip
= ~shader
->reg_maps
.integer_constants
| shader
->reg_maps
.local_int_consts
;
4505 if (int_skip
== 0xffff || gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
])
4507 memset(args
->loop_ctrl
, 0, sizeof(args
->loop_ctrl
));
4511 for(i
= 0; i
< MAX_CONST_I
; i
++)
4513 if(int_skip
& (1 << i
))
4515 args
->loop_ctrl
[i
][0] = 0;
4516 args
->loop_ctrl
[i
][1] = 0;
4517 args
->loop_ctrl
[i
][2] = 0;
4521 args
->loop_ctrl
[i
][0] = state
->ps_consts_i
[i
* 4];
4522 args
->loop_ctrl
[i
][1] = state
->ps_consts_i
[i
* 4 + 1];
4523 args
->loop_ctrl
[i
][2] = state
->ps_consts_i
[i
* 4 + 2];
4528 static void find_arb_vs_compile_args(const struct wined3d_state
*state
,
4529 const struct wined3d_context
*context
, const struct wined3d_shader
*shader
,
4530 struct arb_vs_compile_args
*args
)
4532 const struct wined3d_device
*device
= shader
->device
;
4533 const struct wined3d_adapter
*adapter
= device
->adapter
;
4534 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4535 const struct wined3d_d3d_info
*d3d_info
= context
->d3d_info
;
4539 find_vs_compile_args(state
, shader
, context
->stream_info
.swizzle_map
, &args
->super
);
4541 args
->clip
.boolclip_compare
= 0;
4544 const struct wined3d_shader
*ps
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
4545 const struct arb_pshader_private
*shader_priv
= ps
->backend_data
;
4546 args
->ps_signature
= shader_priv
->input_signature_idx
;
4548 args
->clip
.boolclip
.clip_texcoord
= shader_priv
->clipplane_emulation
+ 1;
4552 args
->ps_signature
= ~0;
4553 if (!d3d_info
->vs_clipping
&& adapter
->fragment_pipe
== &arbfp_fragment_pipeline
)
4554 args
->clip
.boolclip
.clip_texcoord
= ffp_clip_emul(context
) ? d3d_info
->limits
.ffp_blend_stages
: 0;
4555 /* Otherwise: Setting boolclip_compare set clip_texcoord to 0 */
4558 if (args
->clip
.boolclip
.clip_texcoord
)
4560 if (state
->render_states
[WINED3D_RS_CLIPPING
])
4561 args
->clip
.boolclip
.clipplane_mask
= (unsigned char)state
->render_states
[WINED3D_RS_CLIPPLANEENABLE
];
4562 /* clipplane_mask was set to 0 by setting boolclip_compare to 0 */
4565 /* This forces all local boolean constants to 1 to make them stateblock independent */
4566 args
->clip
.boolclip
.bools
= shader
->reg_maps
.local_bool_consts
;
4567 /* TODO: Figure out if it would be better to store bool constants as bitmasks in the stateblock */
4568 for(i
= 0; i
< MAX_CONST_B
; i
++)
4570 if (state
->vs_consts_b
[i
])
4571 args
->clip
.boolclip
.bools
|= ( 1 << i
);
4574 args
->vertex
.samplers
[0] = context
->tex_unit_map
[MAX_FRAGMENT_SAMPLERS
+ 0];
4575 args
->vertex
.samplers
[1] = context
->tex_unit_map
[MAX_FRAGMENT_SAMPLERS
+ 1];
4576 args
->vertex
.samplers
[2] = context
->tex_unit_map
[MAX_FRAGMENT_SAMPLERS
+ 2];
4577 args
->vertex
.samplers
[3] = 0;
4579 /* Skip if unused or local */
4580 int_skip
= ~shader
->reg_maps
.integer_constants
| shader
->reg_maps
.local_int_consts
;
4581 /* This is about flow control, not clipping. */
4582 if (int_skip
== 0xffff || gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
])
4584 memset(args
->loop_ctrl
, 0, sizeof(args
->loop_ctrl
));
4588 for(i
= 0; i
< MAX_CONST_I
; i
++)
4590 if(int_skip
& (1 << i
))
4592 args
->loop_ctrl
[i
][0] = 0;
4593 args
->loop_ctrl
[i
][1] = 0;
4594 args
->loop_ctrl
[i
][2] = 0;
4598 args
->loop_ctrl
[i
][0] = state
->vs_consts_i
[i
* 4];
4599 args
->loop_ctrl
[i
][1] = state
->vs_consts_i
[i
* 4 + 1];
4600 args
->loop_ctrl
[i
][2] = state
->vs_consts_i
[i
* 4 + 2];
4605 /* Context activation is done by the caller. */
4606 static void shader_arb_select(void *shader_priv
, struct wined3d_context
*context
,
4607 const struct wined3d_state
*state
)
4609 struct shader_arb_priv
*priv
= shader_priv
;
4610 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4613 /* Deal with pixel shaders first so the vertex shader arg function has the input signature ready */
4616 struct wined3d_shader
*ps
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
4617 struct arb_ps_compile_args compile_args
;
4618 struct arb_ps_compiled_shader
*compiled
;
4620 TRACE("Using pixel shader %p.\n", ps
);
4621 find_arb_ps_compile_args(state
, context
, ps
, &compile_args
);
4622 compiled
= find_arb_pshader(ps
, &compile_args
);
4623 priv
->current_fprogram_id
= compiled
->prgId
;
4624 priv
->compiled_fprog
= compiled
;
4626 /* Bind the fragment program */
4627 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, priv
->current_fprogram_id
));
4628 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->current_fprogram_id);");
4630 if (!priv
->use_arbfp_fixed_func
)
4631 priv
->fragment_pipe
->enable_extension(gl_info
, FALSE
);
4633 /* Enable OpenGL fragment programs. */
4634 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
4635 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
4637 TRACE("Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", priv
->current_fprogram_id
);
4639 /* Pixel Shader 1.x constants are clamped to [-1;1], Pixel Shader 2.0 constants are not. If switching between
4640 * a 1.x and newer shader, reload the first 8 constants
4642 if (priv
->last_ps_const_clamped
!= ((struct arb_pshader_private
*)ps
->backend_data
)->clamp_consts
)
4644 priv
->last_ps_const_clamped
= ((struct arb_pshader_private
*)ps
->backend_data
)->clamp_consts
;
4645 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, 8);
4646 for(i
= 0; i
< 8; i
++)
4648 priv
->pshader_const_dirty
[i
] = 1;
4650 /* Also takes care of loading local constants */
4651 shader_arb_load_constants_internal(shader_priv
, context
, state
, TRUE
, FALSE
, TRUE
);
4655 UINT rt_height
= state
->fb
->render_targets
[0]->height
;
4656 shader_arb_ps_local_constants(compiled
, context
, state
, rt_height
);
4659 /* Force constant reloading for the NP2 fixup (see comment in shader_glsl_select for more info) */
4660 if (compiled
->np2fixup_info
.super
.active
)
4661 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_NP2_FIXUP
;
4663 if (ps
->load_local_constsF
)
4664 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_F
;
4668 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
] && !priv
->use_arbfp_fixed_func
)
4670 /* Disable only if we're not using arbfp fixed function fragment
4671 * processing. If this is used, keep GL_FRAGMENT_PROGRAM_ARB
4672 * enabled, and the fixed function pipeline will bind the fixed
4673 * function replacement shader. */
4674 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4675 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4676 priv
->current_fprogram_id
= 0;
4678 priv
->fragment_pipe
->enable_extension(gl_info
, TRUE
);
4683 struct wined3d_shader
*vs
= state
->shader
[WINED3D_SHADER_TYPE_VERTEX
];
4684 struct arb_vs_compile_args compile_args
;
4685 struct arb_vs_compiled_shader
*compiled
;
4686 const struct wined3d_shader_signature
*ps_input_sig
;
4688 TRACE("Using vertex shader %p\n", vs
);
4689 find_arb_vs_compile_args(state
, context
, vs
, &compile_args
);
4691 /* Instead of searching for the signature in the signature list, read the one from the
4692 * current pixel shader. It's maybe not the shader where the signature came from, but it
4693 * is the same signature and faster to find. */
4694 if (compile_args
.ps_signature
== ~0U)
4695 ps_input_sig
= NULL
;
4697 ps_input_sig
= &state
->shader
[WINED3D_SHADER_TYPE_PIXEL
]->input_signature
;
4699 compiled
= find_arb_vshader(vs
, context
->gl_info
, context
->stream_info
.use_map
,
4700 &compile_args
, ps_input_sig
);
4701 priv
->current_vprogram_id
= compiled
->prgId
;
4702 priv
->compiled_vprog
= compiled
;
4704 /* Bind the vertex program */
4705 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->current_vprogram_id
));
4706 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->current_vprogram_id);");
4708 priv
->vertex_pipe
->vp_enable(gl_info
, FALSE
);
4710 /* Enable OpenGL vertex programs */
4711 gl_info
->gl_ops
.gl
.p_glEnable(GL_VERTEX_PROGRAM_ARB
);
4712 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
4713 TRACE("Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", priv
->current_vprogram_id
);
4714 shader_arb_vs_local_constants(compiled
, context
, state
);
4716 if(priv
->last_vs_color_unclamp
!= compiled
->need_color_unclamp
) {
4717 priv
->last_vs_color_unclamp
= compiled
->need_color_unclamp
;
4719 if (gl_info
->supported
[ARB_COLOR_BUFFER_FLOAT
])
4721 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, !compiled
->need_color_unclamp
));
4722 checkGLcall("glClampColorARB");
4724 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
4728 if (vs
->load_local_constsF
)
4729 context
->constant_update_mask
|= WINED3D_SHADER_CONST_VS_F
;
4733 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
4735 priv
->current_vprogram_id
= 0;
4736 gl_info
->gl_ops
.gl
.p_glDisable(GL_VERTEX_PROGRAM_ARB
);
4737 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4739 priv
->vertex_pipe
->vp_enable(gl_info
, TRUE
);
4744 /* Context activation is done by the caller. */
4745 static void shader_arb_disable(void *shader_priv
, struct wined3d_context
*context
)
4747 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4748 struct shader_arb_priv
*priv
= shader_priv
;
4750 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
4752 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4753 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4754 priv
->current_fprogram_id
= 0;
4756 priv
->fragment_pipe
->enable_extension(gl_info
, FALSE
);
4758 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
4760 priv
->current_vprogram_id
= 0;
4761 gl_info
->gl_ops
.gl
.p_glDisable(GL_VERTEX_PROGRAM_ARB
);
4762 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4764 priv
->vertex_pipe
->vp_enable(gl_info
, FALSE
);
4766 if (gl_info
->supported
[ARB_COLOR_BUFFER_FLOAT
] && priv
->last_vs_color_unclamp
)
4768 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, GL_FIXED_ONLY_ARB
));
4769 checkGLcall("glClampColorARB");
4770 priv
->last_vs_color_unclamp
= FALSE
;
4773 context
->shader_update_mask
= (1 << WINED3D_SHADER_TYPE_PIXEL
)
4774 | (1 << WINED3D_SHADER_TYPE_VERTEX
)
4775 | (1 << WINED3D_SHADER_TYPE_GEOMETRY
);
4778 /* Context activation is done by the caller. */
4779 static void shader_arb_select_depth_blt(void *shader_priv
, const struct wined3d_gl_info
*gl_info
,
4780 enum wined3d_gl_resource_type tex_type
, const SIZE
*ds_mask_size
)
4782 const float mask
[] = {0.0f
, 0.0f
, (float)ds_mask_size
->cx
, (float)ds_mask_size
->cy
};
4783 BOOL masked
= ds_mask_size
->cx
&& ds_mask_size
->cy
;
4784 struct shader_arb_priv
*priv
= shader_priv
;
4785 GLuint
*blt_fprogram
;
4787 if (!priv
->depth_blt_vprogram_id
) priv
->depth_blt_vprogram_id
= create_arb_blt_vertex_program(gl_info
);
4788 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->depth_blt_vprogram_id
));
4789 gl_info
->gl_ops
.gl
.p_glEnable(GL_VERTEX_PROGRAM_ARB
);
4791 blt_fprogram
= masked
? &priv
->depth_blt_fprogram_id_masked
[tex_type
] : &priv
->depth_blt_fprogram_id_full
[tex_type
];
4792 if (!*blt_fprogram
) *blt_fprogram
= create_arb_blt_fragment_program(gl_info
, tex_type
, masked
);
4793 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, *blt_fprogram
));
4794 if (masked
) GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, 0, mask
));
4795 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
4798 /* Context activation is done by the caller. */
4799 static void shader_arb_deselect_depth_blt(void *shader_priv
, const struct wined3d_gl_info
*gl_info
)
4801 struct shader_arb_priv
*priv
= shader_priv
;
4803 if (priv
->current_vprogram_id
) {
4804 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->current_vprogram_id
));
4805 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
4807 TRACE("Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB.\n", priv
->current_vprogram_id
);
4811 gl_info
->gl_ops
.gl
.p_glDisable(GL_VERTEX_PROGRAM_ARB
);
4812 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4815 if (priv
->current_fprogram_id
) {
4816 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, priv
->current_fprogram_id
));
4817 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
4819 TRACE("Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB.\n", priv
->current_fprogram_id
);
4821 else if(!priv
->use_arbfp_fixed_func
)
4823 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4824 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4828 static void shader_arb_destroy(struct wined3d_shader
*shader
)
4830 struct wined3d_device
*device
= shader
->device
;
4831 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
4833 if (shader_is_pshader_version(shader
->reg_maps
.shader_version
.type
))
4835 struct arb_pshader_private
*shader_data
= shader
->backend_data
;
4838 if(!shader_data
) return; /* This can happen if a shader was never compiled */
4840 if (shader_data
->num_gl_shaders
)
4842 struct wined3d_context
*context
= context_acquire(device
, NULL
);
4844 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4846 GL_EXTCALL(glDeleteProgramsARB(1, &shader_data
->gl_shaders
[i
].prgId
));
4847 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader_data->gl_shaders[i].prgId))");
4850 context_release(context
);
4853 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
);
4854 HeapFree(GetProcessHeap(), 0, shader_data
);
4855 shader
->backend_data
= NULL
;
4859 struct arb_vshader_private
*shader_data
= shader
->backend_data
;
4862 if(!shader_data
) return; /* This can happen if a shader was never compiled */
4864 if (shader_data
->num_gl_shaders
)
4866 struct wined3d_context
*context
= context_acquire(device
, NULL
);
4868 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4870 GL_EXTCALL(glDeleteProgramsARB(1, &shader_data
->gl_shaders
[i
].prgId
));
4871 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader_data->gl_shaders[i].prgId))");
4874 context_release(context
);
4877 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
);
4878 HeapFree(GetProcessHeap(), 0, shader_data
);
4879 shader
->backend_data
= NULL
;
4883 static int sig_tree_compare(const void *key
, const struct wine_rb_entry
*entry
)
4885 struct ps_signature
*e
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
4886 return compare_sig(key
, &e
->sig
);
4889 static const struct wine_rb_functions sig_tree_functions
=
4897 static HRESULT
shader_arb_alloc(struct wined3d_device
*device
, const struct wined3d_vertex_pipe_ops
*vertex_pipe
,
4898 const struct fragment_pipeline
*fragment_pipe
)
4900 struct shader_arb_priv
*priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*priv
));
4901 struct fragment_caps fragment_caps
;
4902 void *vertex_priv
, *fragment_priv
;
4903 const struct wined3d_d3d_info
*d3d_info
= &device
->adapter
->d3d_info
;
4905 if (!(vertex_priv
= vertex_pipe
->vp_alloc(&arb_program_shader_backend
, priv
)))
4907 ERR("Failed to initialize vertex pipe.\n");
4908 HeapFree(GetProcessHeap(), 0, priv
);
4912 if (!(fragment_priv
= fragment_pipe
->alloc_private(&arb_program_shader_backend
, priv
)))
4914 ERR("Failed to initialize fragment pipe.\n");
4915 vertex_pipe
->vp_free(device
);
4916 HeapFree(GetProcessHeap(), 0, priv
);
4920 priv
->vshader_const_dirty
= HeapAlloc(GetProcessHeap(), 0,
4921 sizeof(*priv
->vshader_const_dirty
) * d3d_info
->limits
.vs_uniform_count
);
4922 if (!priv
->vshader_const_dirty
)
4924 memset(priv
->vshader_const_dirty
, 1,
4925 sizeof(*priv
->vshader_const_dirty
) * d3d_info
->limits
.vs_uniform_count
);
4927 priv
->pshader_const_dirty
= HeapAlloc(GetProcessHeap(), 0,
4928 sizeof(*priv
->pshader_const_dirty
) * d3d_info
->limits
.ps_uniform_count
);
4929 if (!priv
->pshader_const_dirty
)
4931 memset(priv
->pshader_const_dirty
, 1,
4932 sizeof(*priv
->pshader_const_dirty
) * d3d_info
->limits
.ps_uniform_count
);
4934 if(wine_rb_init(&priv
->signature_tree
, &sig_tree_functions
) == -1)
4936 ERR("RB tree init failed\n");
4940 priv
->vertex_pipe
= vertex_pipe
;
4941 priv
->fragment_pipe
= fragment_pipe
;
4942 fragment_pipe
->get_caps(&device
->adapter
->gl_info
, &fragment_caps
);
4943 priv
->ffp_proj_control
= fragment_caps
.wined3d_caps
& WINED3D_FRAGMENT_CAP_PROJ_CONTROL
;
4945 device
->vertex_priv
= vertex_priv
;
4946 device
->fragment_priv
= fragment_priv
;
4947 device
->shader_priv
= priv
;
4952 HeapFree(GetProcessHeap(), 0, priv
->pshader_const_dirty
);
4953 HeapFree(GetProcessHeap(), 0, priv
->vshader_const_dirty
);
4954 fragment_pipe
->free_private(device
);
4955 vertex_pipe
->vp_free(device
);
4956 HeapFree(GetProcessHeap(), 0, priv
);
4957 return E_OUTOFMEMORY
;
4960 static void release_signature(struct wine_rb_entry
*entry
, void *context
)
4962 struct ps_signature
*sig
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
4965 for (i
= 0; i
< sig
->sig
.element_count
; ++i
)
4967 HeapFree(GetProcessHeap(), 0, (char *)sig
->sig
.elements
[i
].semantic_name
);
4969 HeapFree(GetProcessHeap(), 0, sig
->sig
.elements
);
4970 HeapFree(GetProcessHeap(), 0, sig
);
4973 /* Context activation is done by the caller. */
4974 static void shader_arb_free(struct wined3d_device
*device
)
4976 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
4977 struct shader_arb_priv
*priv
= device
->shader_priv
;
4980 if (priv
->depth_blt_vprogram_id
)
4981 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_vprogram_id
));
4983 for (i
= 0; i
< WINED3D_GL_RES_TYPE_COUNT
; ++i
)
4985 if (priv
->depth_blt_fprogram_id_full
[i
])
4987 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_fprogram_id_full
[i
]));
4989 if (priv
->depth_blt_fprogram_id_masked
[i
])
4991 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_fprogram_id_masked
[i
]));
4995 wine_rb_destroy(&priv
->signature_tree
, release_signature
, NULL
);
4996 HeapFree(GetProcessHeap(), 0, priv
->pshader_const_dirty
);
4997 HeapFree(GetProcessHeap(), 0, priv
->vshader_const_dirty
);
4998 priv
->fragment_pipe
->free_private(device
);
4999 priv
->vertex_pipe
->vp_free(device
);
5000 HeapFree(GetProcessHeap(), 0, device
->shader_priv
);
5003 static BOOL
shader_arb_allocate_context_data(struct wined3d_context
*context
)
5008 static void shader_arb_free_context_data(struct wined3d_context
*context
)
5010 struct shader_arb_priv
*priv
;
5012 if (!context
->swapchain
)
5015 priv
= context
->swapchain
->device
->shader_priv
;
5016 if (priv
->last_context
== context
)
5017 priv
->last_context
= NULL
;
5020 static void shader_arb_get_caps(const struct wined3d_gl_info
*gl_info
, struct shader_caps
*caps
)
5022 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
5027 /* 96 is the minimum allowed value of MAX_PROGRAM_ENV_PARAMETERS_ARB
5028 * for vertex programs. If the native limit is less than that it's
5029 * not very useful, and e.g. Mesa swrast returns 0, probably to
5030 * indicate it's a software implementation. */
5031 if (gl_info
->limits
.arb_vs_native_constants
< 96)
5032 vs_consts
= gl_info
->limits
.arb_vs_float_constants
;
5034 vs_consts
= min(gl_info
->limits
.arb_vs_float_constants
, gl_info
->limits
.arb_vs_native_constants
);
5036 if (gl_info
->supported
[NV_VERTEX_PROGRAM3
])
5039 TRACE("Hardware vertex shader version 3.0 enabled (NV_VERTEX_PROGRAM3)\n");
5041 else if (vs_consts
>= 256)
5043 /* Shader Model 2.0 requires at least 256 vertex shader constants */
5045 TRACE("Hardware vertex shader version 2.0 enabled (ARB_PROGRAM)\n");
5050 TRACE("Hardware vertex shader version 1.1 enabled (ARB_PROGRAM)\n");
5052 caps
->vs_version
= min(wined3d_settings
.max_sm_vs
, vs_version
);
5053 caps
->vs_uniform_count
= vs_consts
;
5057 caps
->vs_version
= 0;
5058 caps
->vs_uniform_count
= 0;
5061 caps
->gs_version
= 0;
5063 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
5068 /* Similar as above for vertex programs, but the minimum for fragment
5069 * programs is 24. */
5070 if (gl_info
->limits
.arb_ps_native_constants
< 24)
5071 ps_consts
= gl_info
->limits
.arb_ps_float_constants
;
5073 ps_consts
= min(gl_info
->limits
.arb_ps_float_constants
, gl_info
->limits
.arb_ps_native_constants
);
5075 if (gl_info
->supported
[NV_FRAGMENT_PROGRAM2
])
5078 TRACE("Hardware pixel shader version 3.0 enabled (NV_FRAGMENT_PROGRAM2)\n");
5080 else if (ps_consts
>= 32)
5082 /* Shader Model 2.0 requires at least 32 pixel shader constants */
5084 TRACE("Hardware pixel shader version 2.0 enabled (ARB_PROGRAM)\n");
5089 TRACE("Hardware pixel shader version 1.4 enabled (ARB_PROGRAM)\n");
5091 caps
->ps_version
= min(wined3d_settings
.max_sm_ps
, ps_version
);
5092 caps
->ps_uniform_count
= ps_consts
;
5093 caps
->ps_1x_max_value
= 8.0f
;
5097 caps
->ps_version
= 0;
5098 caps
->ps_uniform_count
= 0;
5099 caps
->ps_1x_max_value
= 0.0f
;
5102 caps
->wined3d_caps
= WINED3D_SHADER_CAP_SRGB_WRITE
;
5103 if (use_nv_clip(gl_info
))
5104 caps
->wined3d_caps
|= WINED3D_SHADER_CAP_VS_CLIPPING
;
5107 static BOOL
shader_arb_color_fixup_supported(struct color_fixup_desc fixup
)
5109 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
5111 TRACE("Checking support for color_fixup:\n");
5112 dump_color_fixup_desc(fixup
);
5115 /* We support everything except complex conversions. */
5116 if (!is_complex_fixup(fixup
))
5122 TRACE("[FAILED]\n");
5126 static void shader_arb_add_instruction_modifiers(const struct wined3d_shader_instruction
*ins
) {
5128 char write_mask
[20], regstr
[50];
5129 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
5130 BOOL is_color
= FALSE
;
5131 const struct wined3d_shader_dst_param
*dst
;
5133 if (!ins
->dst_count
) return;
5137 if (!shift
) return; /* Saturate alone is handled by the instructions */
5139 shader_arb_get_write_mask(ins
, dst
, write_mask
);
5140 shader_arb_get_register_name(ins
, &dst
->reg
, regstr
, &is_color
);
5142 /* Generate a line that does the output modifier computation
5143 * FIXME: _SAT vs shift? _SAT alone is already handled in the instructions, if this
5144 * maps problems in e.g. _d4_sat modify shader_arb_get_modifier
5146 shader_addline(buffer
, "MUL%s %s%s, %s, %s;\n", shader_arb_get_modifier(ins
),
5147 regstr
, write_mask
, regstr
, shift_tab
[shift
]);
5150 static const SHADER_HANDLER shader_arb_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
5152 /* WINED3DSIH_ABS */ shader_hw_map2gl
,
5153 /* WINED3DSIH_ADD */ shader_hw_map2gl
,
5154 /* WINED3DSIH_AND */ NULL
,
5155 /* WINED3DSIH_BEM */ pshader_hw_bem
,
5156 /* WINED3DSIH_BREAK */ shader_hw_break
,
5157 /* WINED3DSIH_BREAKC */ shader_hw_breakc
,
5158 /* WINED3DSIH_BREAKP */ NULL
,
5159 /* WINED3DSIH_CALL */ shader_hw_call
,
5160 /* WINED3DSIH_CALLNZ */ NULL
,
5161 /* WINED3DSIH_CMP */ pshader_hw_cmp
,
5162 /* WINED3DSIH_CND */ pshader_hw_cnd
,
5163 /* WINED3DSIH_CRS */ shader_hw_map2gl
,
5164 /* WINED3DSIH_CUT */ NULL
,
5165 /* WINED3DSIH_DCL */ shader_hw_nop
,
5166 /* WINED3DSIH_DCL_CONSTANT_BUFFER */ shader_hw_nop
,
5167 /* WINED3DSIH_DCL_INPUT_PRIMITIVE */ shader_hw_nop
,
5168 /* WINED3DSIH_DCL_OUTPUT_TOPOLOGY */ shader_hw_nop
,
5169 /* WINED3DSIH_DCL_VERTICES_OUT */ shader_hw_nop
,
5170 /* WINED3DSIH_DEF */ shader_hw_nop
,
5171 /* WINED3DSIH_DEFB */ shader_hw_nop
,
5172 /* WINED3DSIH_DEFI */ shader_hw_nop
,
5173 /* WINED3DSIH_DIV */ NULL
,
5174 /* WINED3DSIH_DP2 */ NULL
,
5175 /* WINED3DSIH_DP2ADD */ pshader_hw_dp2add
,
5176 /* WINED3DSIH_DP3 */ shader_hw_map2gl
,
5177 /* WINED3DSIH_DP4 */ shader_hw_map2gl
,
5178 /* WINED3DSIH_DST */ shader_hw_map2gl
,
5179 /* WINED3DSIH_DSX */ shader_hw_map2gl
,
5180 /* WINED3DSIH_DSY */ shader_hw_dsy
,
5181 /* WINED3DSIH_ELSE */ shader_hw_else
,
5182 /* WINED3DSIH_EMIT */ NULL
,
5183 /* WINED3DSIH_ENDIF */ shader_hw_endif
,
5184 /* WINED3DSIH_ENDLOOP */ shader_hw_endloop
,
5185 /* WINED3DSIH_ENDREP */ shader_hw_endrep
,
5186 /* WINED3DSIH_EQ */ NULL
,
5187 /* WINED3DSIH_EXP */ shader_hw_scalar_op
,
5188 /* WINED3DSIH_EXPP */ shader_hw_scalar_op
,
5189 /* WINED3DSIH_FRC */ shader_hw_map2gl
,
5190 /* WINED3DSIH_FTOI */ NULL
,
5191 /* WINED3DSIH_GE */ NULL
,
5192 /* WINED3DSIH_IADD */ NULL
,
5193 /* WINED3DSIH_IEQ */ NULL
,
5194 /* WINED3DSIH_IF */ NULL
/* Hardcoded into the shader */,
5195 /* WINED3DSIH_IFC */ shader_hw_ifc
,
5196 /* WINED3DSIH_IGE */ NULL
,
5197 /* WINED3DSIH_IMUL */ NULL
,
5198 /* WINED3DSIH_ISHL */ NULL
,
5199 /* WINED3DSIH_ITOF */ NULL
,
5200 /* WINED3DSIH_LABEL */ shader_hw_label
,
5201 /* WINED3DSIH_LD */ NULL
,
5202 /* WINED3DSIH_LIT */ shader_hw_map2gl
,
5203 /* WINED3DSIH_LOG */ shader_hw_scalar_op
,
5204 /* WINED3DSIH_LOGP */ shader_hw_scalar_op
,
5205 /* WINED3DSIH_LOOP */ shader_hw_loop
,
5206 /* WINED3DSIH_LRP */ shader_hw_lrp
,
5207 /* WINED3DSIH_LT */ NULL
,
5208 /* WINED3DSIH_M3x2 */ shader_hw_mnxn
,
5209 /* WINED3DSIH_M3x3 */ shader_hw_mnxn
,
5210 /* WINED3DSIH_M3x4 */ shader_hw_mnxn
,
5211 /* WINED3DSIH_M4x3 */ shader_hw_mnxn
,
5212 /* WINED3DSIH_M4x4 */ shader_hw_mnxn
,
5213 /* WINED3DSIH_MAD */ shader_hw_map2gl
,
5214 /* WINED3DSIH_MAX */ shader_hw_map2gl
,
5215 /* WINED3DSIH_MIN */ shader_hw_map2gl
,
5216 /* WINED3DSIH_MOV */ shader_hw_mov
,
5217 /* WINED3DSIH_MOVA */ shader_hw_mov
,
5218 /* WINED3DSIH_MOVC */ NULL
,
5219 /* WINED3DSIH_MUL */ shader_hw_map2gl
,
5220 /* WINED3DSIH_NE */ NULL
,
5221 /* WINED3DSIH_NOP */ shader_hw_nop
,
5222 /* WINED3DSIH_NRM */ shader_hw_nrm
,
5223 /* WINED3DSIH_OR */ NULL
,
5224 /* WINED3DSIH_PHASE */ shader_hw_nop
,
5225 /* WINED3DSIH_POW */ shader_hw_pow
,
5226 /* WINED3DSIH_RCP */ shader_hw_scalar_op
,
5227 /* WINED3DSIH_REP */ shader_hw_rep
,
5228 /* WINED3DSIH_RET */ shader_hw_ret
,
5229 /* WINED3DSIH_ROUND_NI */ NULL
,
5230 /* WINED3DSIH_RSQ */ shader_hw_scalar_op
,
5231 /* WINED3DSIH_SAMPLE */ NULL
,
5232 /* WINED3DSIH_SAMPLE_GRAD */ NULL
,
5233 /* WINED3DSIH_SAMPLE_LOD */ NULL
,
5234 /* WINED3DSIH_SETP */ NULL
,
5235 /* WINED3DSIH_SGE */ shader_hw_map2gl
,
5236 /* WINED3DSIH_SGN */ shader_hw_sgn
,
5237 /* WINED3DSIH_SINCOS */ shader_hw_sincos
,
5238 /* WINED3DSIH_SLT */ shader_hw_map2gl
,
5239 /* WINED3DSIH_SQRT */ NULL
,
5240 /* WINED3DSIH_SUB */ shader_hw_map2gl
,
5241 /* WINED3DSIH_TEX */ pshader_hw_tex
,
5242 /* WINED3DSIH_TEXBEM */ pshader_hw_texbem
,
5243 /* WINED3DSIH_TEXBEML */ pshader_hw_texbem
,
5244 /* WINED3DSIH_TEXCOORD */ pshader_hw_texcoord
,
5245 /* WINED3DSIH_TEXDEPTH */ pshader_hw_texdepth
,
5246 /* WINED3DSIH_TEXDP3 */ pshader_hw_texdp3
,
5247 /* WINED3DSIH_TEXDP3TEX */ pshader_hw_texdp3tex
,
5248 /* WINED3DSIH_TEXKILL */ pshader_hw_texkill
,
5249 /* WINED3DSIH_TEXLDD */ shader_hw_texldd
,
5250 /* WINED3DSIH_TEXLDL */ shader_hw_texldl
,
5251 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_hw_texm3x2depth
,
5252 /* WINED3DSIH_TEXM3x2PAD */ pshader_hw_texm3x2pad
,
5253 /* WINED3DSIH_TEXM3x2TEX */ pshader_hw_texm3x2tex
,
5254 /* WINED3DSIH_TEXM3x3 */ pshader_hw_texm3x3
,
5255 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
5256 /* WINED3DSIH_TEXM3x3PAD */ pshader_hw_texm3x3pad
,
5257 /* WINED3DSIH_TEXM3x3SPEC */ pshader_hw_texm3x3spec
,
5258 /* WINED3DSIH_TEXM3x3TEX */ pshader_hw_texm3x3tex
,
5259 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_hw_texm3x3vspec
,
5260 /* WINED3DSIH_TEXREG2AR */ pshader_hw_texreg2ar
,
5261 /* WINED3DSIH_TEXREG2GB */ pshader_hw_texreg2gb
,
5262 /* WINED3DSIH_TEXREG2RGB */ pshader_hw_texreg2rgb
,
5263 /* WINED3DSIH_UDIV */ NULL
,
5264 /* WINED3DSIH_UGE */ NULL
,
5265 /* WINED3DSIH_USHR */ NULL
,
5266 /* WINED3DSIH_UTOF */ NULL
,
5267 /* WINED3DSIH_XOR */ NULL
,
5270 static BOOL
get_bool_const(const struct wined3d_shader_instruction
*ins
,
5271 const struct wined3d_shader
*shader
, DWORD idx
)
5273 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
5274 BOOL vshader
= shader_is_vshader_version(reg_maps
->shader_version
.type
);
5275 const struct wined3d_shader_lconst
*constant
;
5277 WORD flag
= (1 << idx
);
5278 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5280 if (reg_maps
->local_bool_consts
& flag
)
5282 /* What good is an if(bool) with a hardcoded local constant? I don't know, but handle it */
5283 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsB
, struct wined3d_shader_lconst
, entry
)
5285 if (constant
->idx
== idx
)
5287 return constant
->value
[0];
5290 ERR("Local constant not found\n");
5295 if(vshader
) bools
= priv
->cur_vs_args
->clip
.boolclip
.bools
;
5296 else bools
= priv
->cur_ps_args
->bools
;
5297 return bools
& flag
;
5301 static void get_loop_control_const(const struct wined3d_shader_instruction
*ins
,
5302 const struct wined3d_shader
*shader
, UINT idx
, struct wined3d_shader_loop_control
*loop_control
)
5304 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
5305 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5307 /* Integer constants can either be a local constant, or they can be stored in the shader
5308 * type specific compile args. */
5309 if (reg_maps
->local_int_consts
& (1 << idx
))
5311 const struct wined3d_shader_lconst
*constant
;
5313 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsI
, struct wined3d_shader_lconst
, entry
)
5315 if (constant
->idx
== idx
)
5317 loop_control
->count
= constant
->value
[0];
5318 loop_control
->start
= constant
->value
[1];
5319 /* Step is signed. */
5320 loop_control
->step
= (int)constant
->value
[2];
5324 /* If this happens the flag was set incorrectly */
5325 ERR("Local constant not found\n");
5326 loop_control
->count
= 0;
5327 loop_control
->start
= 0;
5328 loop_control
->step
= 0;
5332 switch (reg_maps
->shader_version
.type
)
5334 case WINED3D_SHADER_TYPE_VERTEX
:
5335 /* Count and aL start value are unsigned */
5336 loop_control
->count
= priv
->cur_vs_args
->loop_ctrl
[idx
][0];
5337 loop_control
->start
= priv
->cur_vs_args
->loop_ctrl
[idx
][1];
5338 /* Step is signed. */
5339 loop_control
->step
= ((char)priv
->cur_vs_args
->loop_ctrl
[idx
][2]);
5342 case WINED3D_SHADER_TYPE_PIXEL
:
5343 loop_control
->count
= priv
->cur_ps_args
->loop_ctrl
[idx
][0];
5344 loop_control
->start
= priv
->cur_ps_args
->loop_ctrl
[idx
][1];
5345 loop_control
->step
= ((char)priv
->cur_ps_args
->loop_ctrl
[idx
][2]);
5349 FIXME("Unhandled shader type %#x.\n", reg_maps
->shader_version
.type
);
5354 static void record_instruction(struct list
*list
, const struct wined3d_shader_instruction
*ins
)
5357 struct wined3d_shader_dst_param
*dst_param
= NULL
;
5358 struct wined3d_shader_src_param
*src_param
= NULL
, *rel_addr
= NULL
;
5359 struct recorded_instruction
*rec
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*rec
));
5362 ERR("Out of memory\n");
5367 dst_param
= HeapAlloc(GetProcessHeap(), 0, sizeof(*dst_param
));
5368 if(!dst_param
) goto free
;
5369 *dst_param
= *ins
->dst
;
5370 if (ins
->dst
->reg
.idx
[0].rel_addr
)
5372 rel_addr
= HeapAlloc(GetProcessHeap(), 0, sizeof(*rel_addr
));
5375 *rel_addr
= *ins
->dst
->reg
.idx
[0].rel_addr
;
5376 dst_param
->reg
.idx
[0].rel_addr
= rel_addr
;
5378 rec
->ins
.dst
= dst_param
;
5380 src_param
= HeapAlloc(GetProcessHeap(), 0, sizeof(*src_param
) * ins
->src_count
);
5383 for (i
= 0; i
< ins
->src_count
; ++i
)
5385 src_param
[i
] = ins
->src
[i
];
5386 if (ins
->src
[i
].reg
.idx
[0].rel_addr
)
5388 rel_addr
= HeapAlloc(GetProcessHeap(), 0, sizeof(*rel_addr
));
5391 *rel_addr
= *ins
->src
[i
].reg
.idx
[0].rel_addr
;
5392 src_param
[i
].reg
.idx
[0].rel_addr
= rel_addr
;
5395 rec
->ins
.src
= src_param
;
5396 list_add_tail(list
, &rec
->entry
);
5400 ERR("Out of memory\n");
5403 HeapFree(GetProcessHeap(), 0, (void *)dst_param
->reg
.idx
[0].rel_addr
);
5404 HeapFree(GetProcessHeap(), 0, dst_param
);
5408 for(i
= 0; i
< ins
->src_count
; i
++)
5410 HeapFree(GetProcessHeap(), 0, (void *)src_param
[i
].reg
.idx
[0].rel_addr
);
5412 HeapFree(GetProcessHeap(), 0, src_param
);
5414 HeapFree(GetProcessHeap(), 0, rec
);
5417 static void free_recorded_instruction(struct list
*list
)
5419 struct recorded_instruction
*rec_ins
, *entry2
;
5422 LIST_FOR_EACH_ENTRY_SAFE(rec_ins
, entry2
, list
, struct recorded_instruction
, entry
)
5424 list_remove(&rec_ins
->entry
);
5425 if (rec_ins
->ins
.dst
)
5427 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.dst
->reg
.idx
[0].rel_addr
);
5428 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.dst
);
5430 if (rec_ins
->ins
.src
)
5432 for (i
= 0; i
< rec_ins
->ins
.src_count
; ++i
)
5434 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.src
[i
].reg
.idx
[0].rel_addr
);
5436 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.src
);
5438 HeapFree(GetProcessHeap(), 0, rec_ins
);
5442 static void pop_control_frame(const struct wined3d_shader_instruction
*ins
)
5444 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5445 struct control_frame
*control_frame
;
5447 if (ins
->handler_idx
== WINED3DSIH_ENDLOOP
|| ins
->handler_idx
== WINED3DSIH_ENDREP
)
5449 struct list
*e
= list_head(&priv
->control_frames
);
5450 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5451 list_remove(&control_frame
->entry
);
5452 HeapFree(GetProcessHeap(), 0, control_frame
);
5455 else if (ins
->handler_idx
== WINED3DSIH_ENDIF
)
5457 /* Non-ifc ENDIFs were already handled previously. */
5458 struct list
*e
= list_head(&priv
->control_frames
);
5459 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5460 list_remove(&control_frame
->entry
);
5461 HeapFree(GetProcessHeap(), 0, control_frame
);
5465 static void shader_arb_handle_instruction(const struct wined3d_shader_instruction
*ins
) {
5466 SHADER_HANDLER hw_fct
;
5467 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5468 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
5469 struct control_frame
*control_frame
;
5470 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
5473 if(ins
->handler_idx
== WINED3DSIH_LOOP
|| ins
->handler_idx
== WINED3DSIH_REP
)
5475 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
5476 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5478 if(ins
->handler_idx
== WINED3DSIH_LOOP
) control_frame
->type
= LOOP
;
5479 if(ins
->handler_idx
== WINED3DSIH_REP
) control_frame
->type
= REP
;
5481 if(priv
->target_version
>= NV2
)
5483 control_frame
->no
.loop
= priv
->num_loops
++;
5488 /* Don't bother recording when we're in a not used if branch */
5494 if(!priv
->recording
)
5496 list_init(&priv
->record
);
5497 priv
->recording
= TRUE
;
5498 control_frame
->outer_loop
= TRUE
;
5499 get_loop_control_const(ins
, shader
, ins
->src
[0].reg
.idx
[0].offset
, &control_frame
->loop_control
);
5500 return; /* Instruction is handled */
5502 /* Record this loop in the outer loop's recording */
5505 else if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
|| ins
->handler_idx
== WINED3DSIH_ENDREP
)
5507 if(priv
->target_version
>= NV2
)
5509 /* Nothing to do. The control frame is popped after the HW instr handler */
5513 struct list
*e
= list_head(&priv
->control_frames
);
5514 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5515 list_remove(&control_frame
->entry
);
5517 if(control_frame
->outer_loop
)
5519 unsigned int iteration
;
5523 /* Turn off recording before playback */
5524 priv
->recording
= FALSE
;
5526 /* Move the recorded instructions to a separate list and get them out of the private data
5527 * structure. If there are nested loops, the shader_arb_handle_instruction below will
5528 * be recorded again, thus priv->record might be overwritten
5531 list_move_tail(©
, &priv
->record
);
5532 list_init(&priv
->record
);
5534 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5536 shader_addline(buffer
, "#unrolling loop: %u iterations, aL=%u, inc %d\n",
5537 control_frame
->loop_control
.count
, control_frame
->loop_control
.start
,
5538 control_frame
->loop_control
.step
);
5539 aL
= control_frame
->loop_control
.start
;
5543 shader_addline(buffer
, "#unrolling rep: %u iterations\n", control_frame
->loop_control
.count
);
5546 for (iteration
= 0; iteration
< control_frame
->loop_control
.count
; ++iteration
)
5548 struct recorded_instruction
*rec_ins
;
5549 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5552 shader_addline(buffer
, "#Iteration %u, aL=%d\n", iteration
, aL
);
5556 shader_addline(buffer
, "#Iteration %u\n", iteration
);
5559 LIST_FOR_EACH_ENTRY(rec_ins
, ©
, struct recorded_instruction
, entry
)
5561 shader_arb_handle_instruction(&rec_ins
->ins
);
5564 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5566 aL
+= control_frame
->loop_control
.step
;
5569 shader_addline(buffer
, "#end loop/rep\n");
5571 free_recorded_instruction(©
);
5572 HeapFree(GetProcessHeap(), 0, control_frame
);
5573 return; /* Instruction is handled */
5577 /* This is a nested loop. Proceed to the normal recording function */
5578 HeapFree(GetProcessHeap(), 0, control_frame
);
5585 record_instruction(&priv
->record
, ins
);
5590 if(ins
->handler_idx
== WINED3DSIH_IF
)
5592 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
5593 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5594 control_frame
->type
= IF
;
5596 bool_const
= get_bool_const(ins
, shader
, ins
->src
[0].reg
.idx
[0].offset
);
5597 if (ins
->src
[0].modifiers
== WINED3DSPSM_NOT
)
5598 bool_const
= !bool_const
;
5599 if (!priv
->muted
&& !bool_const
)
5601 shader_addline(buffer
, "#if(FALSE){\n");
5603 control_frame
->muting
= TRUE
;
5605 else shader_addline(buffer
, "#if(TRUE) {\n");
5607 return; /* Instruction is handled */
5609 else if(ins
->handler_idx
== WINED3DSIH_IFC
)
5611 /* IF(bool) and if_cond(a, b) use the same ELSE and ENDIF tokens */
5612 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
5613 control_frame
->type
= IFC
;
5614 control_frame
->no
.ifc
= priv
->num_ifcs
++;
5615 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5617 else if(ins
->handler_idx
== WINED3DSIH_ELSE
)
5619 struct list
*e
= list_head(&priv
->control_frames
);
5620 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5622 if(control_frame
->type
== IF
)
5624 shader_addline(buffer
, "#} else {\n");
5625 if(!priv
->muted
&& !control_frame
->muting
)
5628 control_frame
->muting
= TRUE
;
5630 else if(control_frame
->muting
) priv
->muted
= FALSE
;
5631 return; /* Instruction is handled. */
5633 /* In case of an ifc, generate a HW shader instruction */
5634 if (control_frame
->type
!= IFC
)
5635 ERR("Control frame does not match.\n");
5637 else if(ins
->handler_idx
== WINED3DSIH_ENDIF
)
5639 struct list
*e
= list_head(&priv
->control_frames
);
5640 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5642 if(control_frame
->type
== IF
)
5644 shader_addline(buffer
, "#} endif\n");
5645 if(control_frame
->muting
) priv
->muted
= FALSE
;
5646 list_remove(&control_frame
->entry
);
5647 HeapFree(GetProcessHeap(), 0, control_frame
);
5648 return; /* Instruction is handled */
5650 /* In case of an ifc, generate a HW shader instruction */
5651 if (control_frame
->type
!= IFC
)
5652 ERR("Control frame does not match.\n");
5657 pop_control_frame(ins
);
5661 /* Select handler */
5662 hw_fct
= shader_arb_instruction_handler_table
[ins
->handler_idx
];
5664 /* Unhandled opcode */
5667 FIXME("Backend can't handle opcode %#x\n", ins
->handler_idx
);
5672 pop_control_frame(ins
);
5674 shader_arb_add_instruction_modifiers(ins
);
5677 static BOOL
shader_arb_has_ffp_proj_control(void *shader_priv
)
5679 struct shader_arb_priv
*priv
= shader_priv
;
5681 return priv
->ffp_proj_control
;
5684 const struct wined3d_shader_backend_ops arb_program_shader_backend
=
5686 shader_arb_handle_instruction
,
5689 shader_arb_select_depth_blt
,
5690 shader_arb_deselect_depth_blt
,
5691 shader_arb_update_float_vertex_constants
,
5692 shader_arb_update_float_pixel_constants
,
5693 shader_arb_load_constants
,
5697 shader_arb_allocate_context_data
,
5698 shader_arb_free_context_data
,
5699 shader_arb_get_caps
,
5700 shader_arb_color_fixup_supported
,
5701 shader_arb_has_ffp_proj_control
,
5704 /* ARB_fragment_program fixed function pipeline replacement definitions */
5705 #define ARB_FFP_CONST_TFACTOR 0
5706 #define ARB_FFP_CONST_COLOR_KEY ((ARB_FFP_CONST_TFACTOR) + 1)
5707 #define ARB_FFP_CONST_SPECULAR_ENABLE ((ARB_FFP_CONST_COLOR_KEY) + 1)
5708 #define ARB_FFP_CONST_CONSTANT(i) ((ARB_FFP_CONST_SPECULAR_ENABLE) + 1 + i)
5709 #define ARB_FFP_CONST_BUMPMAT(i) ((ARB_FFP_CONST_CONSTANT(7)) + 1 + i)
5710 #define ARB_FFP_CONST_LUMINANCE(i) ((ARB_FFP_CONST_BUMPMAT(7)) + 1 + i)
5712 struct arbfp_ffp_desc
5714 struct ffp_frag_desc parent
;
5718 /* Context activation is done by the caller. */
5719 static void arbfp_enable(const struct wined3d_gl_info
*gl_info
, BOOL enable
)
5723 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
5724 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
5728 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
5729 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
5733 static void *arbfp_alloc(const struct wined3d_shader_backend_ops
*shader_backend
, void *shader_priv
)
5735 struct shader_arb_priv
*priv
;
5737 /* Share private data between the shader backend and the pipeline
5738 * replacement, if both are the arb implementation. This is needed to
5739 * figure out whether ARBfp should be disabled if no pixel shader is bound
5741 if (shader_backend
== &arb_program_shader_backend
)
5743 else if (!(priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*priv
))))
5746 if (wine_rb_init(&priv
->fragment_shaders
, &wined3d_ffp_frag_program_rb_functions
) == -1)
5748 ERR("Failed to initialize rbtree.\n");
5749 if (priv
!= shader_priv
)
5750 HeapFree(GetProcessHeap(), 0, priv
);
5753 priv
->use_arbfp_fixed_func
= TRUE
;
5758 /* Context activation is done by the caller. */
5759 static void arbfp_free_ffpshader(struct wine_rb_entry
*entry
, void *context
)
5761 const struct wined3d_gl_info
*gl_info
= context
;
5762 struct arbfp_ffp_desc
*entry_arb
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_ffp_desc
, parent
.entry
);
5764 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb
->shader
));
5765 checkGLcall("glDeleteProgramsARB(1, &entry_arb->shader)");
5766 HeapFree(GetProcessHeap(), 0, entry_arb
);
5769 /* Context activation is done by the caller. */
5770 static void arbfp_free(struct wined3d_device
*device
)
5772 struct shader_arb_priv
*priv
= device
->fragment_priv
;
5774 wine_rb_destroy(&priv
->fragment_shaders
, arbfp_free_ffpshader
, &device
->adapter
->gl_info
);
5775 priv
->use_arbfp_fixed_func
= FALSE
;
5777 if (device
->shader_backend
!= &arb_program_shader_backend
)
5779 HeapFree(GetProcessHeap(), 0, device
->fragment_priv
);
5783 static void arbfp_get_caps(const struct wined3d_gl_info
*gl_info
, struct fragment_caps
*caps
)
5785 caps
->wined3d_caps
= WINED3D_FRAGMENT_CAP_PROJ_CONTROL
5786 | WINED3D_FRAGMENT_CAP_SRGB_WRITE
5787 | WINED3D_FRAGMENT_CAP_COLOR_KEY
;
5788 caps
->PrimitiveMiscCaps
= WINED3DPMISCCAPS_TSSARGTEMP
;
5789 caps
->TextureOpCaps
= WINED3DTEXOPCAPS_DISABLE
|
5790 WINED3DTEXOPCAPS_SELECTARG1
|
5791 WINED3DTEXOPCAPS_SELECTARG2
|
5792 WINED3DTEXOPCAPS_MODULATE4X
|
5793 WINED3DTEXOPCAPS_MODULATE2X
|
5794 WINED3DTEXOPCAPS_MODULATE
|
5795 WINED3DTEXOPCAPS_ADDSIGNED2X
|
5796 WINED3DTEXOPCAPS_ADDSIGNED
|
5797 WINED3DTEXOPCAPS_ADD
|
5798 WINED3DTEXOPCAPS_SUBTRACT
|
5799 WINED3DTEXOPCAPS_ADDSMOOTH
|
5800 WINED3DTEXOPCAPS_BLENDCURRENTALPHA
|
5801 WINED3DTEXOPCAPS_BLENDFACTORALPHA
|
5802 WINED3DTEXOPCAPS_BLENDTEXTUREALPHA
|
5803 WINED3DTEXOPCAPS_BLENDDIFFUSEALPHA
|
5804 WINED3DTEXOPCAPS_BLENDTEXTUREALPHAPM
|
5805 WINED3DTEXOPCAPS_MODULATEALPHA_ADDCOLOR
|
5806 WINED3DTEXOPCAPS_MODULATECOLOR_ADDALPHA
|
5807 WINED3DTEXOPCAPS_MODULATEINVCOLOR_ADDALPHA
|
5808 WINED3DTEXOPCAPS_MODULATEINVALPHA_ADDCOLOR
|
5809 WINED3DTEXOPCAPS_DOTPRODUCT3
|
5810 WINED3DTEXOPCAPS_MULTIPLYADD
|
5811 WINED3DTEXOPCAPS_LERP
|
5812 WINED3DTEXOPCAPS_BUMPENVMAP
|
5813 WINED3DTEXOPCAPS_BUMPENVMAPLUMINANCE
;
5815 /* TODO: Implement WINED3DTEXOPCAPS_PREMODULATE */
5817 caps
->MaxTextureBlendStages
= 8;
5818 caps
->MaxSimultaneousTextures
= min(gl_info
->limits
.fragment_samplers
, 8);
5821 static void state_texfactor_arbfp(struct wined3d_context
*context
,
5822 const struct wined3d_state
*state
, DWORD state_id
)
5824 struct wined3d_device
*device
= context
->swapchain
->device
;
5825 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5828 /* Don't load the parameter if we're using an arbfp pixel shader,
5829 * otherwise we'll overwrite application provided constants. */
5830 if (device
->shader_backend
== &arb_program_shader_backend
)
5832 struct shader_arb_priv
*priv
;
5834 if (use_ps(state
)) return;
5836 priv
= device
->shader_priv
;
5837 priv
->pshader_const_dirty
[ARB_FFP_CONST_TFACTOR
] = 1;
5838 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_TFACTOR
+ 1);
5841 D3DCOLORTOGLFLOAT4(state
->render_states
[WINED3D_RS_TEXTUREFACTOR
], col
);
5842 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_TFACTOR
, col
));
5843 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_TFACTOR, col)");
5846 static void state_arb_specularenable(struct wined3d_context
*context
,
5847 const struct wined3d_state
*state
, DWORD state_id
)
5849 struct wined3d_device
*device
= context
->swapchain
->device
;
5850 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5853 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
5854 * application provided constants
5856 if (device
->shader_backend
== &arb_program_shader_backend
)
5858 struct shader_arb_priv
*priv
;
5860 if (use_ps(state
)) return;
5862 priv
= device
->shader_priv
;
5863 priv
->pshader_const_dirty
[ARB_FFP_CONST_SPECULAR_ENABLE
] = 1;
5864 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_SPECULAR_ENABLE
+ 1);
5867 if (state
->render_states
[WINED3D_RS_SPECULARENABLE
])
5869 /* The specular color has no alpha */
5870 col
[0] = 1.0f
; col
[1] = 1.0f
;
5871 col
[2] = 1.0f
; col
[3] = 0.0f
;
5873 col
[0] = 0.0f
; col
[1] = 0.0f
;
5874 col
[2] = 0.0f
; col
[3] = 0.0f
;
5876 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_SPECULAR_ENABLE
, col
));
5877 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_SPECULAR_ENABLE, col)");
5880 static void set_bumpmat_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
5882 DWORD stage
= (state_id
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
5883 struct wined3d_device
*device
= context
->swapchain
->device
;
5884 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5887 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_BUMP_ENV
;
5889 if (device
->shader_backend
== &arb_program_shader_backend
)
5891 struct shader_arb_priv
*priv
= device
->shader_priv
;
5893 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants. */
5897 priv
->pshader_const_dirty
[ARB_FFP_CONST_BUMPMAT(stage
)] = 1;
5898 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_BUMPMAT(stage
) + 1);
5901 mat
[0][0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT00
]);
5902 mat
[0][1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT01
]);
5903 mat
[1][0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT10
]);
5904 mat
[1][1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT11
]);
5906 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_BUMPMAT(stage
), &mat
[0][0]));
5907 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_BUMPMAT(stage), &mat[0][0])");
5910 static void tex_bumpenvlum_arbfp(struct wined3d_context
*context
,
5911 const struct wined3d_state
*state
, DWORD state_id
)
5913 DWORD stage
= (state_id
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
5914 struct wined3d_device
*device
= context
->swapchain
->device
;
5915 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5918 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_BUMP_ENV
;
5920 if (device
->shader_backend
== &arb_program_shader_backend
)
5922 struct shader_arb_priv
*priv
= device
->shader_priv
;
5924 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants. */
5928 priv
->pshader_const_dirty
[ARB_FFP_CONST_LUMINANCE(stage
)] = 1;
5929 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_LUMINANCE(stage
) + 1);
5932 param
[0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_LSCALE
]);
5933 param
[1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_LOFFSET
]);
5937 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_LUMINANCE(stage
), param
));
5938 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_LUMINANCE(stage), param)");
5941 static void alpha_test_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
5943 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5947 TRACE("context %p, state %p, state_id %#x.\n", context
, state
, state_id
);
5949 if (state
->render_states
[WINED3D_RS_ALPHATESTENABLE
])
5951 gl_info
->gl_ops
.gl
.p_glEnable(GL_ALPHA_TEST
);
5952 checkGLcall("glEnable GL_ALPHA_TEST");
5956 gl_info
->gl_ops
.gl
.p_glDisable(GL_ALPHA_TEST
);
5957 checkGLcall("glDisable GL_ALPHA_TEST");
5961 ref
= ((float)state
->render_states
[WINED3D_RS_ALPHAREF
]) / 255.0f
;
5962 glParm
= wined3d_gl_compare_func(state
->render_states
[WINED3D_RS_ALPHAFUNC
]);
5966 gl_info
->gl_ops
.gl
.p_glAlphaFunc(glParm
, ref
);
5967 checkGLcall("glAlphaFunc");
5971 static void color_key_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
5973 struct wined3d_device
*device
= context
->swapchain
->device
;
5974 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5975 struct wined3d_color float_key
;
5976 const struct wined3d_texture
*texture
= state
->textures
[0];
5981 /* Don't load the parameter if we're using an arbfp pixel shader,
5982 * otherwise we'll overwrite application provided constants. */
5983 if (device
->shader_backend
== &arb_program_shader_backend
)
5985 struct shader_arb_priv
*priv
;
5990 priv
= device
->shader_priv
;
5991 priv
->pshader_const_dirty
[ARB_FFP_CONST_COLOR_KEY
] = 1;
5992 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_COLOR_KEY
+ 1);
5995 wined3d_format_convert_color_to_float(texture
->resource
.format
, NULL
,
5996 texture
->async
.src_blt_color_key
.color_space_high_value
, &float_key
);
5997 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_COLOR_KEY
, &float_key
.r
));
5998 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_COLOR_KEY, &float_key.r)");
6001 static const char *get_argreg(struct wined3d_string_buffer
*buffer
, DWORD argnum
, unsigned int stage
, DWORD arg
)
6005 if(arg
== ARG_UNUSED
) return "unused"; /* This is the marker for unused registers */
6007 switch(arg
& WINED3DTA_SELECTMASK
) {
6008 case WINED3DTA_DIFFUSE
:
6009 ret
= "fragment.color.primary"; break;
6011 case WINED3DTA_CURRENT
:
6012 if (!stage
) ret
= "fragment.color.primary";
6016 case WINED3DTA_TEXTURE
:
6018 case 0: ret
= "tex0"; break;
6019 case 1: ret
= "tex1"; break;
6020 case 2: ret
= "tex2"; break;
6021 case 3: ret
= "tex3"; break;
6022 case 4: ret
= "tex4"; break;
6023 case 5: ret
= "tex5"; break;
6024 case 6: ret
= "tex6"; break;
6025 case 7: ret
= "tex7"; break;
6026 default: ret
= "unknown texture";
6030 case WINED3DTA_TFACTOR
:
6031 ret
= "tfactor"; break;
6033 case WINED3DTA_SPECULAR
:
6034 ret
= "fragment.color.secondary"; break;
6036 case WINED3DTA_TEMP
:
6037 ret
= "tempreg"; break;
6039 case WINED3DTA_CONSTANT
:
6040 FIXME("Implement perstage constants\n");
6042 case 0: ret
= "const0"; break;
6043 case 1: ret
= "const1"; break;
6044 case 2: ret
= "const2"; break;
6045 case 3: ret
= "const3"; break;
6046 case 4: ret
= "const4"; break;
6047 case 5: ret
= "const5"; break;
6048 case 6: ret
= "const6"; break;
6049 case 7: ret
= "const7"; break;
6050 default: ret
= "unknown constant";
6058 if(arg
& WINED3DTA_COMPLEMENT
) {
6059 shader_addline(buffer
, "SUB arg%u, const.x, %s;\n", argnum
, ret
);
6060 if(argnum
== 0) ret
= "arg0";
6061 if(argnum
== 1) ret
= "arg1";
6062 if(argnum
== 2) ret
= "arg2";
6064 if(arg
& WINED3DTA_ALPHAREPLICATE
) {
6065 shader_addline(buffer
, "MOV arg%u, %s.w;\n", argnum
, ret
);
6066 if(argnum
== 0) ret
= "arg0";
6067 if(argnum
== 1) ret
= "arg1";
6068 if(argnum
== 2) ret
= "arg2";
6073 static void gen_ffp_instr(struct wined3d_string_buffer
*buffer
, unsigned int stage
, BOOL color
,
6074 BOOL alpha
, DWORD dst
, DWORD op
, DWORD dw_arg0
, DWORD dw_arg1
, DWORD dw_arg2
)
6076 const char *dstmask
, *dstreg
, *arg0
, *arg1
, *arg2
;
6077 unsigned int mul
= 1;
6079 if(color
&& alpha
) dstmask
= "";
6080 else if(color
) dstmask
= ".xyz";
6081 else dstmask
= ".w";
6083 if(dst
== tempreg
) dstreg
= "tempreg";
6084 else dstreg
= "ret";
6086 arg0
= get_argreg(buffer
, 0, stage
, dw_arg0
);
6087 arg1
= get_argreg(buffer
, 1, stage
, dw_arg1
);
6088 arg2
= get_argreg(buffer
, 2, stage
, dw_arg2
);
6092 case WINED3D_TOP_DISABLE
:
6094 shader_addline(buffer
, "MOV %s%s, fragment.color.primary;\n", dstreg
, dstmask
);
6097 case WINED3D_TOP_SELECT_ARG2
:
6100 case WINED3D_TOP_SELECT_ARG1
:
6101 shader_addline(buffer
, "MOV %s%s, %s;\n", dstreg
, dstmask
, arg1
);
6104 case WINED3D_TOP_MODULATE_4X
:
6107 case WINED3D_TOP_MODULATE_2X
:
6110 case WINED3D_TOP_MODULATE
:
6111 shader_addline(buffer
, "MUL %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6114 case WINED3D_TOP_ADD_SIGNED_2X
:
6117 case WINED3D_TOP_ADD_SIGNED
:
6118 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
6121 case WINED3D_TOP_ADD
:
6122 shader_addline(buffer
, "ADD_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6125 case WINED3D_TOP_SUBTRACT
:
6126 shader_addline(buffer
, "SUB_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6129 case WINED3D_TOP_ADD_SMOOTH
:
6130 shader_addline(buffer
, "SUB arg1, const.x, %s;\n", arg1
);
6131 shader_addline(buffer
, "MAD_SAT %s%s, arg1, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6134 case WINED3D_TOP_BLEND_CURRENT_ALPHA
:
6135 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_CURRENT
);
6136 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6138 case WINED3D_TOP_BLEND_FACTOR_ALPHA
:
6139 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TFACTOR
);
6140 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6142 case WINED3D_TOP_BLEND_TEXTURE_ALPHA
:
6143 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
6144 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6146 case WINED3D_TOP_BLEND_DIFFUSE_ALPHA
:
6147 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_DIFFUSE
);
6148 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6151 case WINED3D_TOP_BLEND_TEXTURE_ALPHA_PM
:
6152 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
6153 shader_addline(buffer
, "SUB arg0.w, const.x, %s.w;\n", arg0
);
6154 shader_addline(buffer
, "MAD_SAT %s%s, %s, arg0.w, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6157 /* D3DTOP_PREMODULATE ???? */
6159 case WINED3D_TOP_MODULATE_INVALPHA_ADD_COLOR
:
6160 shader_addline(buffer
, "SUB arg0.w, const.x, %s;\n", arg1
);
6161 shader_addline(buffer
, "MAD_SAT %s%s, arg0.w, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6163 case WINED3D_TOP_MODULATE_ALPHA_ADD_COLOR
:
6164 shader_addline(buffer
, "MAD_SAT %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
6166 case WINED3D_TOP_MODULATE_INVCOLOR_ADD_ALPHA
:
6167 shader_addline(buffer
, "SUB arg0, const.x, %s;\n", arg1
);
6168 shader_addline(buffer
, "MAD_SAT %s%s, arg0, %s, %s.w;\n", dstreg
, dstmask
, arg2
, arg1
);
6170 case WINED3D_TOP_MODULATE_COLOR_ADD_ALPHA
:
6171 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s.w;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
6174 case WINED3D_TOP_DOTPRODUCT3
:
6176 shader_addline(buffer
, "SUB arg1, %s, const.w;\n", arg1
);
6177 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
6178 shader_addline(buffer
, "DP3_SAT %s%s, arg1, arg2;\n", dstreg
, dstmask
);
6181 case WINED3D_TOP_MULTIPLY_ADD
:
6182 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg0
);
6185 case WINED3D_TOP_LERP
:
6186 /* The msdn is not quite right here */
6187 shader_addline(buffer
, "LRP %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6190 case WINED3D_TOP_BUMPENVMAP
:
6191 case WINED3D_TOP_BUMPENVMAP_LUMINANCE
:
6192 /* Those are handled in the first pass of the shader(generation pass 1 and 2) already */
6196 FIXME("Unhandled texture op %08x\n", op
);
6200 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.y;\n", dstreg
, dstmask
, dstreg
);
6202 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.z;\n", dstreg
, dstmask
, dstreg
);
6205 static const char *arbfp_texture_target(enum wined3d_gl_resource_type type
)
6209 case WINED3D_GL_RES_TYPE_TEX_1D
:
6211 case WINED3D_GL_RES_TYPE_TEX_2D
:
6213 case WINED3D_GL_RES_TYPE_TEX_3D
:
6215 case WINED3D_GL_RES_TYPE_TEX_CUBE
:
6217 case WINED3D_GL_RES_TYPE_TEX_RECT
:
6220 return "unexpected_resource_type";
6224 static GLuint
gen_arbfp_ffp_shader(const struct ffp_frag_settings
*settings
, const struct wined3d_gl_info
*gl_info
)
6227 struct wined3d_string_buffer buffer
;
6228 BOOL tex_read
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
6229 BOOL bump_used
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
6230 BOOL luminance_used
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
6231 UINT lowest_disabled_stage
;
6232 const char *textype
;
6234 char colorcor_dst
[8];
6236 DWORD arg0
, arg1
, arg2
;
6237 BOOL tempreg_used
= FALSE
, tfactor_used
= FALSE
;
6239 const char *final_combiner_src
= "ret";
6240 BOOL custom_linear_fog
= FALSE
;
6242 if (!string_buffer_init(&buffer
))
6244 ERR("Failed to initialize shader buffer.\n");
6248 shader_addline(&buffer
, "!!ARBfp1.0\n");
6250 if (settings
->color_key_enabled
)
6252 shader_addline(&buffer
, "PARAM color_key = program.env[%u];\n", ARB_FFP_CONST_COLOR_KEY
);
6256 /* Find out which textures are read */
6257 for (stage
= 0; stage
< MAX_TEXTURES
; ++stage
)
6259 if (settings
->op
[stage
].cop
== WINED3D_TOP_DISABLE
)
6261 arg0
= settings
->op
[stage
].carg0
& WINED3DTA_SELECTMASK
;
6262 arg1
= settings
->op
[stage
].carg1
& WINED3DTA_SELECTMASK
;
6263 arg2
= settings
->op
[stage
].carg2
& WINED3DTA_SELECTMASK
;
6264 if(arg0
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6265 if(arg1
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6266 if(arg2
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6268 if (settings
->op
[stage
].cop
== WINED3D_TOP_BLEND_TEXTURE_ALPHA
)
6269 tex_read
[stage
] = TRUE
;
6270 if (settings
->op
[stage
].cop
== WINED3D_TOP_BLEND_TEXTURE_ALPHA_PM
)
6271 tex_read
[stage
] = TRUE
;
6272 if (settings
->op
[stage
].cop
== WINED3D_TOP_BUMPENVMAP
)
6274 bump_used
[stage
] = TRUE
;
6275 tex_read
[stage
] = TRUE
;
6277 if (settings
->op
[stage
].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
)
6279 bump_used
[stage
] = TRUE
;
6280 tex_read
[stage
] = TRUE
;
6281 luminance_used
[stage
] = TRUE
;
6283 else if (settings
->op
[stage
].cop
== WINED3D_TOP_BLEND_FACTOR_ALPHA
)
6285 tfactor_used
= TRUE
;
6288 if(arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
) {
6289 tfactor_used
= TRUE
;
6292 if(settings
->op
[stage
].dst
== tempreg
) tempreg_used
= TRUE
;
6293 if(arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
) {
6294 tempreg_used
= TRUE
;
6297 if (settings
->op
[stage
].aop
== WINED3D_TOP_DISABLE
)
6299 arg0
= settings
->op
[stage
].aarg0
& WINED3DTA_SELECTMASK
;
6300 arg1
= settings
->op
[stage
].aarg1
& WINED3DTA_SELECTMASK
;
6301 arg2
= settings
->op
[stage
].aarg2
& WINED3DTA_SELECTMASK
;
6302 if(arg0
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6303 if(arg1
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6304 if(arg2
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6306 if(arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
) {
6307 tempreg_used
= TRUE
;
6309 if(arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
) {
6310 tfactor_used
= TRUE
;
6313 lowest_disabled_stage
= stage
;
6315 switch (settings
->fog
)
6317 case WINED3D_FFP_PS_FOG_OFF
: break;
6318 case WINED3D_FFP_PS_FOG_LINEAR
:
6319 if (gl_info
->quirks
& WINED3D_QUIRK_BROKEN_ARB_FOG
)
6321 custom_linear_fog
= TRUE
;
6324 shader_addline(&buffer
, "OPTION ARB_fog_linear;\n");
6327 case WINED3D_FFP_PS_FOG_EXP
: shader_addline(&buffer
, "OPTION ARB_fog_exp;\n"); break;
6328 case WINED3D_FFP_PS_FOG_EXP2
: shader_addline(&buffer
, "OPTION ARB_fog_exp2;\n"); break;
6329 default: FIXME("Unexpected fog setting %d\n", settings
->fog
);
6332 shader_addline(&buffer
, "PARAM const = {1, 2, 4, 0.5};\n");
6333 shader_addline(&buffer
, "TEMP TMP;\n");
6334 shader_addline(&buffer
, "TEMP ret;\n");
6335 if(tempreg_used
|| settings
->sRGB_write
) shader_addline(&buffer
, "TEMP tempreg;\n");
6336 shader_addline(&buffer
, "TEMP arg0;\n");
6337 shader_addline(&buffer
, "TEMP arg1;\n");
6338 shader_addline(&buffer
, "TEMP arg2;\n");
6339 for(stage
= 0; stage
< MAX_TEXTURES
; stage
++) {
6340 if(!tex_read
[stage
]) continue;
6341 shader_addline(&buffer
, "TEMP tex%u;\n", stage
);
6342 if(!bump_used
[stage
]) continue;
6343 shader_addline(&buffer
, "PARAM bumpmat%u = program.env[%u];\n", stage
, ARB_FFP_CONST_BUMPMAT(stage
));
6344 if(!luminance_used
[stage
]) continue;
6345 shader_addline(&buffer
, "PARAM luminance%u = program.env[%u];\n", stage
, ARB_FFP_CONST_LUMINANCE(stage
));
6348 shader_addline(&buffer
, "PARAM tfactor = program.env[%u];\n", ARB_FFP_CONST_TFACTOR
);
6350 shader_addline(&buffer
, "PARAM specular_enable = program.env[%u];\n", ARB_FFP_CONST_SPECULAR_ENABLE
);
6352 if (settings
->sRGB_write
)
6354 shader_addline(&buffer
, "PARAM srgb_consts0 = ");
6355 shader_arb_append_imm_vec4(&buffer
, wined3d_srgb_const0
);
6356 shader_addline(&buffer
, ";\n");
6357 shader_addline(&buffer
, "PARAM srgb_consts1 = ");
6358 shader_arb_append_imm_vec4(&buffer
, wined3d_srgb_const1
);
6359 shader_addline(&buffer
, ";\n");
6362 if (lowest_disabled_stage
< 7 && settings
->emul_clipplanes
)
6363 shader_addline(&buffer
, "KIL fragment.texcoord[7];\n");
6365 if (tempreg_used
|| settings
->sRGB_write
)
6366 shader_addline(&buffer
, "MOV tempreg, 0.0;\n");
6368 /* Generate texture sampling instructions) */
6369 for (stage
= 0; stage
< MAX_TEXTURES
&& settings
->op
[stage
].cop
!= WINED3D_TOP_DISABLE
; ++stage
)
6371 if (!tex_read
[stage
])
6374 textype
= arbfp_texture_target(settings
->op
[stage
].tex_type
);
6376 if(settings
->op
[stage
].projected
== proj_none
) {
6378 } else if(settings
->op
[stage
].projected
== proj_count4
||
6379 settings
->op
[stage
].projected
== proj_count3
) {
6382 FIXME("Unexpected projection mode %d\n", settings
->op
[stage
].projected
);
6387 && (settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP
6388 || settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
))
6390 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, x, z, 0, 0;\n", stage
- 1);
6391 shader_addline(&buffer
, "DP3 ret.x, arg1, tex%u;\n", stage
- 1);
6392 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, y, w, 0, 0;\n", stage
- 1);
6393 shader_addline(&buffer
, "DP3 ret.y, arg1, tex%u;\n", stage
- 1);
6395 /* with projective textures, texbem only divides the static texture coord, not the displacement,
6396 * so multiply the displacement with the dividing parameter before passing it to TXP
6398 if (settings
->op
[stage
].projected
!= proj_none
) {
6399 if(settings
->op
[stage
].projected
== proj_count4
) {
6400 shader_addline(&buffer
, "MOV ret.w, fragment.texcoord[%u].w;\n", stage
);
6401 shader_addline(&buffer
, "MUL ret.xyz, ret, fragment.texcoord[%u].w, fragment.texcoord[%u];\n", stage
, stage
);
6403 shader_addline(&buffer
, "MOV ret.w, fragment.texcoord[%u].z;\n", stage
);
6404 shader_addline(&buffer
, "MAD ret.xyz, ret, fragment.texcoord[%u].z, fragment.texcoord[%u];\n", stage
, stage
);
6407 shader_addline(&buffer
, "ADD ret, ret, fragment.texcoord[%u];\n", stage
);
6410 shader_addline(&buffer
, "%s tex%u, ret, texture[%u], %s;\n",
6411 instr
, stage
, stage
, textype
);
6412 if (settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
)
6414 shader_addline(&buffer
, "MAD_SAT ret.x, tex%u.z, luminance%u.x, luminance%u.y;\n",
6415 stage
- 1, stage
- 1, stage
- 1);
6416 shader_addline(&buffer
, "MUL tex%u, tex%u, ret.x;\n", stage
, stage
);
6418 } else if(settings
->op
[stage
].projected
== proj_count3
) {
6419 shader_addline(&buffer
, "MOV ret, fragment.texcoord[%u];\n", stage
);
6420 shader_addline(&buffer
, "MOV ret.w, ret.z;\n");
6421 shader_addline(&buffer
, "%s tex%u, ret, texture[%u], %s;\n",
6422 instr
, stage
, stage
, textype
);
6424 shader_addline(&buffer
, "%s tex%u, fragment.texcoord[%u], texture[%u], %s;\n",
6425 instr
, stage
, stage
, stage
, textype
);
6428 sprintf(colorcor_dst
, "tex%u", stage
);
6429 gen_color_correction(&buffer
, colorcor_dst
, WINED3DSP_WRITEMASK_ALL
, "const.x", "const.y",
6430 settings
->op
[stage
].color_fixup
);
6433 if (settings
->color_key_enabled
)
6435 shader_addline(&buffer
, "SUB TMP, tex0, color_key;\n");
6436 shader_addline(&buffer
, "DP4 TMP.b, TMP, TMP;\n");
6437 shader_addline(&buffer
, "SGE TMP, -TMP.b, 0.0;\n");
6438 shader_addline(&buffer
, "KIL -TMP;\n");
6441 /* Generate the main shader */
6442 for (stage
= 0; stage
< MAX_TEXTURES
; ++stage
)
6444 if (settings
->op
[stage
].cop
== WINED3D_TOP_DISABLE
)
6447 final_combiner_src
= "fragment.color.primary";
6451 if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG1
6452 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG1
)
6453 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
;
6454 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG1
6455 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG2
)
6456 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg2
;
6457 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG2
6458 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG1
)
6459 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg1
;
6460 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG2
6461 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG2
)
6462 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
6464 op_equal
= settings
->op
[stage
].aop
== settings
->op
[stage
].cop
6465 && settings
->op
[stage
].carg0
== settings
->op
[stage
].aarg0
6466 && settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
6467 && settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
6469 if (settings
->op
[stage
].aop
== WINED3D_TOP_DISABLE
)
6471 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].dst
,
6472 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6473 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6475 shader_addline(&buffer
, "MOV ret.w, fragment.color.primary.w;\n");
6479 gen_ffp_instr(&buffer
, stage
, TRUE
, TRUE
, settings
->op
[stage
].dst
,
6480 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6481 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6483 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].dst
,
6484 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6485 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6486 gen_ffp_instr(&buffer
, stage
, FALSE
, TRUE
, settings
->op
[stage
].dst
,
6487 settings
->op
[stage
].aop
, settings
->op
[stage
].aarg0
,
6488 settings
->op
[stage
].aarg1
, settings
->op
[stage
].aarg2
);
6492 if (settings
->sRGB_write
|| custom_linear_fog
)
6494 shader_addline(&buffer
, "MAD ret, fragment.color.secondary, specular_enable, %s;\n", final_combiner_src
);
6495 if (settings
->sRGB_write
)
6496 arbfp_add_sRGB_correction(&buffer
, "ret", "arg0", "arg1", "arg2", "tempreg", FALSE
);
6497 if (custom_linear_fog
)
6498 arbfp_add_linear_fog(&buffer
, "ret", "arg0");
6499 shader_addline(&buffer
, "MOV result.color, ret;\n");
6503 shader_addline(&buffer
, "MAD result.color, fragment.color.secondary, specular_enable, %s;\n",
6504 final_combiner_src
);
6508 shader_addline(&buffer
, "END\n");
6510 /* Generate the shader */
6511 GL_EXTCALL(glGenProgramsARB(1, &ret
));
6512 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, ret
));
6513 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
6515 string_buffer_free(&buffer
);
6519 static void fragment_prog_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6521 const struct wined3d_device
*device
= context
->swapchain
->device
;
6522 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
6523 struct shader_arb_priv
*priv
= device
->fragment_priv
;
6524 BOOL use_pshader
= use_ps(state
);
6525 struct ffp_frag_settings settings
;
6526 const struct arbfp_ffp_desc
*desc
;
6529 TRACE("context %p, state %p, state_id %#x.\n", context
, state
, state_id
);
6531 if (isStateDirty(context
, STATE_RENDER(WINED3D_RS_FOGENABLE
)))
6533 if (!use_pshader
&& device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
)
6535 /* Reload fixed function constants since they collide with the
6536 * pixel shader constants. */
6537 for (i
= 0; i
< MAX_TEXTURES
; ++i
)
6539 set_bumpmat_arbfp(context
, state
, STATE_TEXTURESTAGE(i
, WINED3D_TSS_BUMPENV_MAT00
));
6541 state_texfactor_arbfp(context
, state
, STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
));
6542 state_arb_specularenable(context
, state
, STATE_RENDER(WINED3D_RS_SPECULARENABLE
));
6543 color_key_arbfp(context
, state
, STATE_COLOR_KEY
);
6545 else if (use_pshader
)
6547 context
->shader_update_mask
|= 1 << WINED3D_SHADER_TYPE_PIXEL
;
6554 /* Find or create a shader implementing the fixed function pipeline
6555 * settings, then activate it. */
6556 gen_ffp_frag_op(context
, state
, &settings
, FALSE
);
6557 desc
= (const struct arbfp_ffp_desc
*)find_ffp_frag_shader(&priv
->fragment_shaders
, &settings
);
6559 struct arbfp_ffp_desc
*new_desc
= HeapAlloc(GetProcessHeap(), 0, sizeof(*new_desc
));
6562 ERR("Out of memory\n");
6566 new_desc
->parent
.settings
= settings
;
6567 new_desc
->shader
= gen_arbfp_ffp_shader(&settings
, gl_info
);
6568 add_ffp_frag_shader(&priv
->fragment_shaders
, &new_desc
->parent
);
6569 TRACE("Allocated fixed function replacement shader descriptor %p\n", new_desc
);
6573 /* Now activate the replacement program. GL_FRAGMENT_PROGRAM_ARB is already active (however, note the
6574 * comment above the shader_select call below). If e.g. GLSL is active, the shader_select call will
6577 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, desc
->shader
));
6578 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, desc->shader)");
6579 priv
->current_fprogram_id
= desc
->shader
;
6581 if (device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
)
6583 /* Reload fixed function constants since they collide with the
6584 * pixel shader constants. */
6585 for (i
= 0; i
< MAX_TEXTURES
; ++i
)
6587 set_bumpmat_arbfp(context
, state
, STATE_TEXTURESTAGE(i
, WINED3D_TSS_BUMPENV_MAT00
));
6589 state_texfactor_arbfp(context
, state
, STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
));
6590 state_arb_specularenable(context
, state
, STATE_RENDER(WINED3D_RS_SPECULARENABLE
));
6592 context
->last_was_pshader
= FALSE
;
6594 else if (!context
->last_was_pshader
)
6596 if (device
->shader_backend
== &arb_program_shader_backend
)
6597 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_F
;
6598 context
->last_was_pshader
= TRUE
;
6601 context
->shader_update_mask
|= 1 << WINED3D_SHADER_TYPE_PIXEL
;
6604 /* We can't link the fog states to the fragment state directly since the
6605 * vertex pipeline links them to FOGENABLE. A different linking in different
6606 * pipeline parts can't be expressed in the combined state table, so we need
6607 * to handle that with a forwarding function. The other invisible side effect
6608 * is that changing the fog start and fog end (which links to FOGENABLE in
6609 * vertex) results in the fragment_prog_arbfp function being called because
6610 * FOGENABLE is dirty, which calls this function here. */
6611 static void state_arbfp_fog(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6613 enum fogsource new_source
;
6614 DWORD fogstart
= state
->render_states
[WINED3D_RS_FOGSTART
];
6615 DWORD fogend
= state
->render_states
[WINED3D_RS_FOGEND
];
6617 TRACE("context %p, state %p, state_id %#x.\n", context
, state
, state_id
);
6619 if (!isStateDirty(context
, STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
)))
6620 fragment_prog_arbfp(context
, state
, state_id
);
6622 if (!state
->render_states
[WINED3D_RS_FOGENABLE
])
6625 if (state
->render_states
[WINED3D_RS_FOGTABLEMODE
] == WINED3D_FOG_NONE
)
6629 new_source
= FOGSOURCE_VS
;
6633 if (state
->render_states
[WINED3D_RS_FOGVERTEXMODE
] == WINED3D_FOG_NONE
|| context
->last_was_rhw
)
6634 new_source
= FOGSOURCE_COORD
;
6636 new_source
= FOGSOURCE_FFP
;
6641 new_source
= FOGSOURCE_FFP
;
6644 if (new_source
!= context
->fog_source
|| fogstart
== fogend
)
6646 context
->fog_source
= new_source
;
6647 state_fogstartend(context
, state
, STATE_RENDER(WINED3D_RS_FOGSTART
));
6651 static void textransform(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6653 if (!isStateDirty(context
, STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
)))
6654 fragment_prog_arbfp(context
, state
, state_id
);
6657 static const struct StateEntryTemplate arbfp_fragmentstate_template
[] =
6659 {STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
), { STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
), state_texfactor_arbfp
}, WINED3D_GL_EXT_NONE
},
6660 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6661 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6662 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6663 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6664 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6665 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6666 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6667 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6668 {STATE_TEXTURESTAGE(0, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6669 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6670 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6671 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6672 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6673 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6674 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6675 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6676 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6677 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6678 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6679 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6680 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6681 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6682 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6683 {STATE_TEXTURESTAGE(1, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6684 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6685 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6686 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6687 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6688 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6689 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6690 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6691 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6692 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6693 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6694 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6695 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6696 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6697 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6698 {STATE_TEXTURESTAGE(2, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6699 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6700 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6701 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6702 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6703 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6704 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6705 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6706 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6707 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6708 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6709 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6710 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6711 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6712 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6713 {STATE_TEXTURESTAGE(3, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6714 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6715 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6716 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6717 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6718 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6719 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6720 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6721 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6722 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6723 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6724 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6725 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6726 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6727 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6728 {STATE_TEXTURESTAGE(4, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6729 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6730 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6731 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6732 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6733 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6734 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6735 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6736 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6737 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6738 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6739 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6740 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6741 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6742 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6743 {STATE_TEXTURESTAGE(5, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6744 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6745 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6746 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6747 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6748 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6749 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6750 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6751 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6752 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6753 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6754 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6755 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6756 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6757 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6758 {STATE_TEXTURESTAGE(6, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6759 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6760 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6761 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6762 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6763 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6764 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6765 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6766 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6767 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6768 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6769 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6770 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6771 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6772 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6773 {STATE_TEXTURESTAGE(7, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6774 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6775 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6776 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6777 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6778 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6779 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6780 {STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
6781 {STATE_RENDER(WINED3D_RS_ALPHAFUNC
), { STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6782 {STATE_RENDER(WINED3D_RS_ALPHAREF
), { STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6783 {STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), { STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), alpha_test_arbfp
}, WINED3D_GL_EXT_NONE
},
6784 {STATE_RENDER(WINED3D_RS_COLORKEYENABLE
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6785 {STATE_COLOR_KEY
, { STATE_COLOR_KEY
, color_key_arbfp
}, WINED3D_GL_EXT_NONE
},
6786 {STATE_RENDER(WINED3D_RS_FOGENABLE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), state_arbfp_fog
}, WINED3D_GL_EXT_NONE
},
6787 {STATE_RENDER(WINED3D_RS_FOGTABLEMODE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6788 {STATE_RENDER(WINED3D_RS_FOGVERTEXMODE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6789 {STATE_RENDER(WINED3D_RS_FOGSTART
), { STATE_RENDER(WINED3D_RS_FOGSTART
), state_fogstartend
}, WINED3D_GL_EXT_NONE
},
6790 {STATE_RENDER(WINED3D_RS_FOGEND
), { STATE_RENDER(WINED3D_RS_FOGSTART
), NULL
}, WINED3D_GL_EXT_NONE
},
6791 {STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), { STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), state_srgbwrite
}, ARB_FRAMEBUFFER_SRGB
},
6792 {STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6793 {STATE_RENDER(WINED3D_RS_FOGCOLOR
), { STATE_RENDER(WINED3D_RS_FOGCOLOR
), state_fogcolor
}, WINED3D_GL_EXT_NONE
},
6794 {STATE_RENDER(WINED3D_RS_FOGDENSITY
), { STATE_RENDER(WINED3D_RS_FOGDENSITY
), state_fogdensity
}, WINED3D_GL_EXT_NONE
},
6795 {STATE_TEXTURESTAGE(0,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(0, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6796 {STATE_TEXTURESTAGE(1,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(1, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6797 {STATE_TEXTURESTAGE(2,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(2, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6798 {STATE_TEXTURESTAGE(3,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(3, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6799 {STATE_TEXTURESTAGE(4,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(4, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6800 {STATE_TEXTURESTAGE(5,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(5, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6801 {STATE_TEXTURESTAGE(6,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(6, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6802 {STATE_TEXTURESTAGE(7,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(7, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6803 {STATE_RENDER(WINED3D_RS_SPECULARENABLE
), { STATE_RENDER(WINED3D_RS_SPECULARENABLE
), state_arb_specularenable
}, WINED3D_GL_EXT_NONE
},
6804 {0 /* Terminate */, { 0, 0 }, WINED3D_GL_EXT_NONE
},
6807 static BOOL
arbfp_alloc_context_data(struct wined3d_context
*context
)
6812 static void arbfp_free_context_data(struct wined3d_context
*context
)
6816 const struct fragment_pipeline arbfp_fragment_pipeline
= {
6821 arbfp_alloc_context_data
,
6822 arbfp_free_context_data
,
6823 shader_arb_color_fixup_supported
,
6824 arbfp_fragmentstate_template
,
6827 struct arbfp_blit_type
6829 enum complex_fixup fixup
: 4;
6830 enum wined3d_gl_resource_type res_type
: 3;
6831 DWORD use_color_key
: 1;
6835 struct arbfp_blit_desc
6838 struct arbfp_blit_type type
;
6839 struct wine_rb_entry entry
;
6842 #define ARBFP_BLIT_PARAM_SIZE 0
6843 #define ARBFP_BLIT_PARAM_COLOR_KEY 1
6845 struct arbfp_blit_priv
6847 struct wine_rb_tree shaders
;
6848 GLuint palette_texture
;
6851 static int arbfp_blit_type_compare(const void *key
, const struct wine_rb_entry
*entry
)
6853 const struct arbfp_blit_type
*ka
= key
;
6854 const struct arbfp_blit_type
*kb
= &WINE_RB_ENTRY_VALUE(entry
, const struct arbfp_blit_desc
, entry
)->type
;
6856 return memcmp(ka
, kb
, sizeof(*ka
));
6859 /* Context activation is done by the caller. */
6860 static void arbfp_free_blit_shader(struct wine_rb_entry
*entry
, void *context
)
6862 const struct wined3d_gl_info
*gl_info
= context
;
6863 struct arbfp_blit_desc
*entry_arb
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_blit_desc
, entry
);
6865 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb
->shader
));
6866 checkGLcall("glDeleteProgramsARB(1, &entry_arb->shader)");
6867 HeapFree(GetProcessHeap(), 0, entry_arb
);
6870 static const struct wine_rb_functions wined3d_arbfp_blit_rb_functions
=
6875 arbfp_blit_type_compare
,
6878 static HRESULT
arbfp_blit_alloc(struct wined3d_device
*device
)
6880 struct arbfp_blit_priv
*priv
;
6882 if (!(priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*priv
))))
6883 return E_OUTOFMEMORY
;
6885 if (wine_rb_init(&priv
->shaders
, &wined3d_arbfp_blit_rb_functions
) == -1)
6887 ERR("Failed to initialize rbtree.\n");
6888 HeapFree(GetProcessHeap(), 0, priv
);
6889 return E_OUTOFMEMORY
;
6892 device
->blit_priv
= priv
;
6897 /* Context activation is done by the caller. */
6898 static void arbfp_blit_free(struct wined3d_device
*device
)
6900 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
6901 struct arbfp_blit_priv
*priv
= device
->blit_priv
;
6903 wine_rb_destroy(&priv
->shaders
, arbfp_free_blit_shader
, &device
->adapter
->gl_info
);
6904 checkGLcall("Delete blit programs");
6906 if (priv
->palette_texture
)
6907 gl_info
->gl_ops
.gl
.p_glDeleteTextures(1, &priv
->palette_texture
);
6909 HeapFree(GetProcessHeap(), 0, device
->blit_priv
);
6910 device
->blit_priv
= NULL
;
6913 static BOOL
gen_planar_yuv_read(struct wined3d_string_buffer
*buffer
, const struct arbfp_blit_type
*type
,
6917 const char *tex
, *texinstr
= "TXP";
6919 if (type
->fixup
== COMPLEX_FIXUP_UYVY
)
6930 tex
= arbfp_texture_target(type
->res_type
);
6931 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_RECT
)
6934 /* First we have to read the chroma values. This means we need at least two pixels(no filtering),
6935 * or 4 pixels(with filtering). To get the unmodified chromas, we have to rid ourselves of the
6936 * filtering when we sample the texture.
6938 * These are the rules for reading the chroma:
6944 * So we have to get the sampling x position in non-normalized coordinates in integers
6946 if (type
->res_type
!= WINED3D_GL_RES_TYPE_TEX_RECT
)
6948 shader_addline(buffer
, "MUL texcrd.xy, fragment.texcoord[0], size.x;\n");
6949 shader_addline(buffer
, "MOV texcrd.w, size.x;\n");
6953 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
6955 /* We must not allow filtering between pixel x and x+1, this would mix U and V
6956 * Vertical filtering is ok. However, bear in mind that the pixel center is at
6959 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
6960 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, coef.y;\n");
6962 /* Divide the x coordinate by 0.5 and get the fraction. This gives 0.25 and 0.75 for the
6963 * even and odd pixels respectively
6965 shader_addline(buffer
, "MUL texcrd2, texcrd, coef.y;\n");
6966 shader_addline(buffer
, "FRC texcrd2, texcrd2;\n");
6968 /* Sample Pixel 1 */
6969 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
6971 /* Put the value into either of the chroma values */
6972 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
6973 shader_addline(buffer
, "MUL chroma.x, luminance.%c, temp.x;\n", chroma
);
6974 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
6975 shader_addline(buffer
, "MUL chroma.y, luminance.%c, temp.x;\n", chroma
);
6977 /* Sample pixel 2. If we read an even pixel(SLT above returned 1), sample
6978 * the pixel right to the current one. Otherwise, sample the left pixel.
6979 * Bias and scale the SLT result to -1;1 and add it to the texcrd.x.
6981 shader_addline(buffer
, "MAD temp.x, temp.x, coef.z, -coef.x;\n");
6982 shader_addline(buffer
, "ADD texcrd.x, texcrd, temp.x;\n");
6983 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
6985 /* Put the value into the other chroma */
6986 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
6987 shader_addline(buffer
, "MAD chroma.y, luminance.%c, temp.x, chroma.y;\n", chroma
);
6988 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
6989 shader_addline(buffer
, "MAD chroma.x, luminance.%c, temp.x, chroma.x;\n", chroma
);
6991 /* TODO: If filtering is enabled, sample a 2nd pair of pixels left or right of
6992 * the current one and lerp the two U and V values
6995 /* This gives the correctly filtered luminance value */
6996 shader_addline(buffer
, "TEX luminance, fragment.texcoord[0], texture[0], %s;\n", tex
);
7001 static BOOL
gen_yv12_read(struct wined3d_string_buffer
*buffer
, const struct arbfp_blit_type
*type
,
7005 static const float yv12_coef
[]
7006 = {2.0f
/ 3.0f
, 1.0f
/ 6.0f
, (2.0f
/ 3.0f
) + (1.0f
/ 6.0f
), 1.0f
/ 3.0f
};
7008 tex
= arbfp_texture_target(type
->res_type
);
7010 /* YV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
7011 * V and a (W/2)x(H/2) U plane, each with 8 bit per pixel. So the effective
7012 * bitdepth is 12 bits per pixel. Since the U and V planes have only half the
7013 * pitch of the luminance plane, the packing into the gl texture is a bit
7014 * unfortunate. If the whole texture is interpreted as luminance data it looks
7015 * approximately like this:
7017 * +----------------------------------+----
7029 * +----------------+-----------------+----
7031 * | V even rows | V odd rows |
7033 * +----------------+------------------ -
7035 * | U even rows | U odd rows |
7037 * +----------------+-----------------+----
7041 * So it appears as if there are 4 chroma images, but in fact the odd rows
7042 * in the chroma images are in the same row as the even ones. So its is
7043 * kinda tricky to read
7045 * When reading from rectangle textures, keep in mind that the input y coordinates
7046 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height
7048 shader_addline(buffer
, "PARAM yv12_coef = ");
7049 shader_arb_append_imm_vec4(buffer
, yv12_coef
);
7050 shader_addline(buffer
, ";\n");
7052 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7053 /* the chroma planes have only half the width */
7054 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, coef.y;\n");
7056 /* The first value is between 2/3 and 5/6th of the texture's height, so scale+bias
7057 * the coordinate. Also read the right side of the image when reading odd lines
7059 * Don't forget to clamp the y values in into the range, otherwise we'll get filtering
7062 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7065 shader_addline(buffer
, "RCP chroma.w, size.y;\n");
7067 shader_addline(buffer
, "MUL texcrd2.y, texcrd.y, size.y;\n");
7069 shader_addline(buffer
, "FLR texcrd2.y, texcrd2.y;\n");
7070 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, yv12_coef.y, yv12_coef.x;\n");
7072 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
7073 shader_addline(buffer
, "ADD texcrd2.x, texcrd2.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
7074 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
7075 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
7076 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
7078 /* clamp, keep the half pixel origin in mind */
7079 shader_addline(buffer
, "MAD temp.y, coef.y, chroma.w, yv12_coef.x;\n");
7080 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7081 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.z;\n");
7082 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7086 /* Read from [size - size+size/4] */
7087 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
7088 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, coef.w, size.y;\n");
7090 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
7091 shader_addline(buffer
, "ADD texcrd2.x, texcrd.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
7092 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
7093 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
7094 shader_addline(buffer
, "MUL texcrd2.x, texcrd2.x, size.x;\n");
7095 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
7097 /* Make sure to read exactly from the pixel center */
7098 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
7099 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, coef.y;\n");
7102 shader_addline(buffer
, "MAD temp.y, size.y, coef.w, size.y;\n");
7103 shader_addline(buffer
, "ADD temp.y, temp.y, -coef.y;\n");
7104 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7105 shader_addline(buffer
, "ADD temp.y, size.y, coef.y;\n");
7106 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7108 /* Read the texture, put the result into the output register */
7109 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7110 shader_addline(buffer
, "MOV chroma.x, temp.w;\n");
7112 /* The other chroma value is 1/6th of the texture lower, from 5/6th to 6/6th
7113 * No need to clamp because we're just reusing the already clamped value from above
7115 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7116 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, yv12_coef.y;\n");
7118 shader_addline(buffer
, "MAD texcrd.y, size.y, coef.w, texcrd.y;\n");
7119 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7120 shader_addline(buffer
, "MOV chroma.y, temp.w;\n");
7122 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
7123 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
7124 * values due to filtering
7126 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7127 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7129 /* Multiply the y coordinate by 2/3 and clamp it */
7130 shader_addline(buffer
, "MUL texcrd.y, texcrd.y, yv12_coef.x;\n");
7131 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.x;\n");
7132 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7133 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7137 /* Reading from texture_rectangles is pretty straightforward, just use the unmodified
7138 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
7141 shader_addline(buffer
, "ADD temp.x, size.y, -coef.y;\n");
7142 shader_addline(buffer
, "MIN texcrd.y, texcrd.y, size.x;\n");
7143 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7150 static BOOL
gen_nv12_read(struct wined3d_string_buffer
*buffer
, const struct arbfp_blit_type
*type
,
7154 static const float nv12_coef
[]
7155 = {2.0f
/ 3.0f
, 1.0f
/ 3.0f
, 1.0f
, 1.0f
};
7157 tex
= arbfp_texture_target(type
->res_type
);
7159 /* NV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
7160 * sized plane where each component is an UV pair. So the effective
7161 * bitdepth is 12 bits per pixel If the whole texture is interpreted as luminance
7162 * data it looks approximately like this:
7164 * +----------------------------------+----
7176 * +----------------------------------+----
7177 * |UVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUV|
7178 * |UVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUV|
7184 * +----------------------------------+----
7186 * When reading from rectangle textures, keep in mind that the input y coordinates
7187 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height. */
7189 shader_addline(buffer
, "PARAM nv12_coef = ");
7190 shader_arb_append_imm_vec4(buffer
, nv12_coef
);
7191 shader_addline(buffer
, ";\n");
7193 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7194 /* We only have half the number of chroma pixels. */
7195 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, coef.y;\n");
7197 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7199 shader_addline(buffer
, "RCP chroma.w, size.x;\n");
7200 shader_addline(buffer
, "RCP chroma.z, size.y;\n");
7202 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, nv12_coef.y, nv12_coef.x;\n");
7204 /* We must not allow filtering horizontally, this would mix U and V.
7205 * Vertical filtering is ok. However, bear in mind that the pixel center is at
7206 * 0.5, so add 0.5. */
7208 /* Convert to non-normalized coordinates so we can find the
7209 * individual pixel. */
7210 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, size.x;\n");
7211 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
7212 /* Multiply by 2 since chroma components are stored in UV pixel pairs,
7213 * add 0.5 to hit the center of the pixel. */
7214 shader_addline(buffer
, "MAD texcrd.x, texcrd.x, coef.z, coef.y;\n");
7216 /* Convert back to normalized coordinates. */
7217 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, chroma.w;\n");
7219 /* Clamp, keep the half pixel origin in mind. */
7220 shader_addline(buffer
, "MAD temp.y, coef.y, chroma.z, nv12_coef.x;\n");
7221 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7222 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.z, nv12_coef.z;\n");
7223 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7227 /* Read from [size - size+size/2] */
7228 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, coef.y, size.y;\n");
7230 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
7231 /* Multiply by 2 since chroma components are stored in UV pixel pairs,
7232 * add 0.5 to hit the center of the pixel. */
7233 shader_addline(buffer
, "MAD texcrd.x, texcrd.x, coef.z, coef.y;\n");
7236 shader_addline(buffer
, "MAD temp.y, size.y, coef.y, size.y;\n");
7237 shader_addline(buffer
, "ADD temp.y, temp.y, -coef.y;\n");
7238 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7239 shader_addline(buffer
, "ADD temp.y, size.y, coef.y;\n");
7240 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7242 /* Read the texture, put the result into the output register. */
7243 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7244 shader_addline(buffer
, "MOV chroma.y, temp.w;\n");
7246 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7249 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, chroma.w;\n");
7254 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, coef.x;\n");
7256 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7257 shader_addline(buffer
, "MOV chroma.x, temp.w;\n");
7259 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
7260 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
7261 * values due to filtering. */
7262 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7263 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7265 /* Multiply the y coordinate by 2/3 and clamp it */
7266 shader_addline(buffer
, "MUL texcrd.y, texcrd.y, nv12_coef.x;\n");
7267 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, nv12_coef.x;\n");
7268 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7269 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7273 /* Reading from texture_rectangles is pretty straightforward, just use the unmodified
7274 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
7277 shader_addline(buffer
, "ADD temp.x, size.y, -coef.y;\n");
7278 shader_addline(buffer
, "MIN texcrd.y, texcrd.y, size.x;\n");
7279 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7286 /* Context activation is done by the caller. */
7287 static GLuint
gen_p8_shader(struct arbfp_blit_priv
*priv
,
7288 const struct wined3d_gl_info
*gl_info
, const struct arbfp_blit_type
*type
)
7291 struct wined3d_string_buffer buffer
;
7292 const char *tex_target
= arbfp_texture_target(type
->res_type
);
7294 /* This should not happen because we only use this conversion for
7295 * present blits which don't use color keying. */
7296 if (type
->use_color_key
)
7297 FIXME("Implement P8 color keying.\n");
7300 if (!string_buffer_init(&buffer
))
7302 ERR("Failed to initialize shader buffer.\n");
7306 GL_EXTCALL(glGenProgramsARB(1, &shader
));
7307 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7310 string_buffer_free(&buffer
);
7314 shader_addline(&buffer
, "!!ARBfp1.0\n");
7315 shader_addline(&buffer
, "TEMP index;\n");
7317 /* { 255/256, 0.5/255*255/256, 0, 0 } */
7318 shader_addline(&buffer
, "PARAM constants = { 0.996, 0.00195, 0, 0 };\n");
7320 /* The alpha-component contains the palette index */
7321 shader_addline(&buffer
, "TEX index, fragment.texcoord[0], texture[0], %s;\n", tex_target
);
7323 /* Scale the index by 255/256 and add a bias of '0.5' in order to sample in the middle */
7324 shader_addline(&buffer
, "MAD index.a, index.a, constants.x, constants.y;\n");
7326 /* Use the alpha-component as an index in the palette to get the final color */
7327 shader_addline(&buffer
, "TEX result.color, index.a, texture[1], 1D;\n");
7328 shader_addline(&buffer
, "END\n");
7330 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
7332 string_buffer_free(&buffer
);
7337 /* Context activation is done by the caller. */
7338 static void upload_palette(const struct wined3d_texture
*texture
, struct wined3d_context
*context
)
7340 const struct wined3d_palette
*palette
= texture
->swapchain
? texture
->swapchain
->palette
: NULL
;
7341 struct wined3d_device
*device
= texture
->resource
.device
;
7342 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
7343 struct arbfp_blit_priv
*priv
= device
->blit_priv
;
7345 if (!priv
->palette_texture
)
7346 gl_info
->gl_ops
.gl
.p_glGenTextures(1, &priv
->palette_texture
);
7348 GL_EXTCALL(glActiveTexture(GL_TEXTURE1
));
7349 gl_info
->gl_ops
.gl
.p_glBindTexture(GL_TEXTURE_1D
, priv
->palette_texture
);
7351 gl_info
->gl_ops
.gl
.p_glTexEnvi(GL_TEXTURE_ENV
, GL_TEXTURE_ENV_MODE
, GL_REPLACE
);
7353 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_MIN_FILTER
, GL_NEAREST
);
7354 /* Make sure we have discrete color levels. */
7355 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_MAG_FILTER
, GL_NEAREST
);
7356 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_WRAP_S
, GL_CLAMP_TO_EDGE
);
7357 /* TODO: avoid unneeded uploads in the future by adding some SFLAG_PALETTE_DIRTY mechanism */
7360 gl_info
->gl_ops
.gl
.p_glTexImage1D(GL_TEXTURE_1D
, 0, GL_RGB
, 256, 0, GL_BGRA
,
7361 GL_UNSIGNED_INT_8_8_8_8_REV
, palette
->colors
);
7365 static const DWORD black
;
7366 FIXME("P8 surface loaded without a palette.\n");
7367 gl_info
->gl_ops
.gl
.p_glTexImage1D(GL_TEXTURE_1D
, 0, GL_RGB
, 1, 0, GL_BGRA
,
7368 GL_UNSIGNED_INT_8_8_8_8_REV
, &black
);
7371 /* Switch back to unit 0 in which the 2D texture will be stored. */
7372 context_active_texture(context
, gl_info
, 0);
7375 /* Context activation is done by the caller. */
7376 static GLuint
gen_yuv_shader(struct arbfp_blit_priv
*priv
, const struct wined3d_gl_info
*gl_info
,
7377 const struct arbfp_blit_type
*type
)
7380 struct wined3d_string_buffer buffer
;
7381 char luminance_component
;
7383 if (type
->use_color_key
)
7384 FIXME("Implement YUV color keying.\n");
7387 if (!string_buffer_init(&buffer
))
7389 ERR("Failed to initialize shader buffer.\n");
7393 GL_EXTCALL(glGenProgramsARB(1, &shader
));
7394 checkGLcall("GL_EXTCALL(glGenProgramsARB(1, &shader))");
7395 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7396 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
7399 string_buffer_free(&buffer
);
7403 /* The YUY2 and UYVY formats contain two pixels packed into a 32 bit macropixel,
7404 * giving effectively 16 bit per pixel. The color consists of a luminance(Y) and
7405 * two chroma(U and V) values. Each macropixel has two luminance values, one for
7406 * each single pixel it contains, and one U and one V value shared between both
7409 * The data is loaded into an A8L8 texture. With YUY2, the luminance component
7410 * contains the luminance and alpha the chroma. With UYVY it is vice versa. Thus
7411 * take the format into account when generating the read swizzles
7413 * Reading the Y value is straightforward - just sample the texture. The hardware
7414 * takes care of filtering in the horizontal and vertical direction.
7416 * Reading the U and V values is harder. We have to avoid filtering horizontally,
7417 * because that would mix the U and V values of one pixel or two adjacent pixels.
7418 * Thus floor the texture coordinate and add 0.5 to get an unfiltered read,
7419 * regardless of the filtering setting. Vertical filtering works automatically
7420 * though - the U and V values of two rows are mixed nicely.
7422 * Apart of avoiding filtering issues, the code has to know which value it just
7423 * read, and where it can find the other one. To determine this, it checks if
7424 * it sampled an even or odd pixel, and shifts the 2nd read accordingly.
7426 * Handling horizontal filtering of U and V values requires reading a 2nd pair
7427 * of pixels, extracting U and V and mixing them. This is not implemented yet.
7429 * An alternative implementation idea is to load the texture as A8R8G8B8 texture,
7430 * with width / 2. This way one read gives all 3 values, finding U and V is easy
7431 * in an unfiltered situation. Finding the luminance on the other hand requires
7432 * finding out if it is an odd or even pixel. The real drawback of this approach
7433 * is filtering. This would have to be emulated completely in the shader, reading
7434 * up two 2 packed pixels in up to 2 rows and interpolating both horizontally and
7435 * vertically. Beyond that it would require adjustments to the texture handling
7436 * code to deal with the width scaling
7438 shader_addline(&buffer
, "!!ARBfp1.0\n");
7439 shader_addline(&buffer
, "TEMP luminance;\n");
7440 shader_addline(&buffer
, "TEMP temp;\n");
7441 shader_addline(&buffer
, "TEMP chroma;\n");
7442 shader_addline(&buffer
, "TEMP texcrd;\n");
7443 shader_addline(&buffer
, "TEMP texcrd2;\n");
7444 shader_addline(&buffer
, "PARAM coef = {1.0, 0.5, 2.0, 0.25};\n");
7445 shader_addline(&buffer
, "PARAM yuv_coef = {1.403, 0.344, 0.714, 1.770};\n");
7446 shader_addline(&buffer
, "PARAM size = program.local[%u];\n", ARBFP_BLIT_PARAM_SIZE
);
7448 switch (type
->fixup
)
7450 case COMPLEX_FIXUP_UYVY
:
7451 case COMPLEX_FIXUP_YUY2
:
7452 if (!gen_planar_yuv_read(&buffer
, type
, &luminance_component
))
7454 string_buffer_free(&buffer
);
7459 case COMPLEX_FIXUP_YV12
:
7460 if (!gen_yv12_read(&buffer
, type
, &luminance_component
))
7462 string_buffer_free(&buffer
);
7467 case COMPLEX_FIXUP_NV12
:
7468 if (!gen_nv12_read(&buffer
, type
, &luminance_component
))
7470 string_buffer_free(&buffer
);
7476 FIXME("Unsupported YUV fixup %#x\n", type
->fixup
);
7477 string_buffer_free(&buffer
);
7481 /* Calculate the final result. Formula is taken from
7482 * http://www.fourcc.org/fccyvrgb.php. Note that the chroma
7483 * ranges from -0.5 to 0.5
7485 shader_addline(&buffer
, "SUB chroma.xy, chroma, coef.y;\n");
7487 shader_addline(&buffer
, "MAD result.color.x, chroma.x, yuv_coef.x, luminance.%c;\n", luminance_component
);
7488 shader_addline(&buffer
, "MAD temp.x, -chroma.y, yuv_coef.y, luminance.%c;\n", luminance_component
);
7489 shader_addline(&buffer
, "MAD result.color.y, -chroma.x, yuv_coef.z, temp.x;\n");
7490 shader_addline(&buffer
, "MAD result.color.z, chroma.y, yuv_coef.w, luminance.%c;\n", luminance_component
);
7491 shader_addline(&buffer
, "END\n");
7493 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
7495 string_buffer_free(&buffer
);
7500 /* Context activation is done by the caller. */
7501 static GLuint
arbfp_gen_plain_shader(struct arbfp_blit_priv
*priv
,
7502 const struct wined3d_gl_info
*gl_info
, const struct arbfp_blit_type
*type
)
7505 struct wined3d_string_buffer buffer
;
7506 const char *tex_target
= arbfp_texture_target(type
->res_type
);
7509 if (!string_buffer_init(&buffer
))
7511 ERR("Failed to initialize shader buffer.\n");
7515 GL_EXTCALL(glGenProgramsARB(1, &shader
));
7518 string_buffer_free(&buffer
);
7521 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7523 shader_addline(&buffer
, "!!ARBfp1.0\n");
7525 if (type
->use_color_key
)
7527 shader_addline(&buffer
, "TEMP color;\n");
7528 shader_addline(&buffer
, "TEMP compare;\n");
7529 shader_addline(&buffer
, "PARAM color_key = program.local[%u];\n", ARBFP_BLIT_PARAM_COLOR_KEY
);
7530 shader_addline(&buffer
, "TEX color, fragment.texcoord[0], texture[0], %s;\n", tex_target
);
7531 shader_addline(&buffer
, "SUB compare, color, color_key;\n");
7532 shader_addline(&buffer
, "DP4 compare.r, compare, compare;\n");
7533 shader_addline(&buffer
, "SGE compare, -compare.r, 0.0;\n");
7534 shader_addline(&buffer
, "KIL -compare;\n");
7535 shader_addline(&buffer
, "MOV result.color, color;\n");
7539 shader_addline(&buffer
, "TEX result.color, fragment.texcoord[0], texture[0], %s;\n", tex_target
);
7542 shader_addline(&buffer
, "END\n");
7544 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
7546 string_buffer_free(&buffer
);
7551 /* Context activation is done by the caller. */
7552 static HRESULT
arbfp_blit_set(void *blit_priv
, struct wined3d_context
*context
, const struct wined3d_surface
*surface
,
7553 const struct wined3d_color_key
*color_key
)
7556 float size
[4] = {(float) surface
->pow2Width
, (float) surface
->pow2Height
, 1.0f
, 1.0f
};
7557 struct arbfp_blit_priv
*priv
= blit_priv
;
7558 enum complex_fixup fixup
;
7559 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
7560 GLenum gl_texture_type
= surface
->container
->target
;
7561 struct wine_rb_entry
*entry
;
7562 struct arbfp_blit_type type
;
7563 struct arbfp_blit_desc
*desc
;
7564 struct wined3d_color float_color_key
;
7566 if (is_complex_fixup(surface
->resource
.format
->color_fixup
))
7567 fixup
= get_complex_fixup(surface
->resource
.format
->color_fixup
);
7569 fixup
= COMPLEX_FIXUP_NONE
;
7571 switch (gl_texture_type
)
7574 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_1D
;
7578 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_2D
;
7582 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_3D
;
7585 case GL_TEXTURE_CUBE_MAP_ARB
:
7586 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_CUBE
;
7589 case GL_TEXTURE_RECTANGLE_ARB
:
7590 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_RECT
;
7594 ERR("Unexpected GL texture type %x.\n", gl_texture_type
);
7595 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_2D
;
7598 type
.use_color_key
= !!color_key
;
7601 entry
= wine_rb_get(&priv
->shaders
, &type
);
7604 desc
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_blit_desc
, entry
);
7605 shader
= desc
->shader
;
7611 case COMPLEX_FIXUP_NONE
:
7612 if (!is_identity_fixup(surface
->resource
.format
->color_fixup
))
7613 FIXME("Implement support for sign or swizzle fixups.\n");
7614 shader
= arbfp_gen_plain_shader(priv
, gl_info
, &type
);
7617 case COMPLEX_FIXUP_P8
:
7618 shader
= gen_p8_shader(priv
, gl_info
, &type
);
7621 case COMPLEX_FIXUP_YUY2
:
7622 case COMPLEX_FIXUP_UYVY
:
7623 case COMPLEX_FIXUP_YV12
:
7624 case COMPLEX_FIXUP_NV12
:
7625 shader
= gen_yuv_shader(priv
, gl_info
, &type
);
7631 FIXME("Unsupported complex fixup %#x, not setting a shader\n", fixup
);
7635 desc
= HeapAlloc(GetProcessHeap(), 0, sizeof(*desc
));
7640 desc
->shader
= shader
;
7641 if (wine_rb_put(&priv
->shaders
, &desc
->type
, &desc
->entry
) == -1)
7644 ERR("Out of memory\n");
7645 GL_EXTCALL(glDeleteProgramsARB(1, &shader
));
7646 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader))");
7647 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, 0));
7648 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, 0)");
7649 HeapFree(GetProcessHeap(), 0, desc
);
7650 return E_OUTOFMEMORY
;
7654 if (fixup
== COMPLEX_FIXUP_P8
)
7655 upload_palette(surface
->container
, context
);
7657 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
7658 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
7659 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7660 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
7661 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARBFP_BLIT_PARAM_SIZE
, size
));
7662 checkGLcall("glProgramLocalParameter4fvARB");
7663 if (type
.use_color_key
)
7665 wined3d_format_convert_color_to_float(surface
->resource
.format
, NULL
,
7666 surface
->container
->async
.src_blt_color_key
.color_space_high_value
, &float_color_key
);
7667 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
7668 ARBFP_BLIT_PARAM_COLOR_KEY
, &float_color_key
.r
));
7669 checkGLcall("glProgramLocalParameter4fvARB");
7675 /* Context activation is done by the caller. */
7676 static void arbfp_blit_unset(const struct wined3d_gl_info
*gl_info
)
7678 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
7679 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
7682 static BOOL
arbfp_blit_supported(const struct wined3d_gl_info
*gl_info
,
7683 const struct wined3d_d3d_info
*d3d_info
, enum wined3d_blit_op blit_op
,
7684 const RECT
*src_rect
, DWORD src_usage
, enum wined3d_pool src_pool
, const struct wined3d_format
*src_format
,
7685 const RECT
*dst_rect
, DWORD dst_usage
, enum wined3d_pool dst_pool
, const struct wined3d_format
*dst_format
)
7687 enum complex_fixup src_fixup
;
7689 if (!gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
7694 case WINED3D_BLIT_OP_COLOR_BLIT_CKEY
:
7695 if (!d3d_info
->shader_color_key
)
7697 /* The conversion modifies the alpha channel so the color key might no longer match. */
7698 TRACE("Color keying not supported with converted textures.\n");
7701 case WINED3D_BLIT_OP_COLOR_BLIT
:
7705 TRACE("Unsupported blit_op=%d\n", blit_op
);
7709 if (src_pool
== WINED3D_POOL_SYSTEM_MEM
|| dst_pool
== WINED3D_POOL_SYSTEM_MEM
)
7712 src_fixup
= get_complex_fixup(src_format
->color_fixup
);
7713 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
7715 TRACE("Checking support for fixup:\n");
7716 dump_color_fixup_desc(src_format
->color_fixup
);
7719 if (!is_identity_fixup(dst_format
->color_fixup
))
7721 TRACE("Destination fixups are not supported\n");
7725 if (is_identity_fixup(src_format
->color_fixup
))
7731 /* We only support YUV conversions. */
7732 if (!is_complex_fixup(src_format
->color_fixup
))
7734 TRACE("[FAILED]\n");
7740 case COMPLEX_FIXUP_YUY2
:
7741 case COMPLEX_FIXUP_UYVY
:
7742 case COMPLEX_FIXUP_YV12
:
7743 case COMPLEX_FIXUP_NV12
:
7744 case COMPLEX_FIXUP_P8
:
7749 FIXME("Unsupported YUV fixup %#x\n", src_fixup
);
7750 TRACE("[FAILED]\n");
7755 static void arbfp_blit_surface(struct wined3d_device
*device
, DWORD filter
,
7756 struct wined3d_surface
*src_surface
, const RECT
*src_rect_in
,
7757 struct wined3d_surface
*dst_surface
, const RECT
*dst_rect_in
,
7758 const struct wined3d_color_key
*color_key
)
7760 struct wined3d_context
*context
;
7761 RECT src_rect
= *src_rect_in
;
7762 RECT dst_rect
= *dst_rect_in
;
7764 /* Activate the destination context, set it up for blitting */
7765 context
= context_acquire(device
, dst_surface
);
7767 /* Now load the surface */
7768 if (wined3d_settings
.offscreen_rendering_mode
!= ORM_FBO
7769 && (src_surface
->locations
& (WINED3D_LOCATION_TEXTURE_RGB
| WINED3D_LOCATION_DRAWABLE
))
7770 == WINED3D_LOCATION_DRAWABLE
7771 && !wined3d_resource_is_offscreen(&src_surface
->container
->resource
))
7773 /* Without FBO blits transferring from the drawable to the texture is
7774 * expensive, because we have to flip the data in sysmem. Since we can
7775 * flip in the blitter, we don't actually need that flip anyway. So we
7776 * use the surface's texture as scratch texture, and flip the source
7777 * rectangle instead. */
7778 surface_load_fb_texture(src_surface
, FALSE
);
7780 src_rect
.top
= src_surface
->resource
.height
- src_rect
.top
;
7781 src_rect
.bottom
= src_surface
->resource
.height
- src_rect
.bottom
;
7784 wined3d_texture_load(src_surface
->container
, context
, FALSE
);
7786 context_apply_blit_state(context
, device
);
7788 if (!wined3d_resource_is_offscreen(&dst_surface
->container
->resource
))
7789 surface_translate_drawable_coords(dst_surface
, context
->win_handle
, &dst_rect
);
7791 arbfp_blit_set(device
->blit_priv
, context
, src_surface
, color_key
);
7793 /* Draw a textured quad */
7794 draw_textured_quad(src_surface
, context
, &src_rect
, &dst_rect
, filter
);
7796 /* Leave the opengl state valid for blitting */
7797 arbfp_blit_unset(context
->gl_info
);
7799 if (wined3d_settings
.strict_draw_ordering
7800 || (dst_surface
->container
->swapchain
7801 && (dst_surface
->container
->swapchain
->front_buffer
== dst_surface
->container
)))
7802 context
->gl_info
->gl_ops
.gl
.p_glFlush(); /* Flush to ensure ordering across contexts. */
7804 context_release(context
);
7806 surface_validate_location(dst_surface
, dst_surface
->container
->resource
.draw_binding
);
7807 surface_invalidate_location(dst_surface
, ~dst_surface
->container
->resource
.draw_binding
);
7810 static HRESULT
arbfp_blit_color_fill(struct wined3d_device
*device
, struct wined3d_surface
*dst_surface
,
7811 const RECT
*dst_rect
, const struct wined3d_color
*color
)
7813 FIXME("Color filling not implemented by arbfp_blit\n");
7814 return WINED3DERR_INVALIDCALL
;
7817 static HRESULT
arbfp_blit_depth_fill(struct wined3d_device
*device
,
7818 struct wined3d_surface
*surface
, const RECT
*rect
, float depth
)
7820 FIXME("Depth filling not implemented by arbfp_blit.\n");
7821 return WINED3DERR_INVALIDCALL
;
7824 const struct blit_shader arbfp_blit
= {
7829 arbfp_blit_supported
,
7830 arbfp_blit_color_fill
,
7831 arbfp_blit_depth_fill
,