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
);
42 static BOOL
shader_is_pshader_version(enum wined3d_shader_type type
)
44 return type
== WINED3D_SHADER_TYPE_PIXEL
;
47 static BOOL
shader_is_vshader_version(enum wined3d_shader_type type
)
49 return type
== WINED3D_SHADER_TYPE_VERTEX
;
52 /* Extract a line. Note that this modifies the source string. */
53 static char *get_line(char **ptr
)
58 if (!(q
= strstr(p
, "\n")))
70 static void shader_arb_dump_program_source(const char *source
)
73 char *ptr
, *line
, *tmp
;
75 source_size
= strlen(source
) + 1;
76 tmp
= HeapAlloc(GetProcessHeap(), 0, source_size
);
79 ERR("Failed to allocate %u bytes for shader source.\n", source_size
);
82 memcpy(tmp
, source
, source_size
);
85 while ((line
= get_line(&ptr
))) FIXME(" %s\n", line
);
88 HeapFree(GetProcessHeap(), 0, tmp
);
102 static const char *arb_get_helper_value(enum wined3d_shader_type shader
, enum arb_helper_value value
)
104 if (shader
== WINED3D_SHADER_TYPE_GEOMETRY
)
106 ERR("Geometry shaders are unsupported\n");
110 if (shader
== WINED3D_SHADER_TYPE_PIXEL
)
114 case ARB_ZERO
: return "ps_helper_const.x";
115 case ARB_ONE
: return "ps_helper_const.y";
116 case ARB_TWO
: return "coefmul.x";
117 case ARB_0001
: return "ps_helper_const.xxxy";
118 case ARB_EPS
: return "ps_helper_const.z";
126 case ARB_ZERO
: return "helper_const.x";
127 case ARB_ONE
: return "helper_const.y";
128 case ARB_TWO
: return "helper_const.z";
129 case ARB_EPS
: return "helper_const.w";
130 case ARB_0001
: return "helper_const.xxxy";
131 case ARB_VS_REL_OFFSET
: return "rel_addr_const.y";
134 FIXME("Unmanaged %s shader helper constant requested: %u\n",
135 shader
== WINED3D_SHADER_TYPE_PIXEL
? "pixel" : "vertex", value
);
138 case ARB_ZERO
: return "0.0";
139 case ARB_ONE
: return "1.0";
140 case ARB_TWO
: return "2.0";
141 case ARB_0001
: return "{0.0, 0.0, 0.0, 1.0}";
142 case ARB_EPS
: return "1e-8";
143 default: return "bad";
147 static inline BOOL
ffp_clip_emul(const struct wined3d_context
*context
)
149 return context
->lowest_disabled_stage
< 7;
152 /* ARB_program_shader private data */
171 struct wined3d_shader_loop_control loop_control
;
175 struct arb_ps_np2fixup_info
177 struct ps_np2fixup_info super
;
178 /* For ARB we need an offset value:
179 * With both GLSL and ARB mode the NP2 fixup information (the texture dimensions) are stored in a
180 * consecutive way (GLSL uses a uniform array). Since ARB doesn't know the notion of a "standalone"
181 * array we need an offset to the index inside the program local parameter array. */
185 struct arb_ps_compile_args
187 struct ps_compile_args super
;
189 WORD clip
; /* only a boolean, use a WORD for alignment */
190 unsigned char loop_ctrl
[MAX_CONST_I
][3];
193 struct stb_const_desc
195 unsigned char texunit
;
199 struct arb_ps_compiled_shader
201 struct arb_ps_compile_args args
;
202 struct arb_ps_np2fixup_info np2fixup_info
;
203 struct stb_const_desc bumpenvmatconst
[MAX_TEXTURES
];
204 struct stb_const_desc luminanceconst
[MAX_TEXTURES
];
205 UINT int_consts
[MAX_CONST_I
];
208 unsigned char numbumpenvmatconsts
;
212 struct arb_vs_compile_args
214 struct vs_compile_args super
;
220 unsigned char clip_texcoord
;
221 unsigned char clipplane_mask
;
223 DWORD boolclip_compare
;
228 unsigned char samplers
[4];
229 DWORD samplers_compare
;
231 unsigned char loop_ctrl
[MAX_CONST_I
][3];
234 struct arb_vs_compiled_shader
236 struct arb_vs_compile_args args
;
238 UINT int_consts
[MAX_CONST_I
];
240 char need_color_unclamp
;
244 struct recorded_instruction
246 struct wined3d_shader_instruction ins
;
250 struct shader_arb_ctx_priv
255 /* plain GL_ARB_vertex_program or GL_ARB_fragment_program */
257 /* GL_NV_vertex_progam2_option or GL_NV_fragment_program_option */
259 /* GL_NV_vertex_program3 or GL_NV_fragment_program2 */
263 const struct arb_vs_compile_args
*cur_vs_args
;
264 const struct arb_ps_compile_args
*cur_ps_args
;
265 const struct arb_ps_compiled_shader
*compiled_fprog
;
266 const struct arb_vs_compiled_shader
*compiled_vprog
;
267 struct arb_ps_np2fixup_info
*cur_np2fixup_info
;
268 struct list control_frames
;
272 unsigned int num_loops
, loop_depth
, num_ifcs
;
274 BOOL ps_post_process
;
276 unsigned int vs_clipplanes
;
280 /* For 3.0 vertex shaders */
281 const char *vs_output
[MAX_REG_OUTPUT
];
282 /* For 2.x and earlier vertex shaders */
283 const char *texcrd_output
[8], *color_output
[2], *fog_output
;
285 /* 3.0 pshader input for compatibility with fixed function */
286 const char *ps_input
[MAX_REG_INPUT
];
291 struct wined3d_shader_signature_element
*sig
;
293 struct wine_rb_entry entry
;
296 struct arb_pshader_private
{
297 struct arb_ps_compiled_shader
*gl_shaders
;
298 UINT num_gl_shaders
, shader_array_size
;
299 DWORD input_signature_idx
;
300 DWORD clipplane_emulation
;
304 struct arb_vshader_private
{
305 struct arb_vs_compiled_shader
*gl_shaders
;
306 UINT num_gl_shaders
, shader_array_size
;
310 struct shader_arb_priv
312 GLuint current_vprogram_id
;
313 GLuint current_fprogram_id
;
314 const struct arb_ps_compiled_shader
*compiled_fprog
;
315 const struct arb_vs_compiled_shader
*compiled_vprog
;
316 GLuint depth_blt_vprogram_id
;
317 GLuint depth_blt_fprogram_id_full
[tex_type_count
];
318 GLuint depth_blt_fprogram_id_masked
[tex_type_count
];
319 BOOL use_arbfp_fixed_func
;
320 struct wine_rb_tree fragment_shaders
;
321 BOOL last_ps_const_clamped
;
322 BOOL last_vs_color_unclamp
;
324 struct wine_rb_tree signature_tree
;
327 unsigned int highest_dirty_ps_const
, highest_dirty_vs_const
;
328 char *vshader_const_dirty
, *pshader_const_dirty
;
329 const struct wined3d_context
*last_context
;
331 const struct wined3d_vertex_pipe_ops
*vertex_pipe
;
332 const struct fragment_pipeline
*fragment_pipe
;
333 BOOL ffp_proj_control
;
336 /* Context activation for state handlers is done by the caller. */
338 static BOOL
need_rel_addr_const(const struct arb_vshader_private
*shader_data
,
339 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
)
341 if (shader_data
->rel_offset
) return TRUE
;
342 if (!reg_maps
->usesmova
) return FALSE
;
343 return !gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
];
346 /* Returns TRUE if result.clip from GL_NV_vertex_program2 should be used and FALSE otherwise */
347 static inline BOOL
use_nv_clip(const struct wined3d_gl_info
*gl_info
)
349 return gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]
350 && !(gl_info
->quirks
& WINED3D_QUIRK_NV_CLIP_BROKEN
);
353 static BOOL
need_helper_const(const struct arb_vshader_private
*shader_data
,
354 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
)
356 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
)) return TRUE
;
357 if (!gl_info
->supported
[NV_VERTEX_PROGRAM
]) return TRUE
; /* Need to init colors. */
358 if (gl_info
->quirks
& WINED3D_QUIRK_ARB_VS_OFFSET_LIMIT
) return TRUE
; /* Load the immval offset. */
359 if (gl_info
->quirks
& WINED3D_QUIRK_SET_TEXCOORD_W
) return TRUE
; /* Have to init texcoords. */
360 if (!use_nv_clip(gl_info
)) return TRUE
; /* Init the clip texcoord */
361 if (reg_maps
->usesnrm
) return TRUE
; /* 0.0 */
362 if (reg_maps
->usespow
) return TRUE
; /* EPS, 0.0 and 1.0 */
363 if (reg_maps
->fog
) return TRUE
; /* Clamping fog coord, 0.0 and 1.0 */
367 static unsigned int reserved_vs_const(const struct arb_vshader_private
*shader_data
,
368 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
)
370 unsigned int ret
= 1;
371 /* We use one PARAM for the pos fixup, and in some cases one to load
372 * some immediate values into the shader. */
373 if (need_helper_const(shader_data
, reg_maps
, gl_info
)) ++ret
;
374 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
)) ++ret
;
378 /* Loads floating point constants into the currently set ARB_vertex/fragment_program.
379 * When constant_list == NULL, it will load all the constants.
381 * @target_type should be either GL_VERTEX_PROGRAM_ARB (for vertex shaders)
382 * or GL_FRAGMENT_PROGRAM_ARB (for pixel shaders)
384 /* Context activation is done by the caller. */
385 static unsigned int shader_arb_load_constantsF(const struct wined3d_shader
*shader
,
386 const struct wined3d_gl_info
*gl_info
, GLuint target_type
, unsigned int max_constants
,
387 const float *constants
, char *dirty_consts
)
389 struct wined3d_shader_lconst
*lconst
;
393 if (TRACE_ON(d3d_constants
))
395 for(i
= 0; i
< max_constants
; i
++) {
396 if(!dirty_consts
[i
]) continue;
397 TRACE_(d3d_constants
)("Loading constants %i: %f, %f, %f, %f\n", i
,
398 constants
[i
* 4 + 0], constants
[i
* 4 + 1],
399 constants
[i
* 4 + 2], constants
[i
* 4 + 3]);
405 /* In 1.X pixel shaders constants are implicitly clamped in the range [-1;1] */
406 if (target_type
== GL_FRAGMENT_PROGRAM_ARB
&& shader
->reg_maps
.shader_version
.major
== 1)
409 /* ps 1.x supports only 8 constants, clamp only those. When switching between 1.x and higher
410 * shaders, the first 8 constants are marked dirty for reload
412 for(; i
< min(8, max_constants
); i
++) {
413 if(!dirty_consts
[i
]) continue;
417 if (constants
[j
+ 0] > 1.0f
) lcl_const
[0] = 1.0f
;
418 else if (constants
[j
+ 0] < -1.0f
) lcl_const
[0] = -1.0f
;
419 else lcl_const
[0] = constants
[j
+ 0];
421 if (constants
[j
+ 1] > 1.0f
) lcl_const
[1] = 1.0f
;
422 else if (constants
[j
+ 1] < -1.0f
) lcl_const
[1] = -1.0f
;
423 else lcl_const
[1] = constants
[j
+ 1];
425 if (constants
[j
+ 2] > 1.0f
) lcl_const
[2] = 1.0f
;
426 else if (constants
[j
+ 2] < -1.0f
) lcl_const
[2] = -1.0f
;
427 else lcl_const
[2] = constants
[j
+ 2];
429 if (constants
[j
+ 3] > 1.0f
) lcl_const
[3] = 1.0f
;
430 else if (constants
[j
+ 3] < -1.0f
) lcl_const
[3] = -1.0f
;
431 else lcl_const
[3] = constants
[j
+ 3];
433 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, i
, lcl_const
));
436 /* If further constants are dirty, reload them without clamping.
438 * The alternative is not to touch them, but then we cannot reset the dirty constant count
439 * to zero. That's bad for apps that only use PS 1.x shaders, because in that case the code
440 * above would always re-check the first 8 constants since max_constant remains at the init
445 if (gl_info
->supported
[EXT_GPU_PROGRAM_PARAMETERS
])
447 /* TODO: Benchmark if we're better of with finding the dirty constants ourselves,
448 * or just reloading *all* constants at once
450 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type, i, max_constants, constants + (i * 4)));
452 for(; i
< max_constants
; i
++) {
453 if(!dirty_consts
[i
]) continue;
455 /* Find the next block of dirty constants */
458 for(i
++; (i
< max_constants
) && dirty_consts
[i
]; i
++) {
462 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type
, j
, i
- j
, constants
+ (j
* 4)));
465 for(; i
< max_constants
; i
++) {
466 if(dirty_consts
[i
]) {
468 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, i
, constants
+ (i
* 4)));
472 checkGLcall("glProgramEnvParameter4fvARB()");
474 /* Load immediate constants */
475 if (shader
->load_local_constsF
)
477 if (TRACE_ON(d3d_shader
))
479 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
481 GLfloat
* values
= (GLfloat
*)lconst
->value
;
482 TRACE_(d3d_constants
)("Loading local constants %i: %f, %f, %f, %f\n", lconst
->idx
,
483 values
[0], values
[1], values
[2], values
[3]);
486 /* Immediate constants are clamped for 1.X shaders at loading times */
488 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
490 dirty_consts
[lconst
->idx
] = 1; /* Dirtify so the non-immediate constant overwrites it next time */
491 ret
= max(ret
, lconst
->idx
+ 1);
492 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, lconst
->idx
, (GLfloat
*)lconst
->value
));
494 checkGLcall("glProgramEnvParameter4fvARB()");
495 return ret
; /* The loaded immediate constants need reloading for the next shader */
497 return 0; /* No constants are dirty now */
501 /* Loads the texture dimensions for NP2 fixup into the currently set
502 * ARB_[vertex/fragment]_programs. */
503 static void shader_arb_load_np2fixup_constants(const struct arb_ps_np2fixup_info
*fixup
,
504 const struct wined3d_gl_info
*gl_info
, const struct wined3d_state
*state
)
506 GLfloat np2fixup_constants
[4 * MAX_FRAGMENT_SAMPLERS
];
507 WORD active
= fixup
->super
.active
;
513 for (i
= 0; active
; active
>>= 1, ++i
)
515 const struct wined3d_texture
*tex
= state
->textures
[i
];
516 unsigned char idx
= fixup
->super
.idx
[i
];
517 GLfloat
*tex_dim
= &np2fixup_constants
[(idx
>> 1) * 4];
524 ERR("Nonexistent texture is flagged for NP2 texcoord fixup.\n");
530 tex_dim
[2] = tex
->pow2_matrix
[0];
531 tex_dim
[3] = tex
->pow2_matrix
[5];
535 tex_dim
[0] = tex
->pow2_matrix
[0];
536 tex_dim
[1] = tex
->pow2_matrix
[5];
540 for (i
= 0; i
< fixup
->super
.num_consts
; ++i
)
542 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
543 fixup
->offset
+ i
, &np2fixup_constants
[i
* 4]));
547 /* Context activation is done by the caller. */
548 static void shader_arb_ps_local_constants(const struct arb_ps_compiled_shader
*gl_shader
,
549 const struct wined3d_context
*context
, const struct wined3d_state
*state
, UINT rt_height
)
551 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
554 for(i
= 0; i
< gl_shader
->numbumpenvmatconsts
; i
++)
556 int texunit
= gl_shader
->bumpenvmatconst
[i
].texunit
;
558 /* The state manager takes care that this function is always called if the bump env matrix changes */
559 const float *data
= (const float *)&state
->texture_states
[texunit
][WINED3D_TSS_BUMPENV_MAT00
];
560 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
561 gl_shader
->bumpenvmatconst
[i
].const_num
, data
));
563 if (gl_shader
->luminanceconst
[i
].const_num
!= WINED3D_CONST_NUM_UNUSED
)
565 /* WINED3D_TSS_BUMPENVLSCALE and WINED3D_TSS_BUMPENVLOFFSET are next to each other.
566 * point gl to the scale, and load 4 floats. x = scale, y = offset, z and w are junk, we
567 * don't care about them. The pointers are valid for sure because the stateblock is bigger.
568 * (they're WINED3D_TSS_TEXTURETRANSFORMFLAGS and WINED3D_TSS_ADDRESSW, so most likely 0 or NaN
570 const float *scale
= (const float *)&state
->texture_states
[texunit
][WINED3D_TSS_BUMPENV_LSCALE
];
571 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
572 gl_shader
->luminanceconst
[i
].const_num
, scale
));
575 checkGLcall("Load bumpmap consts");
577 if(gl_shader
->ycorrection
!= WINED3D_CONST_NUM_UNUSED
)
579 /* ycorrection.x: Backbuffer height(onscreen) or 0(offscreen).
580 * ycorrection.y: -1.0(onscreen), 1.0(offscreen)
585 val
[0] = context
->render_offscreen
? 0.0f
: (float) rt_height
;
586 val
[1] = context
->render_offscreen
? 1.0f
: -1.0f
;
589 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, gl_shader
->ycorrection
, val
));
590 checkGLcall("y correction loading");
593 if (!gl_shader
->num_int_consts
) return;
595 for(i
= 0; i
< MAX_CONST_I
; i
++)
597 if(gl_shader
->int_consts
[i
] != WINED3D_CONST_NUM_UNUSED
)
600 val
[0] = (float)state
->ps_consts_i
[4 * i
];
601 val
[1] = (float)state
->ps_consts_i
[4 * i
+ 1];
602 val
[2] = (float)state
->ps_consts_i
[4 * i
+ 2];
605 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, gl_shader
->int_consts
[i
], val
));
608 checkGLcall("Load ps int consts");
611 /* Context activation is done by the caller. */
612 static void shader_arb_vs_local_constants(const struct arb_vs_compiled_shader
*gl_shader
,
613 const struct wined3d_context
*context
, const struct wined3d_state
*state
)
615 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
616 float position_fixup
[4];
619 /* Upload the position fixup */
620 shader_get_position_fixup(context
, state
, position_fixup
);
621 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, gl_shader
->pos_fixup
, position_fixup
));
623 if (!gl_shader
->num_int_consts
) return;
625 for(i
= 0; i
< MAX_CONST_I
; i
++)
627 if(gl_shader
->int_consts
[i
] != WINED3D_CONST_NUM_UNUSED
)
630 val
[0] = (float)state
->vs_consts_i
[4 * i
];
631 val
[1] = (float)state
->vs_consts_i
[4 * i
+ 1];
632 val
[2] = (float)state
->vs_consts_i
[4 * i
+ 2];
635 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, gl_shader
->int_consts
[i
], val
));
638 checkGLcall("Load vs int consts");
641 static void shader_arb_select(void *shader_priv
, struct wined3d_context
*context
,
642 const struct wined3d_state
*state
);
645 * Loads the app-supplied constants into the currently set ARB_[vertex/fragment]_programs.
647 * We only support float constants in ARB at the moment, so don't
648 * worry about the Integers or Booleans
650 /* Context activation is done by the caller (state handler). */
651 static void shader_arb_load_constants_internal(struct shader_arb_priv
*priv
,
652 struct wined3d_context
*context
, const struct wined3d_state
*state
,
653 BOOL usePixelShader
, BOOL useVertexShader
, BOOL from_shader_select
)
655 const struct wined3d_d3d_info
*d3d_info
= context
->d3d_info
;
656 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
658 if (!from_shader_select
)
660 const struct wined3d_shader
*vshader
= state
->shader
[WINED3D_SHADER_TYPE_VERTEX
];
661 const struct wined3d_shader
*pshader
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
664 && (vshader
->reg_maps
.boolean_constants
665 || (!gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]
666 && (vshader
->reg_maps
.integer_constants
& ~vshader
->reg_maps
.local_int_consts
))))
668 TRACE("bool/integer vertex shader constants potentially modified, forcing shader reselection.\n");
669 shader_arb_select(priv
, context
, state
);
672 && (pshader
->reg_maps
.boolean_constants
673 || (!gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
]
674 && (pshader
->reg_maps
.integer_constants
& ~pshader
->reg_maps
.local_int_consts
))))
676 TRACE("bool/integer pixel shader constants potentially modified, forcing shader reselection.\n");
677 shader_arb_select(priv
, context
, state
);
681 if (context
!= priv
->last_context
)
683 memset(priv
->vshader_const_dirty
, 1,
684 sizeof(*priv
->vshader_const_dirty
) * d3d_info
->limits
.vs_uniform_count
);
685 priv
->highest_dirty_vs_const
= d3d_info
->limits
.vs_uniform_count
;
687 memset(priv
->pshader_const_dirty
, 1,
688 sizeof(*priv
->pshader_const_dirty
) * d3d_info
->limits
.ps_uniform_count
);
689 priv
->highest_dirty_ps_const
= d3d_info
->limits
.ps_uniform_count
;
691 priv
->last_context
= context
;
696 const struct wined3d_shader
*vshader
= state
->shader
[WINED3D_SHADER_TYPE_VERTEX
];
697 const struct arb_vs_compiled_shader
*gl_shader
= priv
->compiled_vprog
;
699 /* Load DirectX 9 float constants for vertex shader */
700 priv
->highest_dirty_vs_const
= shader_arb_load_constantsF(vshader
, gl_info
, GL_VERTEX_PROGRAM_ARB
,
701 priv
->highest_dirty_vs_const
, state
->vs_consts_f
, priv
->vshader_const_dirty
);
702 shader_arb_vs_local_constants(gl_shader
, context
, state
);
707 const struct wined3d_shader
*pshader
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
708 const struct arb_ps_compiled_shader
*gl_shader
= priv
->compiled_fprog
;
709 UINT rt_height
= state
->fb
->render_targets
[0]->resource
.height
;
711 /* Load DirectX 9 float constants for pixel shader */
712 priv
->highest_dirty_ps_const
= shader_arb_load_constantsF(pshader
, gl_info
, GL_FRAGMENT_PROGRAM_ARB
,
713 priv
->highest_dirty_ps_const
, state
->ps_consts_f
, priv
->pshader_const_dirty
);
714 shader_arb_ps_local_constants(gl_shader
, context
, state
, rt_height
);
716 if (context
->constant_update_mask
& WINED3D_SHADER_CONST_PS_NP2_FIXUP
)
717 shader_arb_load_np2fixup_constants(&gl_shader
->np2fixup_info
, gl_info
, state
);
721 static void shader_arb_load_constants(void *shader_priv
, struct wined3d_context
*context
,
722 const struct wined3d_state
*state
)
724 BOOL vs
= use_vs(state
);
725 BOOL ps
= use_ps(state
);
727 shader_arb_load_constants_internal(shader_priv
, context
, state
, ps
, vs
, FALSE
);
730 static void shader_arb_update_float_vertex_constants(struct wined3d_device
*device
, UINT start
, UINT count
)
732 struct wined3d_context
*context
= context_get_current();
733 struct shader_arb_priv
*priv
= device
->shader_priv
;
736 for (i
= 0; i
< device
->context_count
; ++i
)
738 device
->contexts
[i
]->constant_update_mask
|= WINED3D_SHADER_CONST_VS_F
;
741 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
742 * context. On a context switch the old context will be fully dirtified */
743 if (!context
|| context
->swapchain
->device
!= device
) return;
745 memset(priv
->vshader_const_dirty
+ start
, 1, sizeof(*priv
->vshader_const_dirty
) * count
);
746 priv
->highest_dirty_vs_const
= max(priv
->highest_dirty_vs_const
, start
+ count
);
749 static void shader_arb_update_float_pixel_constants(struct wined3d_device
*device
, UINT start
, UINT count
)
751 struct wined3d_context
*context
= context_get_current();
752 struct shader_arb_priv
*priv
= device
->shader_priv
;
755 for (i
= 0; i
< device
->context_count
; ++i
)
757 device
->contexts
[i
]->constant_update_mask
|= WINED3D_SHADER_CONST_PS_F
;
760 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
761 * context. On a context switch the old context will be fully dirtified */
762 if (!context
|| context
->swapchain
->device
!= device
) return;
764 memset(priv
->pshader_const_dirty
+ start
, 1, sizeof(*priv
->pshader_const_dirty
) * count
);
765 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, start
+ count
);
768 static void shader_arb_append_imm_vec4(struct wined3d_shader_buffer
*buffer
, const float *values
)
772 wined3d_ftoa(values
[0], str
[0]);
773 wined3d_ftoa(values
[1], str
[1]);
774 wined3d_ftoa(values
[2], str
[2]);
775 wined3d_ftoa(values
[3], str
[3]);
776 shader_addline(buffer
, "{%s, %s, %s, %s}", str
[0], str
[1], str
[2], str
[3]);
779 /* Generate the variable & register declarations for the ARB_vertex_program output target */
780 static void shader_generate_arb_declarations(const struct wined3d_shader
*shader
,
781 const struct wined3d_shader_reg_maps
*reg_maps
, struct wined3d_shader_buffer
*buffer
,
782 const struct wined3d_gl_info
*gl_info
, DWORD
*num_clipplanes
,
783 const struct shader_arb_ctx_priv
*ctx
)
786 char pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
787 const struct wined3d_shader_lconst
*lconst
;
788 unsigned max_constantsF
;
791 /* In pixel shaders, all private constants are program local, we don't need anything
792 * from program.env. Thus we can advertise the full set of constants in pixel shaders.
793 * If we need a private constant the GL implementation will squeeze it in somewhere
795 * With vertex shaders we need the posFixup and on some GL implementations 4 helper
796 * immediate values. The posFixup is loaded using program.env for now, so always
797 * subtract one from the number of constants. If the shader uses indirect addressing,
798 * account for the helper const too because we have to declare all available d3d constants
799 * and don't know which are actually used.
803 max_constantsF
= gl_info
->limits
.arb_ps_native_constants
;
804 /* 24 is the minimum MAX_PROGRAM_ENV_PARAMETERS_ARB value. */
805 if (max_constantsF
< 24)
806 max_constantsF
= gl_info
->limits
.arb_ps_float_constants
;
810 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
811 max_constantsF
= gl_info
->limits
.arb_vs_native_constants
;
812 /* 96 is the minimum MAX_PROGRAM_ENV_PARAMETERS_ARB value.
813 * Also prevents max_constantsF from becoming less than 0 and
815 if (max_constantsF
< 96)
816 max_constantsF
= gl_info
->limits
.arb_vs_float_constants
;
818 if (reg_maps
->usesrelconstF
)
820 DWORD highest_constf
= 0, clip_limit
;
822 max_constantsF
-= reserved_vs_const(shader_data
, reg_maps
, gl_info
);
823 max_constantsF
-= count_bits(reg_maps
->integer_constants
);
824 max_constantsF
-= gl_info
->reserved_arb_constants
;
826 for (i
= 0; i
< shader
->limits
.constant_float
; ++i
)
829 DWORD shift
= i
& 0x1f;
830 if(reg_maps
->constf
[idx
] & (1 << shift
)) highest_constf
= i
;
833 if(use_nv_clip(gl_info
) && ctx
->target_version
>= NV2
)
835 if(ctx
->cur_vs_args
->super
.clip_enabled
)
836 clip_limit
= gl_info
->limits
.clipplanes
;
842 unsigned int mask
= ctx
->cur_vs_args
->clip
.boolclip
.clipplane_mask
;
843 clip_limit
= min(count_bits(mask
), 4);
845 *num_clipplanes
= min(clip_limit
, max_constantsF
- highest_constf
- 1);
846 max_constantsF
-= *num_clipplanes
;
847 if(*num_clipplanes
< clip_limit
)
849 WARN("Only %u clipplanes out of %u enabled\n", *num_clipplanes
, gl_info
->limits
.clipplanes
);
854 if (ctx
->target_version
>= NV2
) *num_clipplanes
= gl_info
->limits
.clipplanes
;
855 else *num_clipplanes
= min(gl_info
->limits
.clipplanes
, 4);
859 for (i
= 0, map
= reg_maps
->temporary
; map
; map
>>= 1, ++i
)
861 if (map
& 1) shader_addline(buffer
, "TEMP R%u;\n", i
);
864 for (i
= 0, map
= reg_maps
->address
; map
; map
>>= 1, ++i
)
866 if (map
& 1) shader_addline(buffer
, "ADDRESS A%u;\n", i
);
869 if (pshader
&& reg_maps
->shader_version
.major
== 1 && reg_maps
->shader_version
.minor
<= 3)
871 for (i
= 0, map
= reg_maps
->texcoord
; map
; map
>>= 1, ++i
)
873 if (map
& 1) shader_addline(buffer
, "TEMP T%u;\n", i
);
877 if (!shader
->load_local_constsF
)
879 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
882 value
= (const float *)lconst
->value
;
883 shader_addline(buffer
, "PARAM C%u = ", lconst
->idx
);
884 shader_arb_append_imm_vec4(buffer
, value
);
885 shader_addline(buffer
, ";\n");
889 /* After subtracting privately used constants from the hardware limit(they are loaded as
890 * local constants), make sure the shader doesn't violate the env constant limit
894 max_constantsF
= min(max_constantsF
, gl_info
->limits
.arb_ps_float_constants
);
898 max_constantsF
= min(max_constantsF
, gl_info
->limits
.arb_vs_float_constants
);
901 /* Avoid declaring more constants than needed */
902 max_constantsF
= min(max_constantsF
, shader
->limits
.constant_float
);
904 /* we use the array-based constants array if the local constants are marked for loading,
905 * because then we use indirect addressing, or when the local constant list is empty,
906 * because then we don't know if we're using indirect addressing or not. If we're hardcoding
907 * local constants do not declare the loaded constants as an array because ARB compilers usually
908 * do not optimize unused constants away
910 if (reg_maps
->usesrelconstF
)
912 /* Need to PARAM the environment parameters (constants) so we can use relative addressing */
913 shader_addline(buffer
, "PARAM C[%d] = { program.env[0..%d] };\n",
914 max_constantsF
, max_constantsF
- 1);
916 for(i
= 0; i
< max_constantsF
; i
++) {
919 mask
= 1 << (i
& 0x1f);
920 if (!shader_constant_is_local(shader
, i
) && (reg_maps
->constf
[idx
] & mask
))
922 shader_addline(buffer
, "PARAM C%d = program.env[%d];\n",i
, i
);
928 static const char * const shift_tab
[] = {
929 "dummy", /* 0 (none) */
930 "coefmul.x", /* 1 (x2) */
931 "coefmul.y", /* 2 (x4) */
932 "coefmul.z", /* 3 (x8) */
933 "coefmul.w", /* 4 (x16) */
934 "dummy", /* 5 (x32) */
935 "dummy", /* 6 (x64) */
936 "dummy", /* 7 (x128) */
937 "dummy", /* 8 (d256) */
938 "dummy", /* 9 (d128) */
939 "dummy", /* 10 (d64) */
940 "dummy", /* 11 (d32) */
941 "coefdiv.w", /* 12 (d16) */
942 "coefdiv.z", /* 13 (d8) */
943 "coefdiv.y", /* 14 (d4) */
944 "coefdiv.x" /* 15 (d2) */
947 static void shader_arb_get_write_mask(const struct wined3d_shader_instruction
*ins
,
948 const struct wined3d_shader_dst_param
*dst
, char *write_mask
)
950 char *ptr
= write_mask
;
952 if (dst
->write_mask
!= WINED3DSP_WRITEMASK_ALL
)
955 if (dst
->write_mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
956 if (dst
->write_mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
957 if (dst
->write_mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
958 if (dst
->write_mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
964 static void shader_arb_get_swizzle(const struct wined3d_shader_src_param
*param
, BOOL fixup
, char *swizzle_str
)
966 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
967 * but addressed as "rgba". To fix this we need to swap the register's x
968 * and z components. */
969 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
970 char *ptr
= swizzle_str
;
972 /* swizzle bits fields: wwzzyyxx */
973 DWORD swizzle
= param
->swizzle
;
974 DWORD swizzle_x
= swizzle
& 0x03;
975 DWORD swizzle_y
= (swizzle
>> 2) & 0x03;
976 DWORD swizzle_z
= (swizzle
>> 4) & 0x03;
977 DWORD swizzle_w
= (swizzle
>> 6) & 0x03;
979 /* If the swizzle is the default swizzle (ie, "xyzw"), we don't need to
980 * generate a swizzle string. Unless we need to our own swizzling. */
981 if (swizzle
!= WINED3DSP_NOSWIZZLE
|| fixup
)
984 if (swizzle_x
== swizzle_y
&& swizzle_x
== swizzle_z
&& swizzle_x
== swizzle_w
) {
985 *ptr
++ = swizzle_chars
[swizzle_x
];
987 *ptr
++ = swizzle_chars
[swizzle_x
];
988 *ptr
++ = swizzle_chars
[swizzle_y
];
989 *ptr
++ = swizzle_chars
[swizzle_z
];
990 *ptr
++ = swizzle_chars
[swizzle_w
];
997 static void shader_arb_request_a0(const struct wined3d_shader_instruction
*ins
, const char *src
)
999 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1000 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1002 if (!strcmp(priv
->addr_reg
, src
)) return;
1004 strcpy(priv
->addr_reg
, src
);
1005 shader_addline(buffer
, "ARL A0.x, %s;\n", src
);
1008 static void shader_arb_get_src_param(const struct wined3d_shader_instruction
*ins
,
1009 const struct wined3d_shader_src_param
*src
, unsigned int tmpreg
, char *outregstr
);
1011 static void shader_arb_get_register_name(const struct wined3d_shader_instruction
*ins
,
1012 const struct wined3d_shader_register
*reg
, char *register_name
, BOOL
*is_color
)
1014 /* oPos, oFog and oPts in D3D */
1015 static const char * const rastout_reg_names
[] = {"TMP_OUT", "TMP_FOGCOORD", "result.pointsize"};
1016 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
1017 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
1018 BOOL pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
1019 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1025 case WINED3DSPR_TEMP
:
1026 sprintf(register_name
, "R%u", reg
->idx
[0].offset
);
1029 case WINED3DSPR_INPUT
:
1032 if (reg_maps
->shader_version
.major
< 3)
1034 if (!reg
->idx
[0].offset
)
1035 strcpy(register_name
, "fragment.color.primary");
1037 strcpy(register_name
, "fragment.color.secondary");
1041 if (reg
->idx
[0].rel_addr
)
1044 shader_arb_get_src_param(ins
, reg
->idx
[0].rel_addr
, 0, rel_reg
);
1046 if (!strcmp(rel_reg
, "**aL_emul**"))
1048 DWORD idx
= ctx
->aL
+ reg
->idx
[0].offset
;
1049 if(idx
< MAX_REG_INPUT
)
1051 strcpy(register_name
, ctx
->ps_input
[idx
]);
1055 ERR("Pixel shader input register out of bounds: %u\n", idx
);
1056 sprintf(register_name
, "out_of_bounds_%u", idx
);
1059 else if (reg_maps
->input_registers
& 0x0300)
1061 /* There are two ways basically:
1063 * 1) Use the unrolling code that is used for loop emulation and unroll the loop.
1064 * That means trouble if the loop also contains a breakc or if the control values
1065 * aren't local constants.
1066 * 2) Generate an if block that checks if aL.y < 8, == 8 or == 9 and selects the
1067 * source dynamically. The trouble is that we cannot simply read aL.y because it
1068 * is an ADDRESS register. We could however push it, load .zw with a value and use
1069 * ADAC to load the condition code register and pop it again afterwards
1071 FIXME("Relative input register addressing with more than 8 registers\n");
1073 /* This is better than nothing for now */
1074 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1076 else if(ctx
->cur_ps_args
->super
.vp_mode
!= vertexshader
)
1078 /* This is problematic because we'd have to consult the ctx->ps_input strings
1079 * for where to find the varying. Some may be "0.0", others can be texcoords or
1080 * colors. This needs either a pipeline replacement to make the vertex shader feed
1081 * proper varyings, or loop unrolling
1083 * For now use the texcoords and hope for the best
1085 FIXME("Non-vertex shader varying input with indirect addressing\n");
1086 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1090 /* D3D supports indirect addressing only with aL in loop registers. The loop instruction
1091 * pulls GL_NV_fragment_program2 in
1093 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1098 if (reg
->idx
[0].offset
< MAX_REG_INPUT
)
1100 strcpy(register_name
, ctx
->ps_input
[reg
->idx
[0].offset
]);
1104 ERR("Pixel shader input register out of bounds: %u\n", reg
->idx
[0].offset
);
1105 sprintf(register_name
, "out_of_bounds_%u", reg
->idx
[0].offset
);
1112 if (ctx
->cur_vs_args
->super
.swizzle_map
& (1 << reg
->idx
[0].offset
))
1114 sprintf(register_name
, "vertex.attrib[%u]", reg
->idx
[0].offset
);
1118 case WINED3DSPR_CONST
:
1119 if (!pshader
&& reg
->idx
[0].rel_addr
)
1121 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1122 UINT rel_offset
= ctx
->target_version
== ARB
? shader_data
->rel_offset
: 0;
1125 if (reg_maps
->shader_version
.major
< 2)
1127 sprintf(rel_reg
, "A0.x");
1131 shader_arb_get_src_param(ins
, reg
->idx
[0].rel_addr
, 0, rel_reg
);
1132 if (ctx
->target_version
== ARB
)
1134 if (!strcmp(rel_reg
, "**aL_emul**"))
1138 shader_arb_request_a0(ins
, rel_reg
);
1139 sprintf(rel_reg
, "A0.x");
1144 sprintf(register_name
, "C[%u]", ctx
->aL
+ reg
->idx
[0].offset
);
1145 else if (reg
->idx
[0].offset
>= rel_offset
)
1146 sprintf(register_name
, "C[%s + %u]", rel_reg
, reg
->idx
[0].offset
- rel_offset
);
1148 sprintf(register_name
, "C[%s - %u]", rel_reg
, rel_offset
- reg
->idx
[0].offset
);
1152 if (reg_maps
->usesrelconstF
)
1153 sprintf(register_name
, "C[%u]", reg
->idx
[0].offset
);
1155 sprintf(register_name
, "C%u", reg
->idx
[0].offset
);
1159 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
1162 if (reg_maps
->shader_version
.major
== 1
1163 && reg_maps
->shader_version
.minor
<= 3)
1164 /* In ps <= 1.3, Tx is a temporary register as destination
1165 * to all instructions, and as source to most instructions.
1166 * For some instructions it is the texcoord input. Those
1167 * instructions know about the special use. */
1168 sprintf(register_name
, "T%u", reg
->idx
[0].offset
);
1170 /* In ps 1.4 and 2.x Tx is always a (read-only) varying. */
1171 sprintf(register_name
, "fragment.texcoord[%u]", reg
->idx
[0].offset
);
1175 if (reg_maps
->shader_version
.major
== 1 || ctx
->target_version
>= NV2
)
1176 sprintf(register_name
, "A%u", reg
->idx
[0].offset
);
1178 sprintf(register_name
, "A%u_SHADOW", reg
->idx
[0].offset
);
1182 case WINED3DSPR_COLOROUT
:
1183 if (ctx
->ps_post_process
&& !reg
->idx
[0].offset
)
1185 strcpy(register_name
, "TMP_COLOR");
1189 if (ctx
->cur_ps_args
->super
.srgb_correction
)
1190 FIXME("sRGB correction on higher render targets.\n");
1191 if (reg_maps
->rt_mask
> 1)
1192 sprintf(register_name
, "result.color[%u]", reg
->idx
[0].offset
);
1194 strcpy(register_name
, "result.color");
1198 case WINED3DSPR_RASTOUT
:
1199 if (reg
->idx
[0].offset
== 1)
1200 sprintf(register_name
, "%s", ctx
->fog_output
);
1202 sprintf(register_name
, "%s", rastout_reg_names
[reg
->idx
[0].offset
]);
1205 case WINED3DSPR_DEPTHOUT
:
1206 strcpy(register_name
, "result.depth");
1209 case WINED3DSPR_ATTROUT
:
1210 /* case WINED3DSPR_OUTPUT: */
1212 sprintf(register_name
, "oD[%u]", reg
->idx
[0].offset
);
1214 strcpy(register_name
, ctx
->color_output
[reg
->idx
[0].offset
]);
1217 case WINED3DSPR_TEXCRDOUT
:
1219 sprintf(register_name
, "oT[%u]", reg
->idx
[0].offset
);
1220 else if (reg_maps
->shader_version
.major
< 3)
1221 strcpy(register_name
, ctx
->texcrd_output
[reg
->idx
[0].offset
]);
1223 strcpy(register_name
, ctx
->vs_output
[reg
->idx
[0].offset
]);
1226 case WINED3DSPR_LOOP
:
1227 if(ctx
->target_version
>= NV2
)
1229 /* Pshader has an implicitly declared loop index counter A0.x that cannot be renamed */
1230 if(pshader
) sprintf(register_name
, "A0.x");
1231 else sprintf(register_name
, "aL.y");
1235 /* Unfortunately this code cannot return the value of ctx->aL here. An immediate value
1236 * would be valid, but if aL is used for indexing(its only use), there's likely an offset,
1237 * thus the result would be something like C[15 + 30], which is not valid in the ARB program
1238 * grammar. So return a marker for the emulated aL and intercept it in constant and varying
1241 sprintf(register_name
, "**aL_emul**");
1246 case WINED3DSPR_CONSTINT
:
1247 sprintf(register_name
, "I%u", reg
->idx
[0].offset
);
1250 case WINED3DSPR_MISCTYPE
:
1251 if (!reg
->idx
[0].offset
)
1252 sprintf(register_name
, "vpos");
1253 else if (reg
->idx
[0].offset
== 1)
1254 sprintf(register_name
, "fragment.facing.x");
1256 FIXME("Unknown MISCTYPE register index %u.\n", reg
->idx
[0].offset
);
1260 FIXME("Unhandled register type %#x[%u].\n", reg
->type
, reg
->idx
[0].offset
);
1261 sprintf(register_name
, "unrecognized_register[%u]", reg
->idx
[0].offset
);
1266 static void shader_arb_get_dst_param(const struct wined3d_shader_instruction
*ins
,
1267 const struct wined3d_shader_dst_param
*wined3d_dst
, char *str
)
1269 char register_name
[255];
1273 shader_arb_get_register_name(ins
, &wined3d_dst
->reg
, register_name
, &is_color
);
1274 strcpy(str
, register_name
);
1276 shader_arb_get_write_mask(ins
, wined3d_dst
, write_mask
);
1277 strcat(str
, write_mask
);
1280 static const char *shader_arb_get_fixup_swizzle(enum fixup_channel_source channel_source
)
1282 switch(channel_source
)
1284 case CHANNEL_SOURCE_ZERO
: return "0";
1285 case CHANNEL_SOURCE_ONE
: return "1";
1286 case CHANNEL_SOURCE_X
: return "x";
1287 case CHANNEL_SOURCE_Y
: return "y";
1288 case CHANNEL_SOURCE_Z
: return "z";
1289 case CHANNEL_SOURCE_W
: return "w";
1291 FIXME("Unhandled channel source %#x\n", channel_source
);
1296 static void gen_color_correction(struct wined3d_shader_buffer
*buffer
, const char *reg
,
1297 DWORD dst_mask
, const char *one
, const char *two
, struct color_fixup_desc fixup
)
1301 if (is_complex_fixup(fixup
))
1303 enum complex_fixup complex_fixup
= get_complex_fixup(fixup
);
1304 FIXME("Complex fixup (%#x) not supported\n", complex_fixup
);
1309 if (fixup
.x_source
!= CHANNEL_SOURCE_X
) mask
|= WINED3DSP_WRITEMASK_0
;
1310 if (fixup
.y_source
!= CHANNEL_SOURCE_Y
) mask
|= WINED3DSP_WRITEMASK_1
;
1311 if (fixup
.z_source
!= CHANNEL_SOURCE_Z
) mask
|= WINED3DSP_WRITEMASK_2
;
1312 if (fixup
.w_source
!= CHANNEL_SOURCE_W
) mask
|= WINED3DSP_WRITEMASK_3
;
1317 shader_addline(buffer
, "SWZ %s, %s, %s, %s, %s, %s;\n", reg
, reg
,
1318 shader_arb_get_fixup_swizzle(fixup
.x_source
), shader_arb_get_fixup_swizzle(fixup
.y_source
),
1319 shader_arb_get_fixup_swizzle(fixup
.z_source
), shader_arb_get_fixup_swizzle(fixup
.w_source
));
1323 if (fixup
.x_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_0
;
1324 if (fixup
.y_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_1
;
1325 if (fixup
.z_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_2
;
1326 if (fixup
.w_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_3
;
1332 char *ptr
= reg_mask
;
1334 if (mask
!= WINED3DSP_WRITEMASK_ALL
)
1337 if (mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
1338 if (mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
1339 if (mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
1340 if (mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
1344 shader_addline(buffer
, "MAD %s%s, %s, %s, -%s;\n", reg
, reg_mask
, reg
, two
, one
);
1348 static const char *shader_arb_get_modifier(const struct wined3d_shader_instruction
*ins
)
1351 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1352 if (!ins
->dst_count
) return "";
1354 mod
= ins
->dst
[0].modifiers
;
1356 /* Silently ignore PARTIALPRECISION if its not supported */
1357 if(priv
->target_version
== ARB
) mod
&= ~WINED3DSPDM_PARTIALPRECISION
;
1359 if(mod
& WINED3DSPDM_MSAMPCENTROID
)
1361 FIXME("Unhandled modifier WINED3DSPDM_MSAMPCENTROID\n");
1362 mod
&= ~WINED3DSPDM_MSAMPCENTROID
;
1367 case WINED3DSPDM_SATURATE
| WINED3DSPDM_PARTIALPRECISION
:
1370 case WINED3DSPDM_SATURATE
:
1373 case WINED3DSPDM_PARTIALPRECISION
:
1380 FIXME("Unknown modifiers 0x%08x\n", mod
);
1385 #define TEX_PROJ 0x1
1386 #define TEX_BIAS 0x2
1388 #define TEX_DERIV 0x10
1390 static void shader_hw_sample(const struct wined3d_shader_instruction
*ins
, DWORD sampler_idx
,
1391 const char *dst_str
, const char *coord_reg
, WORD flags
, const char *dsx
, const char *dsy
)
1393 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1394 DWORD sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[sampler_idx
];
1395 const char *tex_type
;
1396 BOOL np2_fixup
= FALSE
;
1397 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1399 BOOL pshader
= shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
1400 const struct wined3d_shader
*shader
;
1401 const struct wined3d_device
*device
;
1402 const struct wined3d_gl_info
*gl_info
;
1404 /* D3D vertex shader sampler IDs are vertex samplers(0-3), not global d3d samplers */
1405 if(!pshader
) sampler_idx
+= MAX_FRAGMENT_SAMPLERS
;
1407 switch(sampler_type
) {
1413 shader
= ins
->ctx
->shader
;
1414 device
= shader
->device
;
1415 gl_info
= &device
->adapter
->gl_info
;
1417 if (pshader
&& priv
->cur_ps_args
->super
.np2_fixup
& (1 << sampler_idx
)
1418 && gl_info
->supported
[ARB_TEXTURE_RECTANGLE
])
1422 if (shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
))
1424 if (priv
->cur_np2fixup_info
->super
.active
& (1 << sampler_idx
))
1426 if (flags
) FIXME("Only ordinary sampling from NP2 textures is supported.\n");
1427 else np2_fixup
= TRUE
;
1432 case WINED3DSTT_VOLUME
:
1436 case WINED3DSTT_CUBE
:
1441 ERR("Unexpected texture type %d\n", sampler_type
);
1445 /* TEX, TXL, TXD and TXP do not support the "H" modifier,
1446 * so don't use shader_arb_get_modifier
1448 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
) mod
= "_SAT";
1451 /* Fragment samplers always have indentity mapping */
1452 if(sampler_idx
>= MAX_FRAGMENT_SAMPLERS
)
1454 sampler_idx
= priv
->cur_vs_args
->vertex
.samplers
[sampler_idx
- MAX_FRAGMENT_SAMPLERS
];
1457 if (flags
& TEX_DERIV
)
1459 if(flags
& TEX_PROJ
) FIXME("Projected texture sampling with custom derivatives\n");
1460 if(flags
& TEX_BIAS
) FIXME("Biased texture sampling with custom derivatives\n");
1461 shader_addline(buffer
, "TXD%s %s, %s, %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
,
1462 dsx
, dsy
,sampler_idx
, tex_type
);
1464 else if(flags
& TEX_LOD
)
1466 if(flags
& TEX_PROJ
) FIXME("Projected texture sampling with explicit lod\n");
1467 if(flags
& TEX_BIAS
) FIXME("Biased texture sampling with explicit lod\n");
1468 shader_addline(buffer
, "TXL%s %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
,
1469 sampler_idx
, tex_type
);
1471 else if (flags
& TEX_BIAS
)
1473 /* Shouldn't be possible, but let's check for it */
1474 if(flags
& TEX_PROJ
) FIXME("Biased and Projected texture sampling\n");
1475 /* TXB takes the 4th component of the source vector automatically, as d3d. Nothing more to do */
1476 shader_addline(buffer
, "TXB%s %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
, sampler_idx
, tex_type
);
1478 else if (flags
& TEX_PROJ
)
1480 shader_addline(buffer
, "TXP%s %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
, sampler_idx
, tex_type
);
1486 const unsigned char idx
= priv
->cur_np2fixup_info
->super
.idx
[sampler_idx
];
1487 shader_addline(buffer
, "MUL TA, np2fixup[%u].%s, %s;\n", idx
>> 1,
1488 (idx
% 2) ? "zwxy" : "xyzw", coord_reg
);
1490 shader_addline(buffer
, "TEX%s %s, TA, texture[%u], %s;\n", mod
, dst_str
, sampler_idx
, tex_type
);
1493 shader_addline(buffer
, "TEX%s %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
, sampler_idx
, tex_type
);
1498 gen_color_correction(buffer
, dst_str
, ins
->dst
[0].write_mask
,
1499 arb_get_helper_value(WINED3D_SHADER_TYPE_PIXEL
, ARB_ONE
),
1500 arb_get_helper_value(WINED3D_SHADER_TYPE_PIXEL
, ARB_TWO
),
1501 priv
->cur_ps_args
->super
.color_fixup
[sampler_idx
]);
1505 static void shader_arb_get_src_param(const struct wined3d_shader_instruction
*ins
,
1506 const struct wined3d_shader_src_param
*src
, unsigned int tmpreg
, char *outregstr
)
1508 /* Generate a line that does the input modifier computation and return the input register to use */
1509 BOOL is_color
= FALSE
, insert_line
;
1512 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1513 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1514 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
1515 const char *two
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_TWO
);
1517 /* Assume a new line will be added */
1520 /* Get register name */
1521 shader_arb_get_register_name(ins
, &src
->reg
, regstr
, &is_color
);
1522 shader_arb_get_swizzle(src
, is_color
, swzstr
);
1524 switch (src
->modifiers
)
1526 case WINED3DSPSM_NONE
:
1527 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
1528 insert_line
= FALSE
;
1530 case WINED3DSPSM_NEG
:
1531 sprintf(outregstr
, "-%s%s", regstr
, swzstr
);
1532 insert_line
= FALSE
;
1534 case WINED3DSPSM_BIAS
:
1535 shader_addline(buffer
, "ADD T%c, %s, -coefdiv.x;\n", 'A' + tmpreg
, regstr
);
1537 case WINED3DSPSM_BIASNEG
:
1538 shader_addline(buffer
, "ADD T%c, -%s, coefdiv.x;\n", 'A' + tmpreg
, regstr
);
1540 case WINED3DSPSM_SIGN
:
1541 shader_addline(buffer
, "MAD T%c, %s, %s, -%s;\n", 'A' + tmpreg
, regstr
, two
, one
);
1543 case WINED3DSPSM_SIGNNEG
:
1544 shader_addline(buffer
, "MAD T%c, %s, -%s, %s;\n", 'A' + tmpreg
, regstr
, two
, one
);
1546 case WINED3DSPSM_COMP
:
1547 shader_addline(buffer
, "SUB T%c, %s, %s;\n", 'A' + tmpreg
, one
, regstr
);
1549 case WINED3DSPSM_X2
:
1550 shader_addline(buffer
, "ADD T%c, %s, %s;\n", 'A' + tmpreg
, regstr
, regstr
);
1552 case WINED3DSPSM_X2NEG
:
1553 shader_addline(buffer
, "ADD T%c, -%s, -%s;\n", 'A' + tmpreg
, regstr
, regstr
);
1555 case WINED3DSPSM_DZ
:
1556 shader_addline(buffer
, "RCP T%c, %s.z;\n", 'A' + tmpreg
, regstr
);
1557 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
1559 case WINED3DSPSM_DW
:
1560 shader_addline(buffer
, "RCP T%c, %s.w;\n", 'A' + tmpreg
, regstr
);
1561 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
1563 case WINED3DSPSM_ABS
:
1564 if(ctx
->target_version
>= NV2
) {
1565 sprintf(outregstr
, "|%s%s|", regstr
, swzstr
);
1566 insert_line
= FALSE
;
1568 shader_addline(buffer
, "ABS T%c, %s;\n", 'A' + tmpreg
, regstr
);
1571 case WINED3DSPSM_ABSNEG
:
1572 if(ctx
->target_version
>= NV2
) {
1573 sprintf(outregstr
, "-|%s%s|", regstr
, swzstr
);
1575 shader_addline(buffer
, "ABS T%c, %s;\n", 'A' + tmpreg
, regstr
);
1576 sprintf(outregstr
, "-T%c%s", 'A' + tmpreg
, swzstr
);
1578 insert_line
= FALSE
;
1581 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
1582 insert_line
= FALSE
;
1585 /* Return modified or original register, with swizzle */
1587 sprintf(outregstr
, "T%c%s", 'A' + tmpreg
, swzstr
);
1590 static void pshader_hw_bem(const struct wined3d_shader_instruction
*ins
)
1592 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1593 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1594 DWORD sampler_code
= dst
->reg
.idx
[0].offset
;
1596 char src_name
[2][50];
1598 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1600 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed
1602 * Keep in mind that src_name[1] can be "TB" and src_name[0] can be "TA" because modifiers like _x2 are valid
1603 * with bem. So delay loading the first parameter until after the perturbation calculation which needs two
1606 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1607 shader_addline(buffer
, "SWZ TA, bumpenvmat%d, x, z, 0, 0;\n", sampler_code
);
1608 shader_addline(buffer
, "DP3 TC.r, TA, %s;\n", src_name
[1]);
1609 shader_addline(buffer
, "SWZ TA, bumpenvmat%d, y, w, 0, 0;\n", sampler_code
);
1610 shader_addline(buffer
, "DP3 TC.g, TA, %s;\n", src_name
[1]);
1612 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1613 shader_addline(buffer
, "ADD %s, %s, TC;\n", dst_name
, src_name
[0]);
1616 static DWORD
negate_modifiers(DWORD mod
, char *extra_char
)
1621 case WINED3DSPSM_NONE
: return WINED3DSPSM_NEG
;
1622 case WINED3DSPSM_NEG
: return WINED3DSPSM_NONE
;
1623 case WINED3DSPSM_BIAS
: return WINED3DSPSM_BIASNEG
;
1624 case WINED3DSPSM_BIASNEG
: return WINED3DSPSM_BIAS
;
1625 case WINED3DSPSM_SIGN
: return WINED3DSPSM_SIGNNEG
;
1626 case WINED3DSPSM_SIGNNEG
: return WINED3DSPSM_SIGN
;
1627 case WINED3DSPSM_COMP
: *extra_char
= '-'; return WINED3DSPSM_COMP
;
1628 case WINED3DSPSM_X2
: return WINED3DSPSM_X2NEG
;
1629 case WINED3DSPSM_X2NEG
: return WINED3DSPSM_X2
;
1630 case WINED3DSPSM_DZ
: *extra_char
= '-'; return WINED3DSPSM_DZ
;
1631 case WINED3DSPSM_DW
: *extra_char
= '-'; return WINED3DSPSM_DW
;
1632 case WINED3DSPSM_ABS
: return WINED3DSPSM_ABSNEG
;
1633 case WINED3DSPSM_ABSNEG
: return WINED3DSPSM_ABS
;
1635 FIXME("Unknown modifier %u\n", mod
);
1639 static void pshader_hw_cnd(const struct wined3d_shader_instruction
*ins
)
1641 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1642 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1644 char src_name
[3][50];
1645 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1646 ins
->ctx
->reg_maps
->shader_version
.minor
);
1648 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1649 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1651 if (shader_version
<= WINED3D_SHADER_VERSION(1, 3) && ins
->coissue
1652 && ins
->dst
->write_mask
!= WINED3DSP_WRITEMASK_3
)
1654 shader_addline(buffer
, "MOV%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
[1]);
1658 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
1661 /* src0 may have a negate srcmod set, so we can't blindly add "-" to the name */
1662 src0_copy
.modifiers
= negate_modifiers(src0_copy
.modifiers
, &extra_neg
);
1664 shader_arb_get_src_param(ins
, &src0_copy
, 0, src_name
[0]);
1665 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1666 shader_addline(buffer
, "ADD TA, %c%s, coefdiv.x;\n", extra_neg
, src_name
[0]);
1667 shader_addline(buffer
, "CMP%s %s, TA, %s, %s;\n", shader_arb_get_modifier(ins
),
1668 dst_name
, src_name
[1], src_name
[2]);
1672 static void pshader_hw_cmp(const struct wined3d_shader_instruction
*ins
)
1674 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1675 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1677 char src_name
[3][50];
1679 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1681 /* Generate input register names (with modifiers) */
1682 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1683 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1684 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1686 shader_addline(buffer
, "CMP%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1687 dst_name
, src_name
[0], src_name
[2], src_name
[1]);
1690 /** Process the WINED3DSIO_DP2ADD instruction in ARB.
1691 * dst = dot2(src0, src1) + src2 */
1692 static void pshader_hw_dp2add(const struct wined3d_shader_instruction
*ins
)
1694 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1695 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1697 char src_name
[3][50];
1698 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1700 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1701 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1702 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1704 if(ctx
->target_version
>= NV3
)
1706 /* GL_NV_fragment_program2 has a 1:1 matching instruction */
1707 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1708 shader_addline(buffer
, "DP2A%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1709 dst_name
, src_name
[0], src_name
[1], src_name
[2]);
1711 else if(ctx
->target_version
>= NV2
)
1713 /* dst.x = src2.?, src0.x, src1.x + src0.y * src1.y
1714 * dst.y = src2.?, src0.x, src1.z + src0.y * src1.w
1715 * dst.z = src2.?, src0.x, src1.x + src0.y * src1.y
1716 * dst.z = src2.?, src0.x, src1.z + src0.y * src1.w
1718 * Make sure that src1.zw = src1.xy, then we get a classic dp2add
1720 * .xyxy and other swizzles that we could get with this are not valid in
1721 * plain ARBfp, but luckily the NV extension grammar lifts this limitation.
1723 struct wined3d_shader_src_param tmp_param
= ins
->src
[1];
1724 DWORD swizzle
= tmp_param
.swizzle
& 0xf; /* Selects .xy */
1725 tmp_param
.swizzle
= swizzle
| (swizzle
<< 4); /* Creates .xyxy */
1727 shader_arb_get_src_param(ins
, &tmp_param
, 1, src_name
[1]);
1729 shader_addline(buffer
, "X2D%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1730 dst_name
, src_name
[2], src_name
[0], src_name
[1]);
1734 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1735 /* 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]
1736 * src_name[0] can be TA, but TA is a private temp for modifiers, so it is save to overwrite
1738 shader_addline(buffer
, "MOV TA, %s;\n", src_name
[0]);
1739 shader_addline(buffer
, "MOV TA.z, 0.0;\n");
1740 shader_addline(buffer
, "DP3 TA, TA, %s;\n", src_name
[1]);
1741 shader_addline(buffer
, "ADD%s %s, TA, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
[2]);
1745 /* Map the opcode 1-to-1 to the GL code */
1746 static void shader_hw_map2gl(const struct wined3d_shader_instruction
*ins
)
1748 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1749 const char *instruction
;
1750 char arguments
[256], dst_str
[50];
1752 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1754 switch (ins
->handler_idx
)
1756 case WINED3DSIH_ABS
: instruction
= "ABS"; break;
1757 case WINED3DSIH_ADD
: instruction
= "ADD"; break;
1758 case WINED3DSIH_CRS
: instruction
= "XPD"; break;
1759 case WINED3DSIH_DP3
: instruction
= "DP3"; break;
1760 case WINED3DSIH_DP4
: instruction
= "DP4"; break;
1761 case WINED3DSIH_DST
: instruction
= "DST"; break;
1762 case WINED3DSIH_FRC
: instruction
= "FRC"; break;
1763 case WINED3DSIH_LIT
: instruction
= "LIT"; break;
1764 case WINED3DSIH_LRP
: instruction
= "LRP"; break;
1765 case WINED3DSIH_MAD
: instruction
= "MAD"; break;
1766 case WINED3DSIH_MAX
: instruction
= "MAX"; break;
1767 case WINED3DSIH_MIN
: instruction
= "MIN"; break;
1768 case WINED3DSIH_MOV
: instruction
= "MOV"; break;
1769 case WINED3DSIH_MUL
: instruction
= "MUL"; break;
1770 case WINED3DSIH_SGE
: instruction
= "SGE"; break;
1771 case WINED3DSIH_SLT
: instruction
= "SLT"; break;
1772 case WINED3DSIH_SUB
: instruction
= "SUB"; break;
1773 case WINED3DSIH_MOVA
:instruction
= "ARR"; break;
1774 case WINED3DSIH_DSX
: instruction
= "DDX"; break;
1775 default: instruction
= "";
1776 FIXME("Unhandled opcode %#x\n", ins
->handler_idx
);
1780 /* Note that shader_arb_add_dst_param() adds spaces. */
1781 arguments
[0] = '\0';
1782 shader_arb_get_dst_param(ins
, dst
, dst_str
);
1783 for (i
= 0; i
< ins
->src_count
; ++i
)
1786 strcat(arguments
, ", ");
1787 shader_arb_get_src_param(ins
, &ins
->src
[i
], i
, operand
);
1788 strcat(arguments
, operand
);
1790 shader_addline(buffer
, "%s%s %s%s;\n", instruction
, shader_arb_get_modifier(ins
), dst_str
, arguments
);
1793 static void shader_hw_nop(const struct wined3d_shader_instruction
*ins
) {}
1795 static DWORD
shader_arb_select_component(DWORD swizzle
, DWORD component
)
1797 return ((swizzle
>> 2 * component
) & 0x3) * 0x55;
1800 static void shader_hw_mov(const struct wined3d_shader_instruction
*ins
)
1802 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
1803 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
1804 BOOL pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
1805 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1806 const char *zero
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ZERO
);
1807 const char *one
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ONE
);
1808 const char *two
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_TWO
);
1810 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1811 char src0_param
[256];
1813 if (ins
->handler_idx
== WINED3DSIH_MOVA
)
1815 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1817 const char *offset
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_VS_REL_OFFSET
);
1819 if(ctx
->target_version
>= NV2
) {
1820 shader_hw_map2gl(ins
);
1823 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_param
);
1824 shader_arb_get_write_mask(ins
, &ins
->dst
[0], write_mask
);
1826 /* This implements the mova formula used in GLSL. The first two instructions
1827 * prepare the sign() part. Note that it is fine to have my_sign(0.0) = 1.0
1831 * A0.x = arl(floor(abs(0.0) + 0.5) * 1.0) = floor(0.5) = 0.0 since arl does a floor
1833 * The ARL is performed when A0 is used - the requested component is read from A0_SHADOW into
1834 * A0.x. We can use the overwritten component of A0_shadow as temporary storage for the sign.
1836 shader_addline(buffer
, "SGE A0_SHADOW%s, %s, %s;\n", write_mask
, src0_param
, zero
);
1837 shader_addline(buffer
, "MAD A0_SHADOW%s, A0_SHADOW, %s, -%s;\n", write_mask
, two
, one
);
1839 shader_addline(buffer
, "ABS TA%s, %s;\n", write_mask
, src0_param
);
1840 shader_addline(buffer
, "ADD TA%s, TA, rel_addr_const.x;\n", write_mask
);
1841 shader_addline(buffer
, "FLR TA%s, TA;\n", write_mask
);
1842 if (shader_data
->rel_offset
)
1844 shader_addline(buffer
, "ADD TA%s, TA, %s;\n", write_mask
, offset
);
1846 shader_addline(buffer
, "MUL A0_SHADOW%s, TA, A0_SHADOW;\n", write_mask
);
1848 ((struct shader_arb_ctx_priv
*)ins
->ctx
->backend_data
)->addr_reg
[0] = '\0';
1850 else if (reg_maps
->shader_version
.major
== 1
1851 && !shader_is_pshader_version(reg_maps
->shader_version
.type
)
1852 && ins
->dst
[0].reg
.type
== WINED3DSPR_ADDR
)
1854 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1855 src0_param
[0] = '\0';
1857 if (shader_data
->rel_offset
&& ctx
->target_version
== ARB
)
1859 const char *offset
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_VS_REL_OFFSET
);
1860 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_param
);
1861 shader_addline(buffer
, "ADD TA.x, %s, %s;\n", src0_param
, offset
);
1862 shader_addline(buffer
, "ARL A0.x, TA.x;\n");
1866 /* Apple's ARB_vertex_program implementation does not accept an ARL source argument
1867 * with more than one component. Thus replicate the first source argument over all
1868 * 4 components. For example, .xyzw -> .x (or better: .xxxx), .zwxy -> .z, etc) */
1869 struct wined3d_shader_src_param tmp_src
= ins
->src
[0];
1870 tmp_src
.swizzle
= shader_arb_select_component(tmp_src
.swizzle
, 0);
1871 shader_arb_get_src_param(ins
, &tmp_src
, 0, src0_param
);
1872 shader_addline(buffer
, "ARL A0.x, %s;\n", src0_param
);
1875 else if (ins
->dst
[0].reg
.type
== WINED3DSPR_COLOROUT
&& !ins
->dst
[0].reg
.idx
[0].offset
&& pshader
)
1877 if (ctx
->ps_post_process
&& shader
->u
.ps
.color0_mov
)
1879 shader_addline(buffer
, "#mov handled in srgb write or fog code\n");
1882 shader_hw_map2gl(ins
);
1886 shader_hw_map2gl(ins
);
1890 static void pshader_hw_texkill(const struct wined3d_shader_instruction
*ins
)
1892 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1893 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1896 /* No swizzles are allowed in d3d's texkill. PS 1.x ignores the 4th component as documented,
1897 * but >= 2.0 honors it (undocumented, but tested by the d3d9 testsuite)
1899 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1901 if (ins
->ctx
->reg_maps
->shader_version
.major
>= 2)
1903 const char *kilsrc
= "TA";
1906 shader_arb_get_register_name(ins
, &dst
->reg
, reg_dest
, &is_color
);
1907 if(dst
->write_mask
== WINED3DSP_WRITEMASK_ALL
)
1913 /* Sigh. KIL doesn't support swizzles/writemasks. KIL passes a writemask, but ".xy" for example
1914 * is not valid as a swizzle in ARB (needs ".xyyy"). Use SWZ to load the register properly, and set
1915 * masked out components to 0(won't kill)
1917 char x
= '0', y
= '0', z
= '0', w
= '0';
1918 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
) x
= 'x';
1919 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
) y
= 'y';
1920 if(dst
->write_mask
& WINED3DSP_WRITEMASK_2
) z
= 'z';
1921 if(dst
->write_mask
& WINED3DSP_WRITEMASK_3
) w
= 'w';
1922 shader_addline(buffer
, "SWZ TA, %s, %c, %c, %c, %c;\n", reg_dest
, x
, y
, z
, w
);
1924 shader_addline(buffer
, "KIL %s;\n", kilsrc
);
1928 /* ARB fp doesn't like swizzles on the parameter of the KIL instruction. To mask the 4th component,
1929 * copy the register into our general purpose TMP variable, overwrite .w and pass TMP to KIL
1931 * ps_1_3 shaders use the texcoord incarnation of the Tx register. ps_1_4 shaders can use the same,
1932 * or pass in any temporary register(in shader phase 2)
1934 if (ins
->ctx
->reg_maps
->shader_version
.minor
<= 3)
1935 sprintf(reg_dest
, "fragment.texcoord[%u]", dst
->reg
.idx
[0].offset
);
1937 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1938 shader_addline(buffer
, "SWZ TA, %s, x, y, z, 1;\n", reg_dest
);
1939 shader_addline(buffer
, "KIL TA;\n");
1943 static void pshader_hw_tex(const struct wined3d_shader_instruction
*ins
)
1945 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1946 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1947 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1948 ins
->ctx
->reg_maps
->shader_version
.minor
);
1949 struct wined3d_shader_src_param src
;
1953 DWORD reg_sampler_code
;
1955 BOOL swizzle_coord
= FALSE
;
1957 /* All versions have a destination register */
1958 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1960 /* 1.0-1.4: Use destination register number as texture code.
1961 2.0+: Use provided sampler number as texure code. */
1962 if (shader_version
< WINED3D_SHADER_VERSION(2,0))
1963 reg_sampler_code
= dst
->reg
.idx
[0].offset
;
1965 reg_sampler_code
= ins
->src
[1].reg
.idx
[0].offset
;
1967 /* 1.0-1.3: Use the texcoord varying.
1968 1.4+: Use provided coordinate source register. */
1969 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
1970 sprintf(reg_coord
, "fragment.texcoord[%u]", reg_sampler_code
);
1972 /* TEX is the only instruction that can handle DW and DZ natively */
1974 if(src
.modifiers
== WINED3DSPSM_DW
) src
.modifiers
= WINED3DSPSM_NONE
;
1975 if(src
.modifiers
== WINED3DSPSM_DZ
) src
.modifiers
= WINED3DSPSM_NONE
;
1976 shader_arb_get_src_param(ins
, &src
, 0, reg_coord
);
1980 * 1.1, 1.2, 1.3: Use WINED3D_TSS_TEXTURETRANSFORMFLAGS
1981 * 1.4: Use WINED3DSPSM_DZ or WINED3DSPSM_DW on src[0]
1982 * 2.0+: Use WINED3DSI_TEXLD_PROJECT on the opcode
1984 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
1987 if (reg_sampler_code
< MAX_TEXTURES
)
1988 flags
= priv
->cur_ps_args
->super
.tex_transform
>> reg_sampler_code
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
;
1989 if (flags
& WINED3D_PSARGS_PROJECTED
)
1991 myflags
|= TEX_PROJ
;
1992 if ((flags
& ~WINED3D_PSARGS_PROJECTED
) == WINED3D_TTFF_COUNT3
)
1993 swizzle_coord
= TRUE
;
1996 else if (shader_version
< WINED3D_SHADER_VERSION(2,0))
1998 enum wined3d_shader_src_modifier src_mod
= ins
->src
[0].modifiers
;
1999 if (src_mod
== WINED3DSPSM_DZ
)
2001 swizzle_coord
= TRUE
;
2002 myflags
|= TEX_PROJ
;
2003 } else if(src_mod
== WINED3DSPSM_DW
) {
2004 myflags
|= TEX_PROJ
;
2007 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) myflags
|= TEX_PROJ
;
2008 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) myflags
|= TEX_BIAS
;
2013 /* TXP cannot handle DZ natively, so move the z coordinate to .w.
2014 * reg_coord is a read-only varying register, so we need a temp reg */
2015 shader_addline(ins
->ctx
->buffer
, "SWZ TA, %s, x, y, z, z;\n", reg_coord
);
2016 strcpy(reg_coord
, "TA");
2019 shader_hw_sample(ins
, reg_sampler_code
, reg_dest
, reg_coord
, myflags
, NULL
, NULL
);
2022 static void pshader_hw_texcoord(const struct wined3d_shader_instruction
*ins
)
2024 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2025 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2026 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
2027 ins
->ctx
->reg_maps
->shader_version
.minor
);
2030 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
2032 DWORD reg
= dst
->reg
.idx
[0].offset
;
2034 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2035 shader_addline(buffer
, "MOV_SAT %s, fragment.texcoord[%u];\n", dst_str
, reg
);
2039 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_src
);
2040 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2041 shader_addline(buffer
, "MOV %s, %s;\n", dst_str
, reg_src
);
2045 static void pshader_hw_texreg2ar(const struct wined3d_shader_instruction
*ins
)
2047 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2050 DWORD reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2054 /* Note that texreg2ar treats Tx as a temporary register, not as a varying */
2055 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2056 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2057 /* Move .x first in case src_str is "TA" */
2058 shader_addline(buffer
, "MOV TA.y, %s.x;\n", src_str
);
2059 shader_addline(buffer
, "MOV TA.x, %s.w;\n", src_str
);
2060 if (reg1
< MAX_TEXTURES
)
2062 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2063 flags
= priv
->cur_ps_args
->super
.tex_transform
>> reg1
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
;
2065 shader_hw_sample(ins
, reg1
, dst_str
, "TA", flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2068 static void pshader_hw_texreg2gb(const struct wined3d_shader_instruction
*ins
)
2070 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2072 DWORD reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2076 /* Note that texreg2gb treats Tx as a temporary register, not as a varying */
2077 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2078 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2079 shader_addline(buffer
, "MOV TA.x, %s.y;\n", src_str
);
2080 shader_addline(buffer
, "MOV TA.y, %s.z;\n", src_str
);
2081 shader_hw_sample(ins
, reg1
, dst_str
, "TA", 0, NULL
, NULL
);
2084 static void pshader_hw_texreg2rgb(const struct wined3d_shader_instruction
*ins
)
2086 DWORD reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2090 /* Note that texreg2rg treats Tx as a temporary register, not as a varying */
2091 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2092 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2093 shader_hw_sample(ins
, reg1
, dst_str
, src_str
, 0, NULL
, NULL
);
2096 static void pshader_hw_texbem(const struct wined3d_shader_instruction
*ins
)
2098 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2099 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2100 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2101 char reg_coord
[40], dst_reg
[50], src_reg
[50];
2102 DWORD reg_dest_code
;
2104 /* All versions have a destination register. The Tx where the texture coordinates come
2105 * from is the varying incarnation of the texture register
2107 reg_dest_code
= dst
->reg
.idx
[0].offset
;
2108 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_reg
);
2109 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_reg
);
2110 sprintf(reg_coord
, "fragment.texcoord[%u]", reg_dest_code
);
2112 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed
2113 * The Tx in which the perturbation map is stored is the tempreg incarnation of the texture register
2115 * GL_NV_fragment_program_option could handle this in one instruction via X2D:
2116 * X2D TA.xy, fragment.texcoord, T%u, bumpenvmat%u.xzyw
2118 * However, the NV extensions are never enabled for <= 2.0 shaders because of the performance penalty that
2119 * comes with it, and texbem is an 1.x only instruction. No 1.x instruction forces us to enable the NV
2122 shader_addline(buffer
, "SWZ TB, bumpenvmat%d, x, z, 0, 0;\n", reg_dest_code
);
2123 shader_addline(buffer
, "DP3 TA.x, TB, %s;\n", src_reg
);
2124 shader_addline(buffer
, "SWZ TB, bumpenvmat%d, y, w, 0, 0;\n", reg_dest_code
);
2125 shader_addline(buffer
, "DP3 TA.y, TB, %s;\n", src_reg
);
2127 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2128 * so we can't let the GL handle this.
2130 if ((priv
->cur_ps_args
->super
.tex_transform
>> reg_dest_code
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
)
2131 & WINED3D_PSARGS_PROJECTED
)
2133 shader_addline(buffer
, "RCP TB.w, %s.w;\n", reg_coord
);
2134 shader_addline(buffer
, "MUL TB.xy, %s, TB.w;\n", reg_coord
);
2135 shader_addline(buffer
, "ADD TA.xy, TA, TB;\n");
2137 shader_addline(buffer
, "ADD TA.xy, TA, %s;\n", reg_coord
);
2140 shader_hw_sample(ins
, reg_dest_code
, dst_reg
, "TA", 0, NULL
, NULL
);
2142 if (ins
->handler_idx
== WINED3DSIH_TEXBEML
)
2144 /* No src swizzles are allowed, so this is ok */
2145 shader_addline(buffer
, "MAD TA, %s.z, luminance%d.x, luminance%d.y;\n",
2146 src_reg
, reg_dest_code
, reg_dest_code
);
2147 shader_addline(buffer
, "MUL %s, %s, TA;\n", dst_reg
, dst_reg
);
2151 static void pshader_hw_texm3x2pad(const struct wined3d_shader_instruction
*ins
)
2153 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2154 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2155 char src0_name
[50], dst_name
[50];
2157 struct wined3d_shader_register tmp_reg
= ins
->dst
[0].reg
;
2159 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2160 /* The next instruction will be a texm3x2tex or texm3x2depth that writes to the uninitialized
2161 * T<reg+1> register. Use this register to store the calculated vector
2163 tmp_reg
.idx
[0].offset
= reg
+ 1;
2164 shader_arb_get_register_name(ins
, &tmp_reg
, dst_name
, &is_color
);
2165 shader_addline(buffer
, "DP3 %s.x, fragment.texcoord[%u], %s;\n", dst_name
, reg
, src0_name
);
2168 static void pshader_hw_texm3x2tex(const struct wined3d_shader_instruction
*ins
)
2170 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2172 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2173 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2179 /* We know that we're writing to the uninitialized T<reg> register, so use it for temporary storage */
2180 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2182 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2183 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2184 shader_addline(buffer
, "DP3 %s.y, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2185 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2186 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2189 static void pshader_hw_texm3x3pad(const struct wined3d_shader_instruction
*ins
)
2191 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2192 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2193 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2194 char src0_name
[50], dst_name
[50];
2195 struct wined3d_shader_register tmp_reg
= ins
->dst
[0].reg
;
2198 /* There are always 2 texm3x3pad instructions followed by one texm3x3[tex,vspec, ...] instruction, with
2199 * incrementing ins->dst[0].register_idx numbers. So the pad instruction already knows the final destination
2200 * register, and this register is uninitialized(otherwise the assembler complains that it is 'redeclared')
2202 tmp_reg
.idx
[0].offset
= reg
+ 2 - tex_mx
->current_row
;
2203 shader_arb_get_register_name(ins
, &tmp_reg
, dst_name
, &is_color
);
2205 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2206 shader_addline(buffer
, "DP3 %s.%c, fragment.texcoord[%u], %s;\n",
2207 dst_name
, 'x' + tex_mx
->current_row
, reg
, src0_name
);
2208 tex_mx
->texcoord_w
[tex_mx
->current_row
++] = reg
;
2211 static void pshader_hw_texm3x3tex(const struct wined3d_shader_instruction
*ins
)
2213 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2214 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2216 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2217 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2219 char src0_name
[50], dst_name
[50];
2222 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2223 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2224 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_name
, reg
, src0_name
);
2226 /* Sample the texture using the calculated coordinates */
2227 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2228 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2229 shader_hw_sample(ins
, reg
, dst_str
, dst_name
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2230 tex_mx
->current_row
= 0;
2233 static void pshader_hw_texm3x3vspec(const struct wined3d_shader_instruction
*ins
)
2235 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2236 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2238 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2239 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2245 /* Get the dst reg without writemask strings. We know this register is uninitialized, so we can use all
2246 * components for temporary data storage
2248 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2249 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2250 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2252 /* Construct the eye-ray vector from w coordinates */
2253 shader_addline(buffer
, "MOV TB.x, fragment.texcoord[%u].w;\n", tex_mx
->texcoord_w
[0]);
2254 shader_addline(buffer
, "MOV TB.y, fragment.texcoord[%u].w;\n", tex_mx
->texcoord_w
[1]);
2255 shader_addline(buffer
, "MOV TB.z, fragment.texcoord[%u].w;\n", reg
);
2257 /* Calculate reflection vector
2259 shader_addline(buffer
, "DP3 %s.w, %s, TB;\n", dst_reg
, dst_reg
);
2260 /* The .w is ignored when sampling, so I can use TB.w to calculate dot(N, N) */
2261 shader_addline(buffer
, "DP3 TB.w, %s, %s;\n", dst_reg
, dst_reg
);
2262 shader_addline(buffer
, "RCP TB.w, TB.w;\n");
2263 shader_addline(buffer
, "MUL %s.w, %s.w, TB.w;\n", dst_reg
, dst_reg
);
2264 shader_addline(buffer
, "MUL %s, %s.w, %s;\n", dst_reg
, dst_reg
, dst_reg
);
2265 shader_addline(buffer
, "MAD %s, coefmul.x, %s, -TB;\n", dst_reg
, dst_reg
);
2267 /* Sample the texture using the calculated coordinates */
2268 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2269 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2270 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2271 tex_mx
->current_row
= 0;
2274 static void pshader_hw_texm3x3spec(const struct wined3d_shader_instruction
*ins
)
2276 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2277 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2279 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2280 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2287 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2288 shader_arb_get_src_param(ins
, &ins
->src
[0], 1, src1_name
);
2289 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2290 /* Note: dst_reg.xy is input here, generated by two texm3x3pad instructions */
2291 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2293 /* Calculate reflection vector.
2296 * dst_reg.xyz = 2 * --------- * N - E
2299 * Which normalizes the normal vector
2301 shader_addline(buffer
, "DP3 %s.w, %s, %s;\n", dst_reg
, dst_reg
, src1_name
);
2302 shader_addline(buffer
, "DP3 TC.w, %s, %s;\n", dst_reg
, dst_reg
);
2303 shader_addline(buffer
, "RCP TC.w, TC.w;\n");
2304 shader_addline(buffer
, "MUL %s.w, %s.w, TC.w;\n", dst_reg
, dst_reg
);
2305 shader_addline(buffer
, "MUL %s, %s.w, %s;\n", dst_reg
, dst_reg
, dst_reg
);
2306 shader_addline(buffer
, "MAD %s, coefmul.x, %s, -%s;\n", dst_reg
, dst_reg
, src1_name
);
2308 /* Sample the texture using the calculated coordinates */
2309 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2310 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2311 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2312 tex_mx
->current_row
= 0;
2315 static void pshader_hw_texdepth(const struct wined3d_shader_instruction
*ins
)
2317 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2318 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2320 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2321 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2323 /* texdepth has an implicit destination, the fragment depth value. It's only parameter,
2324 * which is essentially an input, is the destination register because it is the first
2325 * parameter. According to the msdn, this must be register r5, but let's keep it more flexible
2326 * here(writemasks/swizzles are not valid on texdepth)
2328 shader_arb_get_dst_param(ins
, dst
, dst_name
);
2330 /* According to the msdn, the source register(must be r5) is unusable after
2331 * the texdepth instruction, so we're free to modify it
2333 shader_addline(buffer
, "MIN %s.y, %s.y, %s;\n", dst_name
, dst_name
, one
);
2335 /* How to deal with the special case dst_name.g == 0? if r != 0, then
2336 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
2337 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
2339 shader_addline(buffer
, "RCP %s.y, %s.y;\n", dst_name
, dst_name
);
2340 shader_addline(buffer
, "MUL TA.x, %s.x, %s.y;\n", dst_name
, dst_name
);
2341 shader_addline(buffer
, "MIN TA.x, TA.x, %s;\n", one
);
2342 shader_addline(buffer
, "MAX result.depth, TA.x, %s;\n", zero
);
2345 /** Process the WINED3DSIO_TEXDP3TEX instruction in ARB:
2346 * Take a 3-component dot product of the TexCoord[dstreg] and src,
2347 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
2348 static void pshader_hw_texdp3tex(const struct wined3d_shader_instruction
*ins
)
2350 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2351 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
[0].offset
;
2355 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2356 shader_addline(buffer
, "MOV TB, 0.0;\n");
2357 shader_addline(buffer
, "DP3 TB.x, fragment.texcoord[%u], %s;\n", sampler_idx
, src0
);
2359 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2360 shader_hw_sample(ins
, sampler_idx
, dst_str
, "TB", 0 /* Only one coord, can't be projected */, NULL
, NULL
);
2363 /** Process the WINED3DSIO_TEXDP3 instruction in ARB:
2364 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
2365 static void pshader_hw_texdp3(const struct wined3d_shader_instruction
*ins
)
2367 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2370 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2372 /* Handle output register */
2373 shader_arb_get_dst_param(ins
, dst
, dst_str
);
2374 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2375 shader_addline(buffer
, "DP3 %s, fragment.texcoord[%u], %s;\n", dst_str
, dst
->reg
.idx
[0].offset
, src0
);
2378 /** Process the WINED3DSIO_TEXM3X3 instruction in ARB
2379 * Perform the 3rd row of a 3x3 matrix multiply */
2380 static void pshader_hw_texm3x3(const struct wined3d_shader_instruction
*ins
)
2382 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2383 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2384 char dst_str
[50], dst_name
[50];
2388 shader_arb_get_dst_param(ins
, dst
, dst_str
);
2389 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2390 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2391 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_name
, dst
->reg
.idx
[0].offset
, src0
);
2392 shader_addline(buffer
, "MOV %s, %s;\n", dst_str
, dst_name
);
2395 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in ARB:
2396 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
2397 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
2398 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
2400 static void pshader_hw_texm3x2depth(const struct wined3d_shader_instruction
*ins
)
2402 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2403 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2404 char src0
[50], dst_name
[50];
2406 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2407 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2409 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2410 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2411 shader_addline(buffer
, "DP3 %s.y, fragment.texcoord[%u], %s;\n", dst_name
, dst
->reg
.idx
[0].offset
, src0
);
2413 /* How to deal with the special case dst_name.g == 0? if r != 0, then
2414 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
2415 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
2417 shader_addline(buffer
, "RCP %s.y, %s.y;\n", dst_name
, dst_name
);
2418 shader_addline(buffer
, "MUL %s.x, %s.x, %s.y;\n", dst_name
, dst_name
, dst_name
);
2419 shader_addline(buffer
, "MIN %s.x, %s.x, %s;\n", dst_name
, dst_name
, one
);
2420 shader_addline(buffer
, "MAX result.depth, %s.x, %s;\n", dst_name
, zero
);
2423 /** Handles transforming all WINED3DSIO_M?x? opcodes for
2424 Vertex/Pixel shaders to ARB_vertex_program codes */
2425 static void shader_hw_mnxn(const struct wined3d_shader_instruction
*ins
)
2428 int nComponents
= 0;
2429 struct wined3d_shader_dst_param tmp_dst
= {{0}};
2430 struct wined3d_shader_src_param tmp_src
[2] = {{{0}}};
2431 struct wined3d_shader_instruction tmp_ins
;
2433 memset(&tmp_ins
, 0, sizeof(tmp_ins
));
2435 /* Set constants for the temporary argument */
2436 tmp_ins
.ctx
= ins
->ctx
;
2437 tmp_ins
.dst_count
= 1;
2438 tmp_ins
.dst
= &tmp_dst
;
2439 tmp_ins
.src_count
= 2;
2440 tmp_ins
.src
= tmp_src
;
2442 switch(ins
->handler_idx
)
2444 case WINED3DSIH_M4x4
:
2446 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2448 case WINED3DSIH_M4x3
:
2450 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2452 case WINED3DSIH_M3x4
:
2454 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2456 case WINED3DSIH_M3x3
:
2458 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2460 case WINED3DSIH_M3x2
:
2462 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2465 FIXME("Unhandled opcode %#x\n", ins
->handler_idx
);
2469 tmp_dst
= ins
->dst
[0];
2470 tmp_src
[0] = ins
->src
[0];
2471 tmp_src
[1] = ins
->src
[1];
2472 for (i
= 0; i
< nComponents
; ++i
)
2474 tmp_dst
.write_mask
= WINED3DSP_WRITEMASK_0
<< i
;
2475 shader_hw_map2gl(&tmp_ins
);
2476 ++tmp_src
[1].reg
.idx
[0].offset
;
2480 static DWORD
abs_modifier(DWORD mod
, BOOL
*need_abs
)
2486 case WINED3DSPSM_NONE
: return WINED3DSPSM_ABS
;
2487 case WINED3DSPSM_NEG
: return WINED3DSPSM_ABS
;
2488 case WINED3DSPSM_BIAS
: *need_abs
= TRUE
; return WINED3DSPSM_BIAS
;
2489 case WINED3DSPSM_BIASNEG
: *need_abs
= TRUE
; return WINED3DSPSM_BIASNEG
;
2490 case WINED3DSPSM_SIGN
: *need_abs
= TRUE
; return WINED3DSPSM_SIGN
;
2491 case WINED3DSPSM_SIGNNEG
: *need_abs
= TRUE
; return WINED3DSPSM_SIGNNEG
;
2492 case WINED3DSPSM_COMP
: *need_abs
= TRUE
; return WINED3DSPSM_COMP
;
2493 case WINED3DSPSM_X2
: *need_abs
= TRUE
; return WINED3DSPSM_X2
;
2494 case WINED3DSPSM_X2NEG
: *need_abs
= TRUE
; return WINED3DSPSM_X2NEG
;
2495 case WINED3DSPSM_DZ
: *need_abs
= TRUE
; return WINED3DSPSM_DZ
;
2496 case WINED3DSPSM_DW
: *need_abs
= TRUE
; return WINED3DSPSM_DW
;
2497 case WINED3DSPSM_ABS
: return WINED3DSPSM_ABS
;
2498 case WINED3DSPSM_ABSNEG
: return WINED3DSPSM_ABS
;
2500 FIXME("Unknown modifier %u\n", mod
);
2504 static void shader_hw_scalar_op(const struct wined3d_shader_instruction
*ins
)
2506 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2507 const char *instruction
;
2508 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
2509 BOOL need_abs
= FALSE
;
2514 switch(ins
->handler_idx
)
2516 case WINED3DSIH_RSQ
: instruction
= "RSQ"; break;
2517 case WINED3DSIH_RCP
: instruction
= "RCP"; break;
2518 case WINED3DSIH_EXPP
:
2519 if (ins
->ctx
->reg_maps
->shader_version
.major
< 2)
2521 instruction
= "EXP";
2525 case WINED3DSIH_EXP
:
2526 instruction
= "EX2";
2528 case WINED3DSIH_LOG
:
2529 case WINED3DSIH_LOGP
:
2530 /* The precision requirements suggest that LOGP matches ARBvp's LOG
2531 * instruction, but notice that the output of those instructions is
2533 src0_copy
.modifiers
= abs_modifier(src0_copy
.modifiers
, &need_abs
);
2534 instruction
= "LG2";
2536 default: instruction
= "";
2537 FIXME("Unhandled opcode %#x\n", ins
->handler_idx
);
2541 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
2543 src0_copy
.swizzle
= shader_arb_select_component(src0_copy
.swizzle
, 3);
2545 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
); /* Destination */
2546 shader_arb_get_src_param(ins
, &src0_copy
, 0, src
);
2550 shader_addline(buffer
, "ABS TA.w, %s;\n", src
);
2551 shader_addline(buffer
, "%s%s %s, TA.w;\n", instruction
, shader_arb_get_modifier(ins
), dst
);
2555 shader_addline(buffer
, "%s%s %s, %s;\n", instruction
, shader_arb_get_modifier(ins
), dst
, src
);
2560 static void shader_hw_nrm(const struct wined3d_shader_instruction
*ins
)
2562 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2565 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2566 BOOL pshader
= shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2567 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2569 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2570 shader_arb_get_src_param(ins
, &ins
->src
[0], 1 /* Use TB */, src_name
);
2572 /* In D3D, NRM of a vector with length zero returns zero. Catch this situation, as
2573 * otherwise NRM or RSQ would return NaN */
2574 if(pshader
&& priv
->target_version
>= NV3
)
2576 /* GL_NV_fragment_program2's NRM needs protection against length zero vectors too
2578 * TODO: Find out if DP3+NRM+MOV is really faster than DP3+RSQ+MUL
2580 shader_addline(buffer
, "DP3C TA, %s, %s;\n", src_name
, src_name
);
2581 shader_addline(buffer
, "NRM%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
);
2582 shader_addline(buffer
, "MOV %s (EQ), %s;\n", dst_name
, zero
);
2584 else if(priv
->target_version
>= NV2
)
2586 shader_addline(buffer
, "DP3C TA.x, %s, %s;\n", src_name
, src_name
);
2587 shader_addline(buffer
, "RSQ TA.x (NE), TA.x;\n");
2588 shader_addline(buffer
, "MUL%s %s, %s, TA.x;\n", shader_arb_get_modifier(ins
), dst_name
,
2593 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2595 shader_addline(buffer
, "DP3 TA.x, %s, %s;\n", src_name
, src_name
);
2596 /* Pass any non-zero value to RSQ if the input vector has a length of zero. The
2597 * RSQ result doesn't matter, as long as multiplying it by 0 returns 0.
2599 shader_addline(buffer
, "SGE TA.y, -TA.x, %s;\n", zero
);
2600 shader_addline(buffer
, "MAD TA.x, %s, TA.y, TA.x;\n", one
);
2602 shader_addline(buffer
, "RSQ TA.x, TA.x;\n");
2603 /* dst.w = src[0].w * 1 / (src.x^2 + src.y^2 + src.z^2)^(1/2) according to msdn*/
2604 shader_addline(buffer
, "MUL%s %s, %s, TA.x;\n", shader_arb_get_modifier(ins
), dst_name
,
2609 static void shader_hw_lrp(const struct wined3d_shader_instruction
*ins
)
2611 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2613 char src_name
[3][50];
2615 /* ARB_fragment_program has a convenient LRP instruction */
2616 if(shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)) {
2617 shader_hw_map2gl(ins
);
2621 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2622 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
2623 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
2624 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
2626 shader_addline(buffer
, "SUB TA, %s, %s;\n", src_name
[1], src_name
[2]);
2627 shader_addline(buffer
, "MAD%s %s, %s, TA, %s;\n", shader_arb_get_modifier(ins
),
2628 dst_name
, src_name
[0], src_name
[2]);
2631 static void shader_hw_sincos(const struct wined3d_shader_instruction
*ins
)
2633 /* This instruction exists in ARB, but the d3d instruction takes two extra parameters which
2634 * must contain fixed constants. So we need a separate function to filter those constants and
2637 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2638 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2639 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2641 char src_name0
[50], src_name1
[50], src_name2
[50];
2644 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
2645 if(shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)) {
2646 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2647 /* No modifiers are supported on SCS */
2648 shader_addline(buffer
, "SCS %s, %s;\n", dst_name
, src_name0
);
2650 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
)
2652 shader_arb_get_register_name(ins
, &dst
->reg
, src_name0
, &is_color
);
2653 shader_addline(buffer
, "MOV_SAT %s, %s;\n", dst_name
, src_name0
);
2655 } else if(priv
->target_version
>= NV2
) {
2656 shader_arb_get_register_name(ins
, &dst
->reg
, dst_name
, &is_color
);
2658 /* Sincos writemask must be .x, .y or .xy */
2659 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
)
2660 shader_addline(buffer
, "COS%s %s.x, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name0
);
2661 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
)
2662 shader_addline(buffer
, "SIN%s %s.y, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name0
);
2664 /* Approximate sine and cosine with a taylor series, as per math textbook. The application passes 8
2665 * helper constants(D3DSINCOSCONST1 and D3DSINCOSCONST2) in src1 and src2.
2667 * sin(x) = x - x^3/3! + x^5/5! - x^7/7! + ...
2668 * cos(x) = 1 - x^2/2! + x^4/4! - x^6/6! + ...
2670 * The constants we get are:
2672 * +1 +1, -1 -1 +1 +1 -1 -1
2673 * ---- , ---- , ---- , ----- , ----- , ----- , ------
2674 * 1!*2 2!*4 3!*8 4!*16 5!*32 6!*64 7!*128
2676 * If used with x^2, x^3, x^4 etc they calculate sin(x/2) and cos(x/2):
2680 * (x/2)^4 = x^4 / 16
2681 * (x/2)^5 = x^5 / 32
2684 * To get the final result:
2685 * sin(x) = 2 * sin(x/2) * cos(x/2)
2686 * cos(x) = cos(x/2)^2 - sin(x/2)^2
2687 * (from sin(x+y) and cos(x+y) rules)
2689 * As per MSDN, dst.z is undefined after the operation, and so is
2690 * dst.x and dst.y if they're masked out by the writemask. Ie
2691 * sincos dst.y, src1, c0, c1
2692 * returns the sine in dst.y. dst.x and dst.z are undefined, dst.w is not touched. The assembler
2693 * vsa.exe also stops with an error if the dest register is the same register as the source
2694 * register. This means we can use dest.xyz as temporary storage. The assembler vsa.exe output also
2695 * indicates that sincos consumes 8 instruction slots in vs_2_0(and, strangely, in vs_3_0).
2697 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
2698 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name2
);
2699 shader_arb_get_register_name(ins
, &dst
->reg
, dst_name
, &is_color
);
2701 shader_addline(buffer
, "MUL %s.x, %s, %s;\n", dst_name
, src_name0
, src_name0
); /* x ^ 2 */
2702 shader_addline(buffer
, "MUL TA.y, %s.x, %s;\n", dst_name
, src_name0
); /* x ^ 3 */
2703 shader_addline(buffer
, "MUL %s.y, TA.y, %s;\n", dst_name
, src_name0
); /* x ^ 4 */
2704 shader_addline(buffer
, "MUL TA.z, %s.y, %s;\n", dst_name
, src_name0
); /* x ^ 5 */
2705 shader_addline(buffer
, "MUL %s.z, TA.z, %s;\n", dst_name
, src_name0
); /* x ^ 6 */
2706 shader_addline(buffer
, "MUL TA.w, %s.z, %s;\n", dst_name
, src_name0
); /* x ^ 7 */
2710 * Unfortunately we don't get the constants in a DP4-capable form. Is there a way to
2711 * properly merge that with MULs in the code above?
2712 * The swizzles .yz and xw however fit into the .yzxw swizzle added to ps_2_0. Maybe
2713 * we can merge the sine and cosine MAD rows to calculate them together.
2715 shader_addline(buffer
, "MUL TA.x, %s, %s.w;\n", src_name0
, src_name2
); /* x^1, +1/(1!*2) */
2716 shader_addline(buffer
, "MAD TA.x, TA.y, %s.x, TA.x;\n", src_name2
); /* -1/(3!*8) */
2717 shader_addline(buffer
, "MAD TA.x, TA.z, %s.w, TA.x;\n", src_name1
); /* +1/(5!*32) */
2718 shader_addline(buffer
, "MAD TA.x, TA.w, %s.x, TA.x;\n", src_name1
); /* -1/(7!*128) */
2721 shader_addline(buffer
, "MAD TA.y, %s.x, %s.y, %s.z;\n", dst_name
, src_name2
, src_name2
); /* -1/(2!*4), +1.0 */
2722 shader_addline(buffer
, "MAD TA.y, %s.y, %s.z, TA.y;\n", dst_name
, src_name1
); /* +1/(4!*16) */
2723 shader_addline(buffer
, "MAD TA.y, %s.z, %s.y, TA.y;\n", dst_name
, src_name1
); /* -1/(6!*64) */
2725 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
) {
2727 shader_addline(buffer
, "MUL TA.z, TA.y, TA.y;\n");
2728 shader_addline(buffer
, "MAD %s.x, -TA.x, TA.x, TA.z;\n", dst_name
);
2730 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
) {
2732 shader_addline(buffer
, "MUL %s.y, TA.x, TA.y;\n", dst_name
);
2733 shader_addline(buffer
, "ADD %s.y, %s.y, %s.y;\n", dst_name
, dst_name
, dst_name
);
2738 static void shader_hw_sgn(const struct wined3d_shader_instruction
*ins
)
2740 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2743 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
2745 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2746 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
);
2748 /* SGN is only valid in vertex shaders */
2749 if(ctx
->target_version
>= NV2
) {
2750 shader_addline(buffer
, "SSG%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
);
2754 /* If SRC > 0.0, -SRC < SRC = TRUE, otherwise false.
2755 * if SRC < 0.0, SRC < -SRC = TRUE. If neither is true, src = 0.0
2757 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
) {
2758 shader_addline(buffer
, "SLT %s, -%s, %s;\n", dst_name
, src_name
, src_name
);
2760 /* src contains TA? Write to the dest first. This won't overwrite our destination.
2761 * Then use TA, and calculate the final result
2763 * Not reading from TA? Store the first result in TA to avoid overwriting the
2764 * destination if src reg = dst reg
2766 if(strstr(src_name
, "TA"))
2768 shader_addline(buffer
, "SLT %s, %s, -%s;\n", dst_name
, src_name
, src_name
);
2769 shader_addline(buffer
, "SLT TA, -%s, %s;\n", src_name
, src_name
);
2770 shader_addline(buffer
, "ADD %s, %s, -TA;\n", dst_name
, dst_name
);
2774 shader_addline(buffer
, "SLT TA, -%s, %s;\n", src_name
, src_name
);
2775 shader_addline(buffer
, "SLT %s, %s, -%s;\n", dst_name
, src_name
, src_name
);
2776 shader_addline(buffer
, "ADD %s, TA, -%s;\n", dst_name
, dst_name
);
2781 static void shader_hw_dsy(const struct wined3d_shader_instruction
*ins
)
2783 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2789 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
);
2790 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src
);
2791 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2793 shader_addline(buffer
, "DDY %s, %s;\n", dst
, src
);
2794 shader_addline(buffer
, "MUL%s %s, %s, ycorrection.y;\n", shader_arb_get_modifier(ins
), dst
, dst_name
);
2797 static void shader_hw_pow(const struct wined3d_shader_instruction
*ins
)
2799 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2800 char src0
[50], src1
[50], dst
[50];
2801 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
2802 BOOL need_abs
= FALSE
;
2803 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2804 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2806 /* POW operates on the absolute value of the input */
2807 src0_copy
.modifiers
= abs_modifier(src0_copy
.modifiers
, &need_abs
);
2809 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
);
2810 shader_arb_get_src_param(ins
, &src0_copy
, 0, src0
);
2811 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src1
);
2814 shader_addline(buffer
, "ABS TA.x, %s;\n", src0
);
2816 shader_addline(buffer
, "MOV TA.x, %s;\n", src0
);
2818 if (priv
->target_version
>= NV2
)
2820 shader_addline(buffer
, "MOVC TA.y, %s;\n", src1
);
2821 shader_addline(buffer
, "POW%s %s, TA.x, TA.y;\n", shader_arb_get_modifier(ins
), dst
);
2822 shader_addline(buffer
, "MOV %s (EQ.y), %s;\n", dst
, one
);
2826 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2827 const char *flt_eps
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_EPS
);
2829 shader_addline(buffer
, "ABS TA.y, %s;\n", src1
);
2830 shader_addline(buffer
, "SGE TA.y, -TA.y, %s;\n", zero
);
2831 /* Possibly add flt_eps to avoid getting float special values */
2832 shader_addline(buffer
, "MAD TA.z, TA.y, %s, %s;\n", flt_eps
, src1
);
2833 shader_addline(buffer
, "POW%s TA.x, TA.x, TA.z;\n", shader_arb_get_modifier(ins
));
2834 shader_addline(buffer
, "MAD TA.x, -TA.x, TA.y, TA.x;\n");
2835 shader_addline(buffer
, "MAD %s, TA.y, %s, TA.x;\n", dst
, one
);
2839 static void shader_hw_loop(const struct wined3d_shader_instruction
*ins
)
2841 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2843 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2846 shader_arb_get_src_param(ins
, &ins
->src
[1], 0, src_name
);
2850 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2851 struct list
*e
= list_head(&priv
->control_frames
);
2852 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2854 if(priv
->loop_depth
> 1) shader_addline(buffer
, "PUSHA aL;\n");
2855 /* The constant loader makes sure to load -1 into iX.w */
2856 shader_addline(buffer
, "ARLC aL, %s.xywz;\n", src_name
);
2857 shader_addline(buffer
, "BRA loop_%u_end (LE.x);\n", control_frame
->no
.loop
);
2858 shader_addline(buffer
, "loop_%u_start:\n", control_frame
->no
.loop
);
2862 shader_addline(buffer
, "LOOP %s;\n", src_name
);
2866 static void shader_hw_rep(const struct wined3d_shader_instruction
*ins
)
2868 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2870 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2872 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
);
2874 /* The constant loader makes sure to load -1 into iX.w */
2877 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2878 struct list
*e
= list_head(&priv
->control_frames
);
2879 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2881 if(priv
->loop_depth
> 1) shader_addline(buffer
, "PUSHA aL;\n");
2883 shader_addline(buffer
, "ARLC aL, %s.xywz;\n", src_name
);
2884 shader_addline(buffer
, "BRA loop_%u_end (LE.x);\n", control_frame
->no
.loop
);
2885 shader_addline(buffer
, "loop_%u_start:\n", control_frame
->no
.loop
);
2889 shader_addline(buffer
, "REP %s;\n", src_name
);
2893 static void shader_hw_endloop(const struct wined3d_shader_instruction
*ins
)
2895 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2896 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2900 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2901 struct list
*e
= list_head(&priv
->control_frames
);
2902 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2904 shader_addline(buffer
, "ARAC aL.xy, aL;\n");
2905 shader_addline(buffer
, "BRA loop_%u_start (GT.x);\n", control_frame
->no
.loop
);
2906 shader_addline(buffer
, "loop_%u_end:\n", control_frame
->no
.loop
);
2908 if(priv
->loop_depth
> 1) shader_addline(buffer
, "POPA aL;\n");
2912 shader_addline(buffer
, "ENDLOOP;\n");
2916 static void shader_hw_endrep(const struct wined3d_shader_instruction
*ins
)
2918 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2919 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2923 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2924 struct list
*e
= list_head(&priv
->control_frames
);
2925 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2927 shader_addline(buffer
, "ARAC aL.xy, aL;\n");
2928 shader_addline(buffer
, "BRA loop_%u_start (GT.x);\n", control_frame
->no
.loop
);
2929 shader_addline(buffer
, "loop_%u_end:\n", control_frame
->no
.loop
);
2931 if(priv
->loop_depth
> 1) shader_addline(buffer
, "POPA aL;\n");
2935 shader_addline(buffer
, "ENDREP;\n");
2939 static const struct control_frame
*find_last_loop(const struct shader_arb_ctx_priv
*priv
)
2941 struct control_frame
*control_frame
;
2943 LIST_FOR_EACH_ENTRY(control_frame
, &priv
->control_frames
, struct control_frame
, entry
)
2945 if(control_frame
->type
== LOOP
|| control_frame
->type
== REP
) return control_frame
;
2947 ERR("Could not find loop for break\n");
2951 static void shader_hw_break(const struct wined3d_shader_instruction
*ins
)
2953 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2954 const struct control_frame
*control_frame
= find_last_loop(ins
->ctx
->backend_data
);
2955 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2959 shader_addline(buffer
, "BRA loop_%u_end;\n", control_frame
->no
.loop
);
2963 shader_addline(buffer
, "BRK;\n");
2967 static const char *get_compare(enum wined3d_shader_rel_op op
)
2971 case WINED3D_SHADER_REL_OP_GT
: return "GT";
2972 case WINED3D_SHADER_REL_OP_EQ
: return "EQ";
2973 case WINED3D_SHADER_REL_OP_GE
: return "GE";
2974 case WINED3D_SHADER_REL_OP_LT
: return "LT";
2975 case WINED3D_SHADER_REL_OP_NE
: return "NE";
2976 case WINED3D_SHADER_REL_OP_LE
: return "LE";
2978 FIXME("Unrecognized operator %#x.\n", op
);
2983 static enum wined3d_shader_rel_op
invert_compare(enum wined3d_shader_rel_op op
)
2987 case WINED3D_SHADER_REL_OP_GT
: return WINED3D_SHADER_REL_OP_LE
;
2988 case WINED3D_SHADER_REL_OP_EQ
: return WINED3D_SHADER_REL_OP_NE
;
2989 case WINED3D_SHADER_REL_OP_GE
: return WINED3D_SHADER_REL_OP_LT
;
2990 case WINED3D_SHADER_REL_OP_LT
: return WINED3D_SHADER_REL_OP_GE
;
2991 case WINED3D_SHADER_REL_OP_NE
: return WINED3D_SHADER_REL_OP_EQ
;
2992 case WINED3D_SHADER_REL_OP_LE
: return WINED3D_SHADER_REL_OP_GT
;
2994 FIXME("Unrecognized operator %#x.\n", op
);
2999 static void shader_hw_breakc(const struct wined3d_shader_instruction
*ins
)
3001 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3002 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3003 const struct control_frame
*control_frame
= find_last_loop(ins
->ctx
->backend_data
);
3006 const char *comp
= get_compare(ins
->flags
);
3008 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
3009 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
3013 /* SUBC CC, src0, src1" works only in pixel shaders, so use TA to throw
3014 * away the subtraction result
3016 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3017 shader_addline(buffer
, "BRA loop_%u_end (%s.x);\n", control_frame
->no
.loop
, comp
);
3021 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3022 shader_addline(buffer
, "BRK (%s.x);\n", comp
);
3026 static void shader_hw_ifc(const struct wined3d_shader_instruction
*ins
)
3028 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3029 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3030 struct list
*e
= list_head(&priv
->control_frames
);
3031 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3035 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3037 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
3038 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
3042 /* Invert the flag. We jump to the else label if the condition is NOT true */
3043 comp
= get_compare(invert_compare(ins
->flags
));
3044 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3045 shader_addline(buffer
, "BRA ifc_%u_else (%s.x);\n", control_frame
->no
.ifc
, comp
);
3049 comp
= get_compare(ins
->flags
);
3050 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3051 shader_addline(buffer
, "IF %s.x;\n", comp
);
3055 static void shader_hw_else(const struct wined3d_shader_instruction
*ins
)
3057 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3058 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3059 struct list
*e
= list_head(&priv
->control_frames
);
3060 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3061 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3065 shader_addline(buffer
, "BRA ifc_%u_endif;\n", control_frame
->no
.ifc
);
3066 shader_addline(buffer
, "ifc_%u_else:\n", control_frame
->no
.ifc
);
3067 control_frame
->had_else
= TRUE
;
3071 shader_addline(buffer
, "ELSE;\n");
3075 static void shader_hw_endif(const struct wined3d_shader_instruction
*ins
)
3077 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3078 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3079 struct list
*e
= list_head(&priv
->control_frames
);
3080 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3081 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3085 if(control_frame
->had_else
)
3087 shader_addline(buffer
, "ifc_%u_endif:\n", control_frame
->no
.ifc
);
3091 shader_addline(buffer
, "#No else branch. else is endif\n");
3092 shader_addline(buffer
, "ifc_%u_else:\n", control_frame
->no
.ifc
);
3097 shader_addline(buffer
, "ENDIF;\n");
3101 static void shader_hw_texldd(const struct wined3d_shader_instruction
*ins
)
3103 DWORD sampler_idx
= ins
->src
[1].reg
.idx
[0].offset
;
3105 char reg_src
[3][40];
3106 WORD flags
= TEX_DERIV
;
3108 shader_arb_get_dst_param(ins
, &ins
->dst
[0], reg_dest
);
3109 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_src
[0]);
3110 shader_arb_get_src_param(ins
, &ins
->src
[2], 1, reg_src
[1]);
3111 shader_arb_get_src_param(ins
, &ins
->src
[3], 2, reg_src
[2]);
3113 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) flags
|= TEX_PROJ
;
3114 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) flags
|= TEX_BIAS
;
3116 shader_hw_sample(ins
, sampler_idx
, reg_dest
, reg_src
[0], flags
, reg_src
[1], reg_src
[2]);
3119 static void shader_hw_texldl(const struct wined3d_shader_instruction
*ins
)
3121 DWORD sampler_idx
= ins
->src
[1].reg
.idx
[0].offset
;
3124 WORD flags
= TEX_LOD
;
3126 shader_arb_get_dst_param(ins
, &ins
->dst
[0], reg_dest
);
3127 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_coord
);
3129 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) flags
|= TEX_PROJ
;
3130 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) flags
|= TEX_BIAS
;
3132 shader_hw_sample(ins
, sampler_idx
, reg_dest
, reg_coord
, flags
, NULL
, NULL
);
3135 static void shader_hw_label(const struct wined3d_shader_instruction
*ins
)
3137 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3138 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3140 priv
->in_main_func
= FALSE
;
3141 /* Call instructions activate the NV extensions, not labels and rets. If there is an uncalled
3142 * subroutine, don't generate a label that will make GL complain
3144 if(priv
->target_version
== ARB
) return;
3146 shader_addline(buffer
, "l%u:\n", ins
->src
[0].reg
.idx
[0].offset
);
3149 static void vshader_add_footer(struct shader_arb_ctx_priv
*priv_ctx
,
3150 const struct arb_vshader_private
*shader_data
, const struct arb_vs_compile_args
*args
,
3151 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
,
3152 struct wined3d_shader_buffer
*buffer
)
3156 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
3157 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
3158 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
3159 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
3161 if (args
->super
.fog_src
== VS_FOG_Z
)
3163 shader_addline(buffer
, "MOV result.fogcoord, TMP_OUT.z;\n");
3169 /* posFixup.x is always 1.0, so we can safely use it */
3170 shader_addline(buffer
, "ADD result.fogcoord, posFixup.x, -posFixup.x;\n");
3174 /* Clamp fogcoord */
3175 const char *zero
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ZERO
);
3176 const char *one
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ONE
);
3178 shader_addline(buffer
, "MIN TMP_FOGCOORD.x, TMP_FOGCOORD.x, %s;\n", one
);
3179 shader_addline(buffer
, "MAX result.fogcoord.x, TMP_FOGCOORD.x, %s;\n", zero
);
3183 /* Clipplanes are always stored without y inversion */
3184 if (use_nv_clip(gl_info
) && priv_ctx
->target_version
>= NV2
)
3186 if (args
->super
.clip_enabled
)
3188 for (i
= 0; i
< priv_ctx
->vs_clipplanes
; i
++)
3190 shader_addline(buffer
, "DP4 result.clip[%u].x, TMP_OUT, state.clip[%u].plane;\n", i
, i
);
3194 else if (args
->clip
.boolclip
.clip_texcoord
)
3196 unsigned int cur_clip
= 0;
3197 char component
[4] = {'x', 'y', 'z', 'w'};
3198 const char *zero
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_ZERO
);
3200 for (i
= 0; i
< gl_info
->limits
.clipplanes
; ++i
)
3202 if (args
->clip
.boolclip
.clipplane_mask
& (1 << i
))
3204 shader_addline(buffer
, "DP4 TA.%c, TMP_OUT, state.clip[%u].plane;\n",
3205 component
[cur_clip
++], i
);
3211 shader_addline(buffer
, "MOV TA, %s;\n", zero
);
3214 shader_addline(buffer
, "MOV TA.yzw, %s;\n", zero
);
3217 shader_addline(buffer
, "MOV TA.zw, %s;\n", zero
);
3220 shader_addline(buffer
, "MOV TA.w, %s;\n", zero
);
3223 shader_addline(buffer
, "MOV result.texcoord[%u], TA;\n",
3224 args
->clip
.boolclip
.clip_texcoord
- 1);
3227 /* Write the final position.
3229 * OpenGL coordinates specify the center of the pixel while d3d coords specify
3230 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
3231 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
3232 * contains 1.0 to allow a mad, but arb vs swizzles are too restricted for that.
3234 shader_addline(buffer
, "MUL TA, posFixup, TMP_OUT.w;\n");
3235 shader_addline(buffer
, "ADD TMP_OUT.x, TMP_OUT.x, TA.z;\n");
3236 shader_addline(buffer
, "MAD TMP_OUT.y, TMP_OUT.y, posFixup.y, TA.w;\n");
3238 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
3239 * and the glsl equivalent
3241 if (need_helper_const(shader_data
, reg_maps
, gl_info
))
3243 const char *two
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_TWO
);
3244 shader_addline(buffer
, "MAD TMP_OUT.z, TMP_OUT.z, %s, -TMP_OUT.w;\n", two
);
3248 shader_addline(buffer
, "ADD TMP_OUT.z, TMP_OUT.z, TMP_OUT.z;\n");
3249 shader_addline(buffer
, "ADD TMP_OUT.z, TMP_OUT.z, -TMP_OUT.w;\n");
3252 shader_addline(buffer
, "MOV result.position, TMP_OUT;\n");
3254 priv_ctx
->footer_written
= TRUE
;
3257 static void shader_hw_ret(const struct wined3d_shader_instruction
*ins
)
3259 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3260 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3261 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
3262 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3264 if(priv
->target_version
== ARB
) return;
3268 if (priv
->in_main_func
) vshader_add_footer(priv
, shader
->backend_data
,
3269 priv
->cur_vs_args
, ins
->ctx
->reg_maps
, ins
->ctx
->gl_info
, buffer
);
3272 shader_addline(buffer
, "RET;\n");
3275 static void shader_hw_call(const struct wined3d_shader_instruction
*ins
)
3277 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3278 shader_addline(buffer
, "CAL l%u;\n", ins
->src
[0].reg
.idx
[0].offset
);
3281 /* Context activation is done by the caller. */
3282 static GLuint
create_arb_blt_vertex_program(const struct wined3d_gl_info
*gl_info
)
3284 GLuint program_id
= 0;
3287 static const char blt_vprogram
[] =
3289 "PARAM c[1] = { { 1, 0.5 } };\n"
3290 "MOV result.position, vertex.position;\n"
3291 "MOV result.color, c[0].x;\n"
3292 "MOV result.texcoord[0], vertex.texcoord[0];\n"
3295 GL_EXTCALL(glGenProgramsARB(1, &program_id
));
3296 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, program_id
));
3297 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
3298 strlen(blt_vprogram
), blt_vprogram
));
3299 checkGLcall("glProgramStringARB()");
3301 gl_info
->gl_ops
.gl
.p_glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
3304 FIXME("Vertex program error at position %d: %s\n\n", pos
,
3305 debugstr_a((const char *)gl_info
->gl_ops
.gl
.p_glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
3306 shader_arb_dump_program_source(blt_vprogram
);
3312 GL_EXTCALL(glGetProgramivARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
3313 checkGLcall("glGetProgramivARB()");
3314 if (!native
) WARN("Program exceeds native resource limits.\n");
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 tex_types tex_type
, BOOL masked
)
3324 GLuint program_id
= 0;
3325 const char *fprogram
;
3328 static const char * const blt_fprograms_full
[tex_type_count
] =
3335 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
3336 "MOV result.depth.z, R0.x;\n"
3343 "TEX R0.x, fragment.texcoord[0], texture[0], CUBE;\n"
3344 "MOV result.depth.z, R0.x;\n"
3349 "TEX R0.x, fragment.texcoord[0], texture[0], RECT;\n"
3350 "MOV result.depth.z, R0.x;\n"
3354 static const char * const blt_fprograms_masked
[tex_type_count
] =
3360 "PARAM mask = program.local[0];\n"
3362 "SLT R0.xy, fragment.position, mask.zwzw;\n"
3363 "MUL R0.x, R0.x, R0.y;\n"
3365 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
3366 "MOV result.depth.z, R0.x;\n"
3372 "PARAM mask = program.local[0];\n"
3374 "SLT R0.xy, fragment.position, mask.zwzw;\n"
3375 "MUL R0.x, R0.x, R0.y;\n"
3377 "TEX R0.x, fragment.texcoord[0], texture[0], CUBE;\n"
3378 "MOV result.depth.z, R0.x;\n"
3382 "PARAM mask = program.local[0];\n"
3384 "SLT R0.xy, fragment.position, mask.zwzw;\n"
3385 "MUL R0.x, R0.x, R0.y;\n"
3387 "TEX R0.x, fragment.texcoord[0], texture[0], RECT;\n"
3388 "MOV result.depth.z, R0.x;\n"
3392 fprogram
= masked
? blt_fprograms_masked
[tex_type
] : blt_fprograms_full
[tex_type
];
3395 FIXME("tex_type %#x not supported, falling back to tex_2d\n", tex_type
);
3397 fprogram
= masked
? blt_fprograms_masked
[tex_type
] : blt_fprograms_full
[tex_type
];
3400 GL_EXTCALL(glGenProgramsARB(1, &program_id
));
3401 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, program_id
));
3402 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
, strlen(fprogram
), fprogram
));
3403 checkGLcall("glProgramStringARB()");
3405 gl_info
->gl_ops
.gl
.p_glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
3408 FIXME("Fragment program error at position %d: %s\n\n", pos
,
3409 debugstr_a((const char *)gl_info
->gl_ops
.gl
.p_glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
3410 shader_arb_dump_program_source(fprogram
);
3416 GL_EXTCALL(glGetProgramivARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
3417 checkGLcall("glGetProgramivARB()");
3418 if (!native
) WARN("Program exceeds native resource limits.\n");
3424 static void arbfp_add_sRGB_correction(struct wined3d_shader_buffer
*buffer
, const char *fragcolor
,
3425 const char *tmp1
, const char *tmp2
, const char *tmp3
, const char *tmp4
, BOOL condcode
)
3427 /* Perform sRGB write correction. See GLX_EXT_framebuffer_sRGB */
3431 /* Sigh. MOVC CC doesn't work, so use one of the temps as dummy dest */
3432 shader_addline(buffer
, "SUBC %s, %s.x, srgb_consts1.x;\n", tmp1
, fragcolor
);
3433 /* Calculate the > 0.0031308 case */
3434 shader_addline(buffer
, "POW %s.x (GE), %s.x, srgb_consts0.x;\n", fragcolor
, fragcolor
);
3435 shader_addline(buffer
, "POW %s.y (GE), %s.y, srgb_consts0.x;\n", fragcolor
, fragcolor
);
3436 shader_addline(buffer
, "POW %s.z (GE), %s.z, srgb_consts0.x;\n", fragcolor
, fragcolor
);
3437 shader_addline(buffer
, "MUL %s.xyz (GE), %s, srgb_consts0.y;\n", fragcolor
, fragcolor
);
3438 shader_addline(buffer
, "SUB %s.xyz (GE), %s, srgb_consts0.z;\n", fragcolor
, fragcolor
);
3439 /* Calculate the < case */
3440 shader_addline(buffer
, "MUL %s.xyz (LT), srgb_consts0.w, %s;\n", fragcolor
, fragcolor
);
3444 /* Calculate the > 0.0031308 case */
3445 shader_addline(buffer
, "POW %s.x, %s.x, srgb_consts0.x;\n", tmp1
, fragcolor
);
3446 shader_addline(buffer
, "POW %s.y, %s.y, srgb_consts0.x;\n", tmp1
, fragcolor
);
3447 shader_addline(buffer
, "POW %s.z, %s.z, srgb_consts0.x;\n", tmp1
, fragcolor
);
3448 shader_addline(buffer
, "MUL %s, %s, srgb_consts0.y;\n", tmp1
, tmp1
);
3449 shader_addline(buffer
, "SUB %s, %s, srgb_consts0.z;\n", tmp1
, tmp1
);
3450 /* Calculate the < case */
3451 shader_addline(buffer
, "MUL %s, srgb_consts0.w, %s;\n", tmp2
, fragcolor
);
3452 /* Get 1.0 / 0.0 masks for > 0.0031308 and < 0.0031308 */
3453 shader_addline(buffer
, "SLT %s, srgb_consts1.x, %s;\n", tmp3
, fragcolor
);
3454 shader_addline(buffer
, "SGE %s, srgb_consts1.x, %s;\n", tmp4
, fragcolor
);
3455 /* Store the components > 0.0031308 in the destination */
3456 shader_addline(buffer
, "MUL %s.xyz, %s, %s;\n", fragcolor
, tmp1
, tmp3
);
3457 /* Add the components that are < 0.0031308 */
3458 shader_addline(buffer
, "MAD %s.xyz, %s, %s, %s;\n", fragcolor
, tmp2
, tmp4
, fragcolor
);
3459 /* Move everything into result.color at once. Nvidia hardware cannot handle partial
3460 * result.color writes(.rgb first, then .a), or handle overwriting already written
3461 * components. The assembler uses a temporary register in this case, which is usually
3462 * not allocated from one of our registers that were used earlier.
3465 /* [0.0;1.0] clamping. Not needed, this is done implicitly */
3468 static const DWORD
*find_loop_control_values(const struct wined3d_shader
*shader
, DWORD idx
)
3470 const struct wined3d_shader_lconst
*constant
;
3472 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsI
, struct wined3d_shader_lconst
, entry
)
3474 if (constant
->idx
== idx
)
3476 return constant
->value
;
3482 static void init_ps_input(const struct wined3d_shader
*shader
,
3483 const struct arb_ps_compile_args
*args
, struct shader_arb_ctx_priv
*priv
)
3485 static const char * const texcoords
[8] =
3487 "fragment.texcoord[0]", "fragment.texcoord[1]", "fragment.texcoord[2]", "fragment.texcoord[3]",
3488 "fragment.texcoord[4]", "fragment.texcoord[5]", "fragment.texcoord[6]", "fragment.texcoord[7]"
3491 const struct wined3d_shader_signature_element
*sig
= shader
->input_signature
;
3492 const char *semantic_name
;
3495 switch(args
->super
.vp_mode
)
3497 case pretransformed
:
3499 /* The pixelshader has to collect the varyings on its own. In any case properly load
3500 * color0 and color1. In the case of pretransformed vertices also load texcoords. Set
3501 * other attribs to 0.0.
3503 * For fixedfunction this behavior is correct, according to the tests. For pretransformed
3504 * we'd either need a replacement shader that can load other attribs like BINORMAL, or
3505 * load the texcoord attrib pointers to match the pixel shader signature
3507 for(i
= 0; i
< MAX_REG_INPUT
; i
++)
3509 semantic_name
= sig
[i
].semantic_name
;
3510 semantic_idx
= sig
[i
].semantic_idx
;
3511 if (!semantic_name
) continue;
3513 if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_COLOR
))
3515 if (!semantic_idx
) priv
->ps_input
[i
] = "fragment.color.primary";
3516 else if(semantic_idx
== 1) priv
->ps_input
[i
] = "fragment.color.secondary";
3517 else priv
->ps_input
[i
] = "0.0";
3519 else if(args
->super
.vp_mode
== fixedfunction
)
3521 priv
->ps_input
[i
] = "0.0";
3523 else if(shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
3525 if(semantic_idx
< 8) priv
->ps_input
[i
] = texcoords
[semantic_idx
];
3526 else priv
->ps_input
[i
] = "0.0";
3528 else if(shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_FOG
))
3530 if (!semantic_idx
) priv
->ps_input
[i
] = "fragment.fogcoord";
3531 else priv
->ps_input
[i
] = "0.0";
3535 priv
->ps_input
[i
] = "0.0";
3538 TRACE("v%u, semantic %s%u is %s\n", i
, semantic_name
, semantic_idx
, priv
->ps_input
[i
]);
3543 /* That one is easy. The vertex shaders provide v0-v7 in fragment.texcoord and v8 and v9 in
3546 for(i
= 0; i
< 8; i
++)
3548 priv
->ps_input
[i
] = texcoords
[i
];
3550 priv
->ps_input
[8] = "fragment.color.primary";
3551 priv
->ps_input
[9] = "fragment.color.secondary";
3556 static void arbfp_add_linear_fog(struct wined3d_shader_buffer
*buffer
,
3557 const char *fragcolor
, const char *tmp
)
3559 shader_addline(buffer
, "SUB %s.x, state.fog.params.z, fragment.fogcoord.x;\n", tmp
);
3560 shader_addline(buffer
, "MUL_SAT %s.x, %s.x, state.fog.params.w;\n", tmp
, tmp
);
3561 shader_addline(buffer
, "LRP %s.rgb, %s.x, %s, state.fog.color;\n", fragcolor
, tmp
, fragcolor
);
3564 /* Context activation is done by the caller. */
3565 static GLuint
shader_arb_generate_pshader(const struct wined3d_shader
*shader
,
3566 const struct wined3d_gl_info
*gl_info
, struct wined3d_shader_buffer
*buffer
,
3567 const struct arb_ps_compile_args
*args
, struct arb_ps_compiled_shader
*compiled
)
3569 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
3570 const DWORD
*function
= shader
->function
;
3573 DWORD next_local
= 0;
3574 struct shader_arb_ctx_priv priv_ctx
;
3575 BOOL dcl_td
= FALSE
;
3576 BOOL want_nv_prog
= FALSE
;
3577 struct arb_pshader_private
*shader_priv
= shader
->backend_data
;
3580 BOOL custom_linear_fog
= FALSE
;
3584 unsigned int i
, found
= 0;
3586 for (i
= 0, map
= reg_maps
->temporary
; map
; map
>>= 1, ++i
)
3589 || (shader
->u
.ps
.color0_mov
&& i
== shader
->u
.ps
.color0_reg
)
3590 || (reg_maps
->shader_version
.major
< 2 && !i
))
3593 sprintf(srgbtmp
[found
], "R%u", i
);
3595 if (found
== 4) break;
3600 sprintf(srgbtmp
[0], "TA");
3601 sprintf(srgbtmp
[1], "TB");
3602 sprintf(srgbtmp
[2], "TC");
3603 sprintf(srgbtmp
[3], "TD");
3607 sprintf(srgbtmp
[1], "TA");
3608 sprintf(srgbtmp
[2], "TB");
3609 sprintf(srgbtmp
[3], "TC");
3612 sprintf(srgbtmp
[2], "TA");
3613 sprintf(srgbtmp
[3], "TB");
3616 sprintf(srgbtmp
[3], "TA");
3622 /* Create the hw ARB shader */
3623 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
3624 priv_ctx
.cur_ps_args
= args
;
3625 priv_ctx
.compiled_fprog
= compiled
;
3626 priv_ctx
.cur_np2fixup_info
= &compiled
->np2fixup_info
;
3627 init_ps_input(shader
, args
, &priv_ctx
);
3628 list_init(&priv_ctx
.control_frames
);
3629 priv_ctx
.ps_post_process
= args
->super
.srgb_correction
;
3631 /* Avoid enabling NV_fragment_program* if we do not need it.
3633 * Enabling GL_NV_fragment_program_option causes the driver to occupy a temporary register,
3634 * and it slows down the shader execution noticeably(about 5%). Usually our instruction emulation
3635 * is faster than what we gain from using higher native instructions. There are some things though
3636 * that cannot be emulated. In that case enable the extensions.
3637 * If the extension is enabled, instruction handlers that support both ways will use it.
3639 * Testing shows no performance difference between OPTION NV_fragment_program2 and NV_fragment_program.
3640 * So enable the best we can get.
3642 if(reg_maps
->usesdsx
|| reg_maps
->usesdsy
|| reg_maps
->loop_depth
> 0 || reg_maps
->usestexldd
||
3643 reg_maps
->usestexldl
|| reg_maps
->usesfacing
|| reg_maps
->usesifc
|| reg_maps
->usescall
)
3645 want_nv_prog
= TRUE
;
3648 shader_addline(buffer
, "!!ARBfp1.0\n");
3649 if (want_nv_prog
&& gl_info
->supported
[NV_FRAGMENT_PROGRAM2
])
3651 shader_addline(buffer
, "OPTION NV_fragment_program2;\n");
3652 priv_ctx
.target_version
= NV3
;
3654 else if (want_nv_prog
&& gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
])
3656 shader_addline(buffer
, "OPTION NV_fragment_program;\n");
3657 priv_ctx
.target_version
= NV2
;
3661 /* This is an error - either we're advertising the wrong shader version, or aren't enforcing some
3664 ERR("The shader requires instructions that are not available in plain GL_ARB_fragment_program\n");
3667 priv_ctx
.target_version
= ARB
;
3670 if (reg_maps
->rt_mask
> 1)
3672 shader_addline(buffer
, "OPTION ARB_draw_buffers;\n");
3675 if (reg_maps
->shader_version
.major
< 3)
3677 switch (args
->super
.fog
)
3679 case WINED3D_FFP_PS_FOG_OFF
:
3681 case WINED3D_FFP_PS_FOG_LINEAR
:
3682 if (gl_info
->quirks
& WINED3D_QUIRK_BROKEN_ARB_FOG
)
3684 custom_linear_fog
= TRUE
;
3685 priv_ctx
.ps_post_process
= TRUE
;
3688 shader_addline(buffer
, "OPTION ARB_fog_linear;\n");
3690 case WINED3D_FFP_PS_FOG_EXP
:
3691 shader_addline(buffer
, "OPTION ARB_fog_exp;\n");
3693 case WINED3D_FFP_PS_FOG_EXP2
:
3694 shader_addline(buffer
, "OPTION ARB_fog_exp2;\n");
3699 /* For now always declare the temps. At least the Nvidia assembler optimizes completely
3700 * unused temps away(but occupies them for the whole shader if they're used once). Always
3701 * declaring them avoids tricky bookkeeping work
3703 shader_addline(buffer
, "TEMP TA;\n"); /* Used for modifiers */
3704 shader_addline(buffer
, "TEMP TB;\n"); /* Used for modifiers */
3705 shader_addline(buffer
, "TEMP TC;\n"); /* Used for modifiers */
3706 if(dcl_td
) shader_addline(buffer
, "TEMP TD;\n"); /* Used for sRGB writing */
3707 shader_addline(buffer
, "PARAM coefdiv = { 0.5, 0.25, 0.125, 0.0625 };\n");
3708 shader_addline(buffer
, "PARAM coefmul = { 2, 4, 8, 16 };\n");
3709 wined3d_ftoa(eps
, ftoa_tmp
);
3710 shader_addline(buffer
, "PARAM ps_helper_const = { 0.0, 1.0, %s, 0.0 };\n", ftoa_tmp
);
3712 if (reg_maps
->shader_version
.major
< 2)
3714 strcpy(fragcolor
, "R0");
3718 if (priv_ctx
.ps_post_process
)
3720 if (shader
->u
.ps
.color0_mov
)
3722 sprintf(fragcolor
, "R%u", shader
->u
.ps
.color0_reg
);
3726 shader_addline(buffer
, "TEMP TMP_COLOR;\n");
3727 strcpy(fragcolor
, "TMP_COLOR");
3730 strcpy(fragcolor
, "result.color");
3734 if (args
->super
.srgb_correction
)
3736 shader_addline(buffer
, "PARAM srgb_consts0 = ");
3737 shader_arb_append_imm_vec4(buffer
, wined3d_srgb_const0
);
3738 shader_addline(buffer
, ";\n");
3739 shader_addline(buffer
, "PARAM srgb_consts1 = ");
3740 shader_arb_append_imm_vec4(buffer
, wined3d_srgb_const1
);
3741 shader_addline(buffer
, ";\n");
3744 /* Base Declarations */
3745 shader_generate_arb_declarations(shader
, reg_maps
, buffer
, gl_info
, NULL
, &priv_ctx
);
3747 for (i
= 0, map
= reg_maps
->bumpmat
; map
; map
>>= 1, ++i
)
3749 unsigned char bump_const
;
3751 if (!(map
& 1)) continue;
3753 bump_const
= compiled
->numbumpenvmatconsts
;
3754 compiled
->bumpenvmatconst
[bump_const
].const_num
= WINED3D_CONST_NUM_UNUSED
;
3755 compiled
->bumpenvmatconst
[bump_const
].texunit
= i
;
3756 compiled
->luminanceconst
[bump_const
].const_num
= WINED3D_CONST_NUM_UNUSED
;
3757 compiled
->luminanceconst
[bump_const
].texunit
= i
;
3759 /* We can fit the constants into the constant limit for sure because texbem, texbeml, bem and beml are only supported
3760 * in 1.x shaders, and GL_ARB_fragment_program has a constant limit of 24 constants. So in the worst case we're loading
3761 * 8 shader constants, 8 bump matrices and 8 luminance parameters and are perfectly fine. (No NP2 fixup on bumpmapped
3762 * textures due to conditional NP2 restrictions)
3764 * Use local constants to load the bump env parameters, not program.env. This avoids collisions with d3d constants of
3765 * shaders in newer shader models. Since the bump env parameters have to share their space with NP2 fixup constants,
3766 * their location is shader dependent anyway and they cannot be loaded globally.
3768 compiled
->bumpenvmatconst
[bump_const
].const_num
= next_local
++;
3769 shader_addline(buffer
, "PARAM bumpenvmat%d = program.local[%d];\n",
3770 i
, compiled
->bumpenvmatconst
[bump_const
].const_num
);
3771 compiled
->numbumpenvmatconsts
= bump_const
+ 1;
3773 if (!(reg_maps
->luminanceparams
& (1 << i
))) continue;
3775 compiled
->luminanceconst
[bump_const
].const_num
= next_local
++;
3776 shader_addline(buffer
, "PARAM luminance%d = program.local[%d];\n",
3777 i
, compiled
->luminanceconst
[bump_const
].const_num
);
3780 for(i
= 0; i
< MAX_CONST_I
; i
++)
3782 compiled
->int_consts
[i
] = WINED3D_CONST_NUM_UNUSED
;
3783 if (reg_maps
->integer_constants
& (1 << i
) && priv_ctx
.target_version
>= NV2
)
3785 const DWORD
*control_values
= find_loop_control_values(shader
, i
);
3789 shader_addline(buffer
, "PARAM I%u = {%u, %u, %u, -1};\n", i
,
3790 control_values
[0], control_values
[1], control_values
[2]);
3794 compiled
->int_consts
[i
] = next_local
;
3795 compiled
->num_int_consts
++;
3796 shader_addline(buffer
, "PARAM I%u = program.local[%u];\n", i
, next_local
++);
3801 if(reg_maps
->vpos
|| reg_maps
->usesdsy
)
3803 compiled
->ycorrection
= next_local
;
3804 shader_addline(buffer
, "PARAM ycorrection = program.local[%u];\n", next_local
++);
3808 shader_addline(buffer
, "TEMP vpos;\n");
3809 /* ycorrection.x: Backbuffer height(onscreen) or 0(offscreen).
3810 * ycorrection.y: -1.0(onscreen), 1.0(offscreen)
3811 * ycorrection.z: 1.0
3812 * ycorrection.w: 0.0
3814 shader_addline(buffer
, "MAD vpos, fragment.position, ycorrection.zyww, ycorrection.wxww;\n");
3815 shader_addline(buffer
, "FLR vpos.xy, vpos;\n");
3820 compiled
->ycorrection
= WINED3D_CONST_NUM_UNUSED
;
3823 /* Load constants to fixup NP2 texcoords if there are still free constants left:
3824 * Constants (texture dimensions) for the NP2 fixup are loaded as local program parameters. This will consume
3825 * at most 8 (MAX_FRAGMENT_SAMPLERS / 2) parameters, which is highly unlikely, since the application had to
3826 * use 16 NP2 textures at the same time. In case that we run out of constants the fixup is simply not
3827 * applied / activated. This will probably result in wrong rendering of the texture, but will save us from
3828 * shader compilation errors and the subsequent errors when drawing with this shader. */
3829 if (priv_ctx
.cur_ps_args
->super
.np2_fixup
) {
3830 unsigned char cur_fixup_sampler
= 0;
3832 struct arb_ps_np2fixup_info
* const fixup
= priv_ctx
.cur_np2fixup_info
;
3833 const WORD map
= priv_ctx
.cur_ps_args
->super
.np2_fixup
;
3834 const UINT max_lconsts
= gl_info
->limits
.arb_ps_local_constants
;
3836 fixup
->offset
= next_local
;
3837 fixup
->super
.active
= 0;
3839 for (i
= 0; i
< MAX_FRAGMENT_SAMPLERS
; ++i
) {
3840 if (!(map
& (1 << i
))) continue;
3842 if (fixup
->offset
+ (cur_fixup_sampler
>> 1) < max_lconsts
) {
3843 fixup
->super
.active
|= (1 << i
);
3844 fixup
->super
.idx
[i
] = cur_fixup_sampler
++;
3846 FIXME("No free constant found to load NP2 fixup data into shader. "
3847 "Sampling from this texture will probably look wrong.\n");
3852 fixup
->super
.num_consts
= (cur_fixup_sampler
+ 1) >> 1;
3853 if (fixup
->super
.num_consts
) {
3854 shader_addline(buffer
, "PARAM np2fixup[%u] = { program.env[%u..%u] };\n",
3855 fixup
->super
.num_consts
, fixup
->offset
, fixup
->super
.num_consts
+ fixup
->offset
- 1);
3859 if (shader_priv
->clipplane_emulation
!= ~0U && args
->clip
)
3861 shader_addline(buffer
, "KIL fragment.texcoord[%u];\n", shader_priv
->clipplane_emulation
);
3864 /* Base Shader Body */
3865 shader_generate_main(shader
, buffer
, reg_maps
, function
, &priv_ctx
);
3867 if(args
->super
.srgb_correction
) {
3868 arbfp_add_sRGB_correction(buffer
, fragcolor
, srgbtmp
[0], srgbtmp
[1], srgbtmp
[2], srgbtmp
[3],
3869 priv_ctx
.target_version
>= NV2
);
3872 if (custom_linear_fog
)
3873 arbfp_add_linear_fog(buffer
, fragcolor
, "TA");
3875 if(strcmp(fragcolor
, "result.color")) {
3876 shader_addline(buffer
, "MOV result.color, %s;\n", fragcolor
);
3878 shader_addline(buffer
, "END\n");
3880 /* TODO: change to resource.glObjectHandle or something like that */
3881 GL_EXTCALL(glGenProgramsARB(1, &retval
));
3883 TRACE("Creating a hw pixel shader, prg=%d\n", retval
);
3884 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, retval
));
3886 TRACE("Created hw pixel shader, prg=%d\n", retval
);
3887 /* Create the program and check for errors */
3888 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
3889 buffer
->content_size
, buffer
->buffer
));
3890 checkGLcall("glProgramStringARB()");
3892 gl_info
->gl_ops
.gl
.p_glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &errPos
);
3895 FIXME("HW PixelShader Error at position %d: %s\n\n",
3896 errPos
, debugstr_a((const char *)gl_info
->gl_ops
.gl
.p_glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
3897 shader_arb_dump_program_source(buffer
->buffer
);
3904 GL_EXTCALL(glGetProgramivARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
3905 checkGLcall("glGetProgramivARB()");
3906 if (!native
) WARN("Program exceeds native resource limits.\n");
3912 static int compare_sig(const struct wined3d_shader_signature_element
*sig1
, const struct wined3d_shader_signature_element
*sig2
)
3917 for(i
= 0; i
< MAX_REG_INPUT
; i
++)
3919 if (!sig1
[i
].semantic_name
|| !sig2
[i
].semantic_name
)
3921 /* Compare pointers, not contents. One string is NULL(element does not exist), the other one is not NULL */
3922 if(sig1
[i
].semantic_name
!= sig2
[i
].semantic_name
) return sig1
[i
].semantic_name
< sig2
[i
].semantic_name
? -1 : 1;
3926 if ((ret
= strcmp(sig1
[i
].semantic_name
, sig2
[i
].semantic_name
))) return ret
;
3927 if(sig1
[i
].semantic_idx
!= sig2
[i
].semantic_idx
) return sig1
[i
].semantic_idx
< sig2
[i
].semantic_idx
? -1 : 1;
3928 if(sig1
[i
].sysval_semantic
!= sig2
[i
].sysval_semantic
) return sig1
[i
].sysval_semantic
< sig2
[i
].sysval_semantic
? -1 : 1;
3929 if(sig1
[i
].component_type
!= sig2
[i
].component_type
) return sig1
[i
].component_type
< sig2
[i
].component_type
? -1 : 1;
3930 if(sig1
[i
].register_idx
!= sig2
[i
].register_idx
) return sig1
[i
].register_idx
< sig2
[i
].register_idx
? -1 : 1;
3931 if(sig1
[i
].mask
!= sig2
[i
].mask
) return sig1
[i
].mask
< sig2
[i
].mask
? -1 : 1;
3936 static struct wined3d_shader_signature_element
*clone_sig(const struct wined3d_shader_signature_element
*sig
)
3938 struct wined3d_shader_signature_element
*new;
3942 new = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*new) * MAX_REG_INPUT
);
3943 for(i
= 0; i
< MAX_REG_INPUT
; i
++)
3945 if (!sig
[i
].semantic_name
) continue;
3948 /* Clone the semantic string */
3949 name
= HeapAlloc(GetProcessHeap(), 0, strlen(sig
[i
].semantic_name
) + 1);
3950 strcpy(name
, sig
[i
].semantic_name
);
3951 new[i
].semantic_name
= name
;
3956 static DWORD
find_input_signature(struct shader_arb_priv
*priv
, const struct wined3d_shader_signature_element
*sig
)
3958 struct wine_rb_entry
*entry
= wine_rb_get(&priv
->signature_tree
, sig
);
3959 struct ps_signature
*found_sig
;
3963 found_sig
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
3964 TRACE("Found existing signature %u\n", found_sig
->idx
);
3965 return found_sig
->idx
;
3967 found_sig
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*found_sig
));
3968 found_sig
->sig
= clone_sig(sig
);
3969 found_sig
->idx
= priv
->ps_sig_number
++;
3970 TRACE("New signature stored and assigned number %u\n", found_sig
->idx
);
3971 if(wine_rb_put(&priv
->signature_tree
, sig
, &found_sig
->entry
) == -1)
3973 ERR("Failed to insert program entry.\n");
3975 return found_sig
->idx
;
3978 static void init_output_registers(const struct wined3d_shader
*shader
,
3979 const struct wined3d_shader_signature_element
*ps_input_sig
,
3980 struct shader_arb_ctx_priv
*priv_ctx
, struct arb_vs_compiled_shader
*compiled
)
3983 static const char * const texcoords
[8] =
3985 "result.texcoord[0]", "result.texcoord[1]", "result.texcoord[2]", "result.texcoord[3]",
3986 "result.texcoord[4]", "result.texcoord[5]", "result.texcoord[6]", "result.texcoord[7]"
3988 const char *semantic_name
;
3989 DWORD semantic_idx
, reg_idx
;
3991 /* Write generic input varyings 0 to 7 to result.texcoord[], varying 8 to result.color.primary
3992 * and varying 9 to result.color.secondary
3994 static const char * const decl_idx_to_string
[MAX_REG_INPUT
] =
3996 "result.texcoord[0]", "result.texcoord[1]", "result.texcoord[2]", "result.texcoord[3]",
3997 "result.texcoord[4]", "result.texcoord[5]", "result.texcoord[6]", "result.texcoord[7]",
3998 "result.color.primary", "result.color.secondary"
4003 TRACE("Pixel shader uses builtin varyings\n");
4004 /* Map builtins to builtins */
4005 for(i
= 0; i
< 8; i
++)
4007 priv_ctx
->texcrd_output
[i
] = texcoords
[i
];
4009 priv_ctx
->color_output
[0] = "result.color.primary";
4010 priv_ctx
->color_output
[1] = "result.color.secondary";
4011 priv_ctx
->fog_output
= "TMP_FOGCOORD";
4013 /* Map declared regs to builtins. Use "TA" to /dev/null unread output */
4014 for (i
= 0; i
< (sizeof(shader
->output_signature
) / sizeof(*shader
->output_signature
)); ++i
)
4016 semantic_name
= shader
->output_signature
[i
].semantic_name
;
4017 if (!semantic_name
) continue;
4019 if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_POSITION
))
4021 TRACE("o%u is TMP_OUT\n", i
);
4022 if (!shader
->output_signature
[i
].semantic_idx
) priv_ctx
->vs_output
[i
] = "TMP_OUT";
4023 else priv_ctx
->vs_output
[i
] = "TA";
4025 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_PSIZE
))
4027 TRACE("o%u is result.pointsize\n", i
);
4028 if (!shader
->output_signature
[i
].semantic_idx
) priv_ctx
->vs_output
[i
] = "result.pointsize";
4029 else priv_ctx
->vs_output
[i
] = "TA";
4031 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_COLOR
))
4033 TRACE("o%u is result.color.?, idx %u\n", i
, shader
->output_signature
[i
].semantic_idx
);
4034 if (!shader
->output_signature
[i
].semantic_idx
)
4035 priv_ctx
->vs_output
[i
] = "result.color.primary";
4036 else if (shader
->output_signature
[i
].semantic_idx
== 1)
4037 priv_ctx
->vs_output
[i
] = "result.color.secondary";
4038 else priv_ctx
->vs_output
[i
] = "TA";
4040 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
4042 TRACE("o%u is %s\n", i
, texcoords
[shader
->output_signature
[i
].semantic_idx
]);
4043 if (shader
->output_signature
[i
].semantic_idx
>= 8) priv_ctx
->vs_output
[i
] = "TA";
4044 else priv_ctx
->vs_output
[i
] = texcoords
[shader
->output_signature
[i
].semantic_idx
];
4046 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_FOG
))
4048 TRACE("o%u is result.fogcoord\n", i
);
4049 if (shader
->output_signature
[i
].semantic_idx
> 0) priv_ctx
->vs_output
[i
] = "TA";
4050 else priv_ctx
->vs_output
[i
] = "result.fogcoord";
4054 priv_ctx
->vs_output
[i
] = "TA";
4060 TRACE("Pixel shader uses declared varyings\n");
4062 /* Map builtin to declared. /dev/null the results by default to the TA temp reg */
4063 for(i
= 0; i
< 8; i
++)
4065 priv_ctx
->texcrd_output
[i
] = "TA";
4067 priv_ctx
->color_output
[0] = "TA";
4068 priv_ctx
->color_output
[1] = "TA";
4069 priv_ctx
->fog_output
= "TA";
4071 for(i
= 0; i
< MAX_REG_INPUT
; i
++)
4073 semantic_name
= ps_input_sig
[i
].semantic_name
;
4074 semantic_idx
= ps_input_sig
[i
].semantic_idx
;
4075 reg_idx
= ps_input_sig
[i
].register_idx
;
4076 if (!semantic_name
) continue;
4078 /* If a declared input register is not written by builtin arguments, don't write to it.
4079 * GL_NV_vertex_program makes sure the input defaults to 0.0, which is correct with D3D
4081 * Don't care about POSITION and PSIZE here - this is a builtin vertex shader, position goes
4082 * to TMP_OUT in any case
4084 if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
4086 if (semantic_idx
< 8)
4087 priv_ctx
->texcrd_output
[semantic_idx
] = decl_idx_to_string
[reg_idx
];
4089 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_COLOR
))
4091 if (semantic_idx
< 2)
4092 priv_ctx
->color_output
[semantic_idx
] = decl_idx_to_string
[reg_idx
];
4094 else if(shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_FOG
))
4097 priv_ctx
->fog_output
= decl_idx_to_string
[reg_idx
];
4104 if (!strcmp(decl_idx_to_string
[reg_idx
], "result.color.primary")
4105 || !strcmp(decl_idx_to_string
[reg_idx
], "result.color.secondary"))
4107 compiled
->need_color_unclamp
= TRUE
;
4111 /* Map declared to declared */
4112 for (i
= 0; i
< (sizeof(shader
->output_signature
) / sizeof(*shader
->output_signature
)); ++i
)
4114 /* Write unread output to TA to throw them away */
4115 priv_ctx
->vs_output
[i
] = "TA";
4116 semantic_name
= shader
->output_signature
[i
].semantic_name
;
4117 if (!semantic_name
) continue;
4119 if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_POSITION
)
4120 && !shader
->output_signature
[i
].semantic_idx
)
4122 priv_ctx
->vs_output
[i
] = "TMP_OUT";
4125 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_PSIZE
)
4126 && !shader
->output_signature
[i
].semantic_idx
)
4128 priv_ctx
->vs_output
[i
] = "result.pointsize";
4132 for(j
= 0; j
< MAX_REG_INPUT
; j
++)
4134 if (!ps_input_sig
[j
].semantic_name
) continue;
4136 if (!strcmp(ps_input_sig
[j
].semantic_name
, semantic_name
)
4137 && ps_input_sig
[j
].semantic_idx
== shader
->output_signature
[i
].semantic_idx
)
4139 priv_ctx
->vs_output
[i
] = decl_idx_to_string
[ps_input_sig
[j
].register_idx
];
4141 if (!strcmp(priv_ctx
->vs_output
[i
], "result.color.primary")
4142 || !strcmp(priv_ctx
->vs_output
[i
], "result.color.secondary"))
4144 compiled
->need_color_unclamp
= TRUE
;
4151 /* Context activation is done by the caller. */
4152 static GLuint
shader_arb_generate_vshader(const struct wined3d_shader
*shader
,
4153 const struct wined3d_gl_info
*gl_info
, struct wined3d_shader_buffer
*buffer
,
4154 const struct arb_vs_compile_args
*args
, struct arb_vs_compiled_shader
*compiled
,
4155 const struct wined3d_shader_signature_element
*ps_input_sig
)
4157 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
4158 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
4159 struct shader_arb_priv
*priv
= shader
->device
->shader_priv
;
4160 const DWORD
*function
= shader
->function
;
4162 DWORD next_local
= 0;
4163 struct shader_arb_ctx_priv priv_ctx
;
4167 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
4168 priv_ctx
.cur_vs_args
= args
;
4169 list_init(&priv_ctx
.control_frames
);
4170 init_output_registers(shader
, ps_input_sig
, &priv_ctx
, compiled
);
4172 /* Create the hw ARB shader */
4173 shader_addline(buffer
, "!!ARBvp1.0\n");
4175 /* Always enable the NV extension if available. Unlike fragment shaders, there is no
4176 * mesurable performance penalty, and we can always make use of it for clipplanes.
4178 if (gl_info
->supported
[NV_VERTEX_PROGRAM3
])
4180 shader_addline(buffer
, "OPTION NV_vertex_program3;\n");
4181 priv_ctx
.target_version
= NV3
;
4182 shader_addline(buffer
, "ADDRESS aL;\n");
4184 else if (gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
])
4186 shader_addline(buffer
, "OPTION NV_vertex_program2;\n");
4187 priv_ctx
.target_version
= NV2
;
4188 shader_addline(buffer
, "ADDRESS aL;\n");
4190 priv_ctx
.target_version
= ARB
;
4193 shader_addline(buffer
, "TEMP TMP_OUT;\n");
4195 shader_addline(buffer
, "TEMP TMP_FOGCOORD;\n");
4196 if (need_helper_const(shader_data
, reg_maps
, gl_info
))
4199 wined3d_ftoa(eps
, ftoa_tmp
);
4200 shader_addline(buffer
, "PARAM helper_const = { 0.0, 1.0, 2.0, %s};\n", ftoa_tmp
);
4202 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
))
4204 shader_addline(buffer
, "PARAM rel_addr_const = { 0.5, %d.0, 0.0, 0.0 };\n", shader_data
->rel_offset
);
4205 shader_addline(buffer
, "TEMP A0_SHADOW;\n");
4208 shader_addline(buffer
, "TEMP TA;\n");
4209 shader_addline(buffer
, "TEMP TB;\n");
4211 /* Base Declarations */
4212 shader_generate_arb_declarations(shader
, reg_maps
, buffer
, gl_info
,
4213 &priv_ctx
.vs_clipplanes
, &priv_ctx
);
4215 for(i
= 0; i
< MAX_CONST_I
; i
++)
4217 compiled
->int_consts
[i
] = WINED3D_CONST_NUM_UNUSED
;
4218 if(reg_maps
->integer_constants
& (1 << i
) && priv_ctx
.target_version
>= NV2
)
4220 const DWORD
*control_values
= find_loop_control_values(shader
, i
);
4224 shader_addline(buffer
, "PARAM I%u = {%u, %u, %u, -1};\n", i
,
4225 control_values
[0], control_values
[1], control_values
[2]);
4229 compiled
->int_consts
[i
] = next_local
;
4230 compiled
->num_int_consts
++;
4231 shader_addline(buffer
, "PARAM I%u = program.local[%u];\n", i
, next_local
++);
4236 /* We need a constant to fixup the final position */
4237 shader_addline(buffer
, "PARAM posFixup = program.local[%u];\n", next_local
);
4238 compiled
->pos_fixup
= next_local
++;
4240 /* Initialize output parameters. GL_ARB_vertex_program does not require special initialization values
4241 * for output parameters. D3D in theory does not do that either, but some applications depend on a
4242 * proper initialization of the secondary color, and programs using the fixed function pipeline without
4243 * a replacement shader depend on the texcoord.w being set properly.
4245 * GL_NV_vertex_program defines that all output values are initialized to {0.0, 0.0, 0.0, 1.0}. This
4246 * assertion is in effect even when using GL_ARB_vertex_program without any NV specific additions. So
4247 * skip this if NV_vertex_program is supported. Otherwise, initialize the secondary color. For the tex-
4248 * coords, we have a flag in the opengl caps. Many cards do not require the texcoord being set, and
4249 * this can eat a number of instructions, so skip it unless this cap is set as well
4251 if (!gl_info
->supported
[NV_VERTEX_PROGRAM
])
4253 const char *color_init
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_0001
);
4254 shader_addline(buffer
, "MOV result.color.secondary, %s;\n", color_init
);
4256 if (gl_info
->quirks
& WINED3D_QUIRK_SET_TEXCOORD_W
&& !priv
->ffp_proj_control
)
4259 const char *one
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_ONE
);
4260 for(i
= 0; i
< min(8, MAX_REG_TEXCRD
); i
++)
4262 if (reg_maps
->texcoord_mask
[i
] && reg_maps
->texcoord_mask
[i
] != WINED3DSP_WRITEMASK_ALL
)
4263 shader_addline(buffer
, "MOV result.texcoord[%u].w, %s\n", i
, one
);
4268 /* The shader starts with the main function */
4269 priv_ctx
.in_main_func
= TRUE
;
4270 /* Base Shader Body */
4271 shader_generate_main(shader
, buffer
, reg_maps
, function
, &priv_ctx
);
4273 if (!priv_ctx
.footer_written
) vshader_add_footer(&priv_ctx
,
4274 shader_data
, args
, reg_maps
, gl_info
, buffer
);
4276 shader_addline(buffer
, "END\n");
4278 /* TODO: change to resource.glObjectHandle or something like that */
4279 GL_EXTCALL(glGenProgramsARB(1, &ret
));
4281 TRACE("Creating a hw vertex shader, prg=%d\n", ret
);
4282 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, ret
));
4284 TRACE("Created hw vertex shader, prg=%d\n", ret
);
4285 /* Create the program and check for errors */
4286 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
4287 buffer
->content_size
, buffer
->buffer
));
4288 checkGLcall("glProgramStringARB()");
4290 gl_info
->gl_ops
.gl
.p_glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &errPos
);
4293 FIXME("HW VertexShader Error at position %d: %s\n\n",
4294 errPos
, debugstr_a((const char *)gl_info
->gl_ops
.gl
.p_glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
4295 shader_arb_dump_program_source(buffer
->buffer
);
4302 GL_EXTCALL(glGetProgramivARB(GL_VERTEX_PROGRAM_ARB
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
4303 checkGLcall("glGetProgramivARB()");
4304 if (!native
) WARN("Program exceeds native resource limits.\n");
4310 /* Context activation is done by the caller. */
4311 static struct arb_ps_compiled_shader
*find_arb_pshader(struct wined3d_shader
*shader
,
4312 const struct arb_ps_compile_args
*args
)
4314 struct wined3d_device
*device
= shader
->device
;
4315 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
4316 const struct wined3d_d3d_info
*d3d_info
= &device
->adapter
->d3d_info
;
4319 struct arb_ps_compiled_shader
*new_array
;
4320 struct wined3d_shader_buffer buffer
;
4321 struct arb_pshader_private
*shader_data
;
4324 if (!shader
->backend_data
)
4326 struct shader_arb_priv
*priv
= device
->shader_priv
;
4328 shader
->backend_data
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
4329 shader_data
= shader
->backend_data
;
4330 shader_data
->clamp_consts
= shader
->reg_maps
.shader_version
.major
== 1;
4332 if (shader
->reg_maps
.shader_version
.major
< 3)
4333 shader_data
->input_signature_idx
= ~0U;
4335 shader_data
->input_signature_idx
= find_input_signature(priv
, shader
->input_signature
);
4337 TRACE("Shader got assigned input signature index %u\n", shader_data
->input_signature_idx
);
4339 if (!d3d_info
->vs_clipping
)
4340 shader_data
->clipplane_emulation
= shader_find_free_input_register(&shader
->reg_maps
,
4341 d3d_info
->limits
.ffp_blend_stages
- 1);
4343 shader_data
->clipplane_emulation
= ~0U;
4345 shader_data
= shader
->backend_data
;
4347 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4348 * so a linear search is more performant than a hashmap or a binary search
4349 * (cache coherency etc)
4351 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4353 if (!memcmp(&shader_data
->gl_shaders
[i
].args
, args
, sizeof(*args
)))
4354 return &shader_data
->gl_shaders
[i
];
4357 TRACE("No matching GL shader found, compiling a new shader\n");
4358 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4359 if (shader_data
->num_gl_shaders
)
4361 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4362 new_array
= HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, shader_data
->gl_shaders
,
4363 new_size
* sizeof(*shader_data
->gl_shaders
));
4365 new_array
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
->gl_shaders
));
4370 ERR("Out of memory\n");
4373 shader_data
->gl_shaders
= new_array
;
4374 shader_data
->shader_array_size
= new_size
;
4377 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4379 pixelshader_update_samplers(shader
, args
->super
.tex_types
);
4381 if (!shader_buffer_init(&buffer
))
4383 ERR("Failed to initialize shader buffer.\n");
4387 ret
= shader_arb_generate_pshader(shader
, gl_info
, &buffer
, args
,
4388 &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
]);
4389 shader_buffer_free(&buffer
);
4390 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].prgId
= ret
;
4392 return &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++];
4395 static inline BOOL
vs_args_equal(const struct arb_vs_compile_args
*stored
, const struct arb_vs_compile_args
*new,
4396 const DWORD use_map
, BOOL skip_int
) {
4397 if((stored
->super
.swizzle_map
& use_map
) != new->super
.swizzle_map
) return FALSE
;
4398 if(stored
->super
.clip_enabled
!= new->super
.clip_enabled
) return FALSE
;
4399 if(stored
->super
.fog_src
!= new->super
.fog_src
) return FALSE
;
4400 if(stored
->clip
.boolclip_compare
!= new->clip
.boolclip_compare
) return FALSE
;
4401 if(stored
->ps_signature
!= new->ps_signature
) return FALSE
;
4402 if(stored
->vertex
.samplers_compare
!= new->vertex
.samplers_compare
) return FALSE
;
4403 if(skip_int
) return TRUE
;
4405 return !memcmp(stored
->loop_ctrl
, new->loop_ctrl
, sizeof(stored
->loop_ctrl
));
4408 static struct arb_vs_compiled_shader
*find_arb_vshader(struct wined3d_shader
*shader
,
4409 const struct wined3d_gl_info
*gl_info
, DWORD use_map
, const struct arb_vs_compile_args
*args
,
4410 const struct wined3d_shader_signature_element
*ps_input_sig
)
4414 struct arb_vs_compiled_shader
*new_array
;
4415 struct wined3d_shader_buffer buffer
;
4416 struct arb_vshader_private
*shader_data
;
4419 if (!shader
->backend_data
)
4421 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
4423 shader
->backend_data
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
4424 shader_data
= shader
->backend_data
;
4426 if ((gl_info
->quirks
& WINED3D_QUIRK_ARB_VS_OFFSET_LIMIT
)
4427 && reg_maps
->min_rel_offset
<= reg_maps
->max_rel_offset
)
4429 if (reg_maps
->max_rel_offset
- reg_maps
->min_rel_offset
> 127)
4431 FIXME("The difference between the minimum and maximum relative offset is > 127.\n");
4432 FIXME("Which this OpenGL implementation does not support. Try using GLSL.\n");
4433 FIXME("Min: %u, Max: %u.\n", reg_maps
->min_rel_offset
, reg_maps
->max_rel_offset
);
4435 else if (reg_maps
->max_rel_offset
- reg_maps
->min_rel_offset
> 63)
4436 shader_data
->rel_offset
= reg_maps
->min_rel_offset
+ 63;
4437 else if (reg_maps
->max_rel_offset
> 63)
4438 shader_data
->rel_offset
= reg_maps
->min_rel_offset
;
4441 shader_data
= shader
->backend_data
;
4443 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4444 * so a linear search is more performant than a hashmap or a binary search
4445 * (cache coherency etc)
4447 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
4448 if (vs_args_equal(&shader_data
->gl_shaders
[i
].args
, args
,
4449 use_map
, gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]))
4451 return &shader_data
->gl_shaders
[i
];
4455 TRACE("No matching GL shader found, compiling a new shader\n");
4457 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4458 if (shader_data
->num_gl_shaders
)
4460 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4461 new_array
= HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, shader_data
->gl_shaders
,
4462 new_size
* sizeof(*shader_data
->gl_shaders
));
4464 new_array
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
->gl_shaders
));
4469 ERR("Out of memory\n");
4472 shader_data
->gl_shaders
= new_array
;
4473 shader_data
->shader_array_size
= new_size
;
4476 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4478 if (!shader_buffer_init(&buffer
))
4480 ERR("Failed to initialize shader buffer.\n");
4484 ret
= shader_arb_generate_vshader(shader
, gl_info
, &buffer
, args
,
4485 &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
],
4487 shader_buffer_free(&buffer
);
4488 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].prgId
= ret
;
4490 return &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++];
4493 static void find_arb_ps_compile_args(const struct wined3d_state
*state
,
4494 const struct wined3d_context
*context
, const struct wined3d_shader
*shader
,
4495 struct arb_ps_compile_args
*args
)
4497 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4498 const struct wined3d_d3d_info
*d3d_info
= context
->d3d_info
;
4502 find_ps_compile_args(state
, shader
, context
->stream_info
.position_transformed
, &args
->super
, gl_info
);
4504 /* This forces all local boolean constants to 1 to make them stateblock independent */
4505 args
->bools
= shader
->reg_maps
.local_bool_consts
;
4507 for(i
= 0; i
< MAX_CONST_B
; i
++)
4509 if (state
->ps_consts_b
[i
])
4510 args
->bools
|= ( 1 << i
);
4513 /* Only enable the clip plane emulation KIL if at least one clipplane is enabled. The KIL instruction
4514 * is quite expensive because it forces the driver to disable early Z discards. It is cheaper to
4515 * duplicate the shader than have a no-op KIL instruction in every shader
4517 if (!d3d_info
->vs_clipping
&& use_vs(state
)
4518 && state
->render_states
[WINED3D_RS_CLIPPING
]
4519 && state
->render_states
[WINED3D_RS_CLIPPLANEENABLE
])
4524 /* Skip if unused or local, or supported natively */
4525 int_skip
= ~shader
->reg_maps
.integer_constants
| shader
->reg_maps
.local_int_consts
;
4526 if (int_skip
== 0xffff || gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
])
4528 memset(args
->loop_ctrl
, 0, sizeof(args
->loop_ctrl
));
4532 for(i
= 0; i
< MAX_CONST_I
; i
++)
4534 if(int_skip
& (1 << i
))
4536 args
->loop_ctrl
[i
][0] = 0;
4537 args
->loop_ctrl
[i
][1] = 0;
4538 args
->loop_ctrl
[i
][2] = 0;
4542 args
->loop_ctrl
[i
][0] = state
->ps_consts_i
[i
* 4];
4543 args
->loop_ctrl
[i
][1] = state
->ps_consts_i
[i
* 4 + 1];
4544 args
->loop_ctrl
[i
][2] = state
->ps_consts_i
[i
* 4 + 2];
4549 static void find_arb_vs_compile_args(const struct wined3d_state
*state
,
4550 const struct wined3d_context
*context
, const struct wined3d_shader
*shader
,
4551 struct arb_vs_compile_args
*args
)
4553 const struct wined3d_device
*device
= shader
->device
;
4554 const struct wined3d_adapter
*adapter
= device
->adapter
;
4555 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4556 const struct wined3d_d3d_info
*d3d_info
= context
->d3d_info
;
4560 find_vs_compile_args(state
, shader
, context
->stream_info
.swizzle_map
, &args
->super
);
4562 args
->clip
.boolclip_compare
= 0;
4565 const struct wined3d_shader
*ps
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
4566 const struct arb_pshader_private
*shader_priv
= ps
->backend_data
;
4567 args
->ps_signature
= shader_priv
->input_signature_idx
;
4569 args
->clip
.boolclip
.clip_texcoord
= shader_priv
->clipplane_emulation
+ 1;
4573 args
->ps_signature
= ~0;
4574 if (!d3d_info
->vs_clipping
&& adapter
->fragment_pipe
== &arbfp_fragment_pipeline
)
4575 args
->clip
.boolclip
.clip_texcoord
= ffp_clip_emul(context
) ? d3d_info
->limits
.ffp_blend_stages
: 0;
4576 /* Otherwise: Setting boolclip_compare set clip_texcoord to 0 */
4579 if (args
->clip
.boolclip
.clip_texcoord
)
4581 if (state
->render_states
[WINED3D_RS_CLIPPING
])
4582 args
->clip
.boolclip
.clipplane_mask
= (unsigned char)state
->render_states
[WINED3D_RS_CLIPPLANEENABLE
];
4583 /* clipplane_mask was set to 0 by setting boolclip_compare to 0 */
4586 /* This forces all local boolean constants to 1 to make them stateblock independent */
4587 args
->clip
.boolclip
.bools
= shader
->reg_maps
.local_bool_consts
;
4588 /* TODO: Figure out if it would be better to store bool constants as bitmasks in the stateblock */
4589 for(i
= 0; i
< MAX_CONST_B
; i
++)
4591 if (state
->vs_consts_b
[i
])
4592 args
->clip
.boolclip
.bools
|= ( 1 << i
);
4595 args
->vertex
.samplers
[0] = context
->tex_unit_map
[MAX_FRAGMENT_SAMPLERS
+ 0];
4596 args
->vertex
.samplers
[1] = context
->tex_unit_map
[MAX_FRAGMENT_SAMPLERS
+ 1];
4597 args
->vertex
.samplers
[2] = context
->tex_unit_map
[MAX_FRAGMENT_SAMPLERS
+ 2];
4598 args
->vertex
.samplers
[3] = 0;
4600 /* Skip if unused or local */
4601 int_skip
= ~shader
->reg_maps
.integer_constants
| shader
->reg_maps
.local_int_consts
;
4602 /* This is about flow control, not clipping. */
4603 if (int_skip
== 0xffff || gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
])
4605 memset(args
->loop_ctrl
, 0, sizeof(args
->loop_ctrl
));
4609 for(i
= 0; i
< MAX_CONST_I
; i
++)
4611 if(int_skip
& (1 << i
))
4613 args
->loop_ctrl
[i
][0] = 0;
4614 args
->loop_ctrl
[i
][1] = 0;
4615 args
->loop_ctrl
[i
][2] = 0;
4619 args
->loop_ctrl
[i
][0] = state
->vs_consts_i
[i
* 4];
4620 args
->loop_ctrl
[i
][1] = state
->vs_consts_i
[i
* 4 + 1];
4621 args
->loop_ctrl
[i
][2] = state
->vs_consts_i
[i
* 4 + 2];
4626 /* Context activation is done by the caller. */
4627 static void shader_arb_select(void *shader_priv
, struct wined3d_context
*context
,
4628 const struct wined3d_state
*state
)
4630 struct shader_arb_priv
*priv
= shader_priv
;
4631 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4634 /* Deal with pixel shaders first so the vertex shader arg function has the input signature ready */
4637 struct wined3d_shader
*ps
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
4638 struct arb_ps_compile_args compile_args
;
4639 struct arb_ps_compiled_shader
*compiled
;
4641 TRACE("Using pixel shader %p.\n", ps
);
4642 find_arb_ps_compile_args(state
, context
, ps
, &compile_args
);
4643 compiled
= find_arb_pshader(ps
, &compile_args
);
4644 priv
->current_fprogram_id
= compiled
->prgId
;
4645 priv
->compiled_fprog
= compiled
;
4647 /* Bind the fragment program */
4648 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, priv
->current_fprogram_id
));
4649 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->current_fprogram_id);");
4651 if (!priv
->use_arbfp_fixed_func
)
4652 priv
->fragment_pipe
->enable_extension(gl_info
, FALSE
);
4654 /* Enable OpenGL fragment programs. */
4655 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
4656 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
4658 TRACE("Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", priv
->current_fprogram_id
);
4660 /* Pixel Shader 1.x constants are clamped to [-1;1], Pixel Shader 2.0 constants are not. If switching between
4661 * a 1.x and newer shader, reload the first 8 constants
4663 if (priv
->last_ps_const_clamped
!= ((struct arb_pshader_private
*)ps
->backend_data
)->clamp_consts
)
4665 priv
->last_ps_const_clamped
= ((struct arb_pshader_private
*)ps
->backend_data
)->clamp_consts
;
4666 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, 8);
4667 for(i
= 0; i
< 8; i
++)
4669 priv
->pshader_const_dirty
[i
] = 1;
4671 /* Also takes care of loading local constants */
4672 shader_arb_load_constants_internal(shader_priv
, context
, state
, TRUE
, FALSE
, TRUE
);
4676 UINT rt_height
= state
->fb
->render_targets
[0]->resource
.height
;
4677 shader_arb_ps_local_constants(compiled
, context
, state
, rt_height
);
4680 /* Force constant reloading for the NP2 fixup (see comment in shader_glsl_select for more info) */
4681 if (compiled
->np2fixup_info
.super
.active
)
4682 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_NP2_FIXUP
;
4684 if (ps
->load_local_constsF
)
4685 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_F
;
4689 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
] && !priv
->use_arbfp_fixed_func
)
4691 /* Disable only if we're not using arbfp fixed function fragment
4692 * processing. If this is used, keep GL_FRAGMENT_PROGRAM_ARB
4693 * enabled, and the fixed function pipeline will bind the fixed
4694 * function replacement shader. */
4695 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4696 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4697 priv
->current_fprogram_id
= 0;
4699 priv
->fragment_pipe
->enable_extension(gl_info
, TRUE
);
4704 struct wined3d_shader
*vs
= state
->shader
[WINED3D_SHADER_TYPE_VERTEX
];
4705 struct arb_vs_compile_args compile_args
;
4706 struct arb_vs_compiled_shader
*compiled
;
4707 const struct wined3d_shader_signature_element
*ps_input_sig
;
4709 TRACE("Using vertex shader %p\n", vs
);
4710 find_arb_vs_compile_args(state
, context
, vs
, &compile_args
);
4712 /* Instead of searching for the signature in the signature list, read the one from the
4713 * current pixel shader. It's maybe not the shader where the signature came from, but it
4714 * is the same signature and faster to find. */
4715 if (compile_args
.ps_signature
== ~0U)
4716 ps_input_sig
= NULL
;
4718 ps_input_sig
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
]->input_signature
;
4720 compiled
= find_arb_vshader(vs
, context
->gl_info
, context
->stream_info
.use_map
,
4721 &compile_args
, ps_input_sig
);
4722 priv
->current_vprogram_id
= compiled
->prgId
;
4723 priv
->compiled_vprog
= compiled
;
4725 /* Bind the vertex program */
4726 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->current_vprogram_id
));
4727 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->current_vprogram_id);");
4729 priv
->vertex_pipe
->vp_enable(gl_info
, FALSE
);
4731 /* Enable OpenGL vertex programs */
4732 gl_info
->gl_ops
.gl
.p_glEnable(GL_VERTEX_PROGRAM_ARB
);
4733 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
4734 TRACE("Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", priv
->current_vprogram_id
);
4735 shader_arb_vs_local_constants(compiled
, context
, state
);
4737 if(priv
->last_vs_color_unclamp
!= compiled
->need_color_unclamp
) {
4738 priv
->last_vs_color_unclamp
= compiled
->need_color_unclamp
;
4740 if (gl_info
->supported
[ARB_COLOR_BUFFER_FLOAT
])
4742 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, !compiled
->need_color_unclamp
));
4743 checkGLcall("glClampColorARB");
4745 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
4749 if (vs
->load_local_constsF
)
4750 context
->constant_update_mask
|= WINED3D_SHADER_CONST_VS_F
;
4754 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
4756 priv
->current_vprogram_id
= 0;
4757 gl_info
->gl_ops
.gl
.p_glDisable(GL_VERTEX_PROGRAM_ARB
);
4758 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4760 priv
->vertex_pipe
->vp_enable(gl_info
, TRUE
);
4765 /* Context activation is done by the caller. */
4766 static void shader_arb_disable(void *shader_priv
, struct wined3d_context
*context
)
4768 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4769 struct shader_arb_priv
*priv
= shader_priv
;
4771 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
4773 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4774 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4775 priv
->current_fprogram_id
= 0;
4777 priv
->fragment_pipe
->enable_extension(gl_info
, FALSE
);
4779 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
4781 priv
->current_vprogram_id
= 0;
4782 gl_info
->gl_ops
.gl
.p_glDisable(GL_VERTEX_PROGRAM_ARB
);
4783 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4785 priv
->vertex_pipe
->vp_enable(gl_info
, FALSE
);
4787 if (gl_info
->supported
[ARB_COLOR_BUFFER_FLOAT
] && priv
->last_vs_color_unclamp
)
4789 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, GL_FIXED_ONLY_ARB
));
4790 checkGLcall("glClampColorARB");
4791 priv
->last_vs_color_unclamp
= FALSE
;
4794 context
->shader_update_mask
= (1 << WINED3D_SHADER_TYPE_PIXEL
)
4795 | (1 << WINED3D_SHADER_TYPE_VERTEX
)
4796 | (1 << WINED3D_SHADER_TYPE_GEOMETRY
);
4799 /* Context activation is done by the caller. */
4800 static void shader_arb_select_depth_blt(void *shader_priv
, const struct wined3d_gl_info
*gl_info
,
4801 enum tex_types tex_type
, const SIZE
*ds_mask_size
)
4803 const float mask
[] = {0.0f
, 0.0f
, (float)ds_mask_size
->cx
, (float)ds_mask_size
->cy
};
4804 BOOL masked
= ds_mask_size
->cx
&& ds_mask_size
->cy
;
4805 struct shader_arb_priv
*priv
= shader_priv
;
4806 GLuint
*blt_fprogram
;
4808 if (!priv
->depth_blt_vprogram_id
) priv
->depth_blt_vprogram_id
= create_arb_blt_vertex_program(gl_info
);
4809 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->depth_blt_vprogram_id
));
4810 gl_info
->gl_ops
.gl
.p_glEnable(GL_VERTEX_PROGRAM_ARB
);
4812 blt_fprogram
= masked
? &priv
->depth_blt_fprogram_id_masked
[tex_type
] : &priv
->depth_blt_fprogram_id_full
[tex_type
];
4813 if (!*blt_fprogram
) *blt_fprogram
= create_arb_blt_fragment_program(gl_info
, tex_type
, masked
);
4814 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, *blt_fprogram
));
4815 if (masked
) GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, 0, mask
));
4816 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
4819 /* Context activation is done by the caller. */
4820 static void shader_arb_deselect_depth_blt(void *shader_priv
, const struct wined3d_gl_info
*gl_info
)
4822 struct shader_arb_priv
*priv
= shader_priv
;
4824 if (priv
->current_vprogram_id
) {
4825 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->current_vprogram_id
));
4826 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
4828 TRACE("Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB.\n", priv
->current_vprogram_id
);
4832 gl_info
->gl_ops
.gl
.p_glDisable(GL_VERTEX_PROGRAM_ARB
);
4833 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4836 if (priv
->current_fprogram_id
) {
4837 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, priv
->current_fprogram_id
));
4838 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
4840 TRACE("Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB.\n", priv
->current_fprogram_id
);
4842 else if(!priv
->use_arbfp_fixed_func
)
4844 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4845 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4849 static void shader_arb_destroy(struct wined3d_shader
*shader
)
4851 struct wined3d_device
*device
= shader
->device
;
4852 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
4854 if (shader_is_pshader_version(shader
->reg_maps
.shader_version
.type
))
4856 struct arb_pshader_private
*shader_data
= shader
->backend_data
;
4859 if(!shader_data
) return; /* This can happen if a shader was never compiled */
4861 if (shader_data
->num_gl_shaders
)
4863 struct wined3d_context
*context
= context_acquire(device
, NULL
);
4865 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4867 GL_EXTCALL(glDeleteProgramsARB(1, &shader_data
->gl_shaders
[i
].prgId
));
4868 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader_data->gl_shaders[i].prgId))");
4871 context_release(context
);
4874 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
);
4875 HeapFree(GetProcessHeap(), 0, shader_data
);
4876 shader
->backend_data
= NULL
;
4880 struct arb_vshader_private
*shader_data
= shader
->backend_data
;
4883 if(!shader_data
) return; /* This can happen if a shader was never compiled */
4885 if (shader_data
->num_gl_shaders
)
4887 struct wined3d_context
*context
= context_acquire(device
, NULL
);
4889 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4891 GL_EXTCALL(glDeleteProgramsARB(1, &shader_data
->gl_shaders
[i
].prgId
));
4892 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader_data->gl_shaders[i].prgId))");
4895 context_release(context
);
4898 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
);
4899 HeapFree(GetProcessHeap(), 0, shader_data
);
4900 shader
->backend_data
= NULL
;
4904 static int sig_tree_compare(const void *key
, const struct wine_rb_entry
*entry
)
4906 struct ps_signature
*e
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
4907 return compare_sig(key
, e
->sig
);
4910 static const struct wine_rb_functions sig_tree_functions
=
4918 static HRESULT
shader_arb_alloc(struct wined3d_device
*device
, const struct wined3d_vertex_pipe_ops
*vertex_pipe
,
4919 const struct fragment_pipeline
*fragment_pipe
)
4921 struct shader_arb_priv
*priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*priv
));
4922 struct fragment_caps fragment_caps
;
4923 void *vertex_priv
, *fragment_priv
;
4924 const struct wined3d_d3d_info
*d3d_info
= &device
->adapter
->d3d_info
;
4926 if (!(vertex_priv
= vertex_pipe
->vp_alloc(&arb_program_shader_backend
, priv
)))
4928 ERR("Failed to initialize vertex pipe.\n");
4929 HeapFree(GetProcessHeap(), 0, priv
);
4933 if (!(fragment_priv
= fragment_pipe
->alloc_private(&arb_program_shader_backend
, priv
)))
4935 ERR("Failed to initialize fragment pipe.\n");
4936 vertex_pipe
->vp_free(device
);
4937 HeapFree(GetProcessHeap(), 0, priv
);
4941 priv
->vshader_const_dirty
= HeapAlloc(GetProcessHeap(), 0,
4942 sizeof(*priv
->vshader_const_dirty
) * d3d_info
->limits
.vs_uniform_count
);
4943 if (!priv
->vshader_const_dirty
)
4945 memset(priv
->vshader_const_dirty
, 1,
4946 sizeof(*priv
->vshader_const_dirty
) * d3d_info
->limits
.vs_uniform_count
);
4948 priv
->pshader_const_dirty
= HeapAlloc(GetProcessHeap(), 0,
4949 sizeof(*priv
->pshader_const_dirty
) * d3d_info
->limits
.ps_uniform_count
);
4950 if (!priv
->pshader_const_dirty
)
4952 memset(priv
->pshader_const_dirty
, 1,
4953 sizeof(*priv
->pshader_const_dirty
) * d3d_info
->limits
.ps_uniform_count
);
4955 if(wine_rb_init(&priv
->signature_tree
, &sig_tree_functions
) == -1)
4957 ERR("RB tree init failed\n");
4961 priv
->vertex_pipe
= vertex_pipe
;
4962 priv
->fragment_pipe
= fragment_pipe
;
4963 fragment_pipe
->get_caps(&device
->adapter
->gl_info
, &fragment_caps
);
4964 priv
->ffp_proj_control
= fragment_caps
.wined3d_caps
& WINED3D_FRAGMENT_CAP_PROJ_CONTROL
;
4966 device
->vertex_priv
= vertex_priv
;
4967 device
->fragment_priv
= fragment_priv
;
4968 device
->shader_priv
= priv
;
4973 HeapFree(GetProcessHeap(), 0, priv
->pshader_const_dirty
);
4974 HeapFree(GetProcessHeap(), 0, priv
->vshader_const_dirty
);
4975 fragment_pipe
->free_private(device
);
4976 vertex_pipe
->vp_free(device
);
4977 HeapFree(GetProcessHeap(), 0, priv
);
4978 return E_OUTOFMEMORY
;
4981 static void release_signature(struct wine_rb_entry
*entry
, void *context
)
4983 struct ps_signature
*sig
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
4985 for(i
= 0; i
< MAX_REG_INPUT
; i
++)
4987 HeapFree(GetProcessHeap(), 0, (char *) sig
->sig
[i
].semantic_name
);
4989 HeapFree(GetProcessHeap(), 0, sig
->sig
);
4990 HeapFree(GetProcessHeap(), 0, sig
);
4993 /* Context activation is done by the caller. */
4994 static void shader_arb_free(struct wined3d_device
*device
)
4996 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
4997 struct shader_arb_priv
*priv
= device
->shader_priv
;
5000 if (priv
->depth_blt_vprogram_id
)
5001 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_vprogram_id
));
5003 for (i
= 0; i
< tex_type_count
; ++i
)
5005 if (priv
->depth_blt_fprogram_id_full
[i
])
5007 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_fprogram_id_full
[i
]));
5009 if (priv
->depth_blt_fprogram_id_masked
[i
])
5011 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_fprogram_id_masked
[i
]));
5015 wine_rb_destroy(&priv
->signature_tree
, release_signature
, NULL
);
5016 HeapFree(GetProcessHeap(), 0, priv
->pshader_const_dirty
);
5017 HeapFree(GetProcessHeap(), 0, priv
->vshader_const_dirty
);
5018 priv
->fragment_pipe
->free_private(device
);
5019 priv
->vertex_pipe
->vp_free(device
);
5020 HeapFree(GetProcessHeap(), 0, device
->shader_priv
);
5023 static BOOL
shader_arb_allocate_context_data(struct wined3d_context
*context
)
5028 static void shader_arb_free_context_data(struct wined3d_context
*context
)
5030 struct shader_arb_priv
*priv
;
5032 if (!context
->swapchain
)
5035 priv
= context
->swapchain
->device
->shader_priv
;
5036 if (priv
->last_context
== context
)
5037 priv
->last_context
= NULL
;
5040 static void shader_arb_get_caps(const struct wined3d_gl_info
*gl_info
, struct shader_caps
*caps
)
5042 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
5047 /* 96 is the minimum allowed value of MAX_PROGRAM_ENV_PARAMETERS_ARB
5048 * for vertex programs. If the native limit is less than that it's
5049 * not very useful, and e.g. Mesa swrast returns 0, probably to
5050 * indicate it's a software implementation. */
5051 if (gl_info
->limits
.arb_vs_native_constants
< 96)
5052 vs_consts
= gl_info
->limits
.arb_vs_float_constants
;
5054 vs_consts
= min(gl_info
->limits
.arb_vs_float_constants
, gl_info
->limits
.arb_vs_native_constants
);
5056 if (gl_info
->supported
[NV_VERTEX_PROGRAM3
])
5059 TRACE("Hardware vertex shader version 3.0 enabled (NV_VERTEX_PROGRAM3)\n");
5061 else if (vs_consts
>= 256)
5063 /* Shader Model 2.0 requires at least 256 vertex shader constants */
5065 TRACE("Hardware vertex shader version 2.0 enabled (ARB_PROGRAM)\n");
5070 TRACE("Hardware vertex shader version 1.1 enabled (ARB_PROGRAM)\n");
5072 caps
->vs_version
= min(wined3d_settings
.max_sm_vs
, vs_version
);
5073 caps
->vs_uniform_count
= vs_consts
;
5077 caps
->vs_version
= 0;
5078 caps
->vs_uniform_count
= 0;
5081 caps
->gs_version
= 0;
5083 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
5088 /* Similar as above for vertex programs, but the minimum for fragment
5089 * programs is 24. */
5090 if (gl_info
->limits
.arb_ps_native_constants
< 24)
5091 ps_consts
= gl_info
->limits
.arb_ps_float_constants
;
5093 ps_consts
= min(gl_info
->limits
.arb_ps_float_constants
, gl_info
->limits
.arb_ps_native_constants
);
5095 if (gl_info
->supported
[NV_FRAGMENT_PROGRAM2
])
5098 TRACE("Hardware pixel shader version 3.0 enabled (NV_FRAGMENT_PROGRAM2)\n");
5100 else if (ps_consts
>= 32)
5102 /* Shader Model 2.0 requires at least 32 pixel shader constants */
5104 TRACE("Hardware pixel shader version 2.0 enabled (ARB_PROGRAM)\n");
5109 TRACE("Hardware pixel shader version 1.4 enabled (ARB_PROGRAM)\n");
5111 caps
->ps_version
= min(wined3d_settings
.max_sm_ps
, ps_version
);
5112 caps
->ps_uniform_count
= ps_consts
;
5113 caps
->ps_1x_max_value
= 8.0f
;
5117 caps
->ps_version
= 0;
5118 caps
->ps_uniform_count
= 0;
5119 caps
->ps_1x_max_value
= 0.0f
;
5122 caps
->wined3d_caps
= WINED3D_SHADER_CAP_SRGB_WRITE
;
5123 if (use_nv_clip(gl_info
))
5124 caps
->wined3d_caps
|= WINED3D_SHADER_CAP_VS_CLIPPING
;
5127 static BOOL
shader_arb_color_fixup_supported(struct color_fixup_desc fixup
)
5129 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
5131 TRACE("Checking support for color_fixup:\n");
5132 dump_color_fixup_desc(fixup
);
5135 /* We support everything except complex conversions. */
5136 if (!is_complex_fixup(fixup
))
5142 TRACE("[FAILED]\n");
5146 static void shader_arb_add_instruction_modifiers(const struct wined3d_shader_instruction
*ins
) {
5148 char write_mask
[20], regstr
[50];
5149 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
5150 BOOL is_color
= FALSE
;
5151 const struct wined3d_shader_dst_param
*dst
;
5153 if (!ins
->dst_count
) return;
5157 if (!shift
) return; /* Saturate alone is handled by the instructions */
5159 shader_arb_get_write_mask(ins
, dst
, write_mask
);
5160 shader_arb_get_register_name(ins
, &dst
->reg
, regstr
, &is_color
);
5162 /* Generate a line that does the output modifier computation
5163 * FIXME: _SAT vs shift? _SAT alone is already handled in the instructions, if this
5164 * maps problems in e.g. _d4_sat modify shader_arb_get_modifier
5166 shader_addline(buffer
, "MUL%s %s%s, %s, %s;\n", shader_arb_get_modifier(ins
),
5167 regstr
, write_mask
, regstr
, shift_tab
[shift
]);
5170 static const SHADER_HANDLER shader_arb_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
5172 /* WINED3DSIH_ABS */ shader_hw_map2gl
,
5173 /* WINED3DSIH_ADD */ shader_hw_map2gl
,
5174 /* WINED3DSIH_AND */ NULL
,
5175 /* WINED3DSIH_BEM */ pshader_hw_bem
,
5176 /* WINED3DSIH_BREAK */ shader_hw_break
,
5177 /* WINED3DSIH_BREAKC */ shader_hw_breakc
,
5178 /* WINED3DSIH_BREAKP */ NULL
,
5179 /* WINED3DSIH_CALL */ shader_hw_call
,
5180 /* WINED3DSIH_CALLNZ */ NULL
,
5181 /* WINED3DSIH_CMP */ pshader_hw_cmp
,
5182 /* WINED3DSIH_CND */ pshader_hw_cnd
,
5183 /* WINED3DSIH_CRS */ shader_hw_map2gl
,
5184 /* WINED3DSIH_CUT */ NULL
,
5185 /* WINED3DSIH_DCL */ shader_hw_nop
,
5186 /* WINED3DSIH_DCL_CONSTANT_BUFFER */ shader_hw_nop
,
5187 /* WINED3DSIH_DCL_INPUT_PRIMITIVE */ shader_hw_nop
,
5188 /* WINED3DSIH_DCL_OUTPUT_TOPOLOGY */ shader_hw_nop
,
5189 /* WINED3DSIH_DCL_VERTICES_OUT */ shader_hw_nop
,
5190 /* WINED3DSIH_DEF */ shader_hw_nop
,
5191 /* WINED3DSIH_DEFB */ shader_hw_nop
,
5192 /* WINED3DSIH_DEFI */ shader_hw_nop
,
5193 /* WINED3DSIH_DIV */ NULL
,
5194 /* WINED3DSIH_DP2 */ NULL
,
5195 /* WINED3DSIH_DP2ADD */ pshader_hw_dp2add
,
5196 /* WINED3DSIH_DP3 */ shader_hw_map2gl
,
5197 /* WINED3DSIH_DP4 */ shader_hw_map2gl
,
5198 /* WINED3DSIH_DST */ shader_hw_map2gl
,
5199 /* WINED3DSIH_DSX */ shader_hw_map2gl
,
5200 /* WINED3DSIH_DSY */ shader_hw_dsy
,
5201 /* WINED3DSIH_ELSE */ shader_hw_else
,
5202 /* WINED3DSIH_EMIT */ NULL
,
5203 /* WINED3DSIH_ENDIF */ shader_hw_endif
,
5204 /* WINED3DSIH_ENDLOOP */ shader_hw_endloop
,
5205 /* WINED3DSIH_ENDREP */ shader_hw_endrep
,
5206 /* WINED3DSIH_EQ */ NULL
,
5207 /* WINED3DSIH_EXP */ shader_hw_scalar_op
,
5208 /* WINED3DSIH_EXPP */ shader_hw_scalar_op
,
5209 /* WINED3DSIH_FRC */ shader_hw_map2gl
,
5210 /* WINED3DSIH_FTOI */ NULL
,
5211 /* WINED3DSIH_GE */ NULL
,
5212 /* WINED3DSIH_IADD */ NULL
,
5213 /* WINED3DSIH_IEQ */ NULL
,
5214 /* WINED3DSIH_IF */ NULL
/* Hardcoded into the shader */,
5215 /* WINED3DSIH_IFC */ shader_hw_ifc
,
5216 /* WINED3DSIH_IGE */ NULL
,
5217 /* WINED3DSIH_IMUL */ NULL
,
5218 /* WINED3DSIH_ISHL */ NULL
,
5219 /* WINED3DSIH_ITOF */ NULL
,
5220 /* WINED3DSIH_LABEL */ shader_hw_label
,
5221 /* WINED3DSIH_LD */ NULL
,
5222 /* WINED3DSIH_LIT */ shader_hw_map2gl
,
5223 /* WINED3DSIH_LOG */ shader_hw_scalar_op
,
5224 /* WINED3DSIH_LOGP */ shader_hw_scalar_op
,
5225 /* WINED3DSIH_LOOP */ shader_hw_loop
,
5226 /* WINED3DSIH_LRP */ shader_hw_lrp
,
5227 /* WINED3DSIH_LT */ NULL
,
5228 /* WINED3DSIH_M3x2 */ shader_hw_mnxn
,
5229 /* WINED3DSIH_M3x3 */ shader_hw_mnxn
,
5230 /* WINED3DSIH_M3x4 */ shader_hw_mnxn
,
5231 /* WINED3DSIH_M4x3 */ shader_hw_mnxn
,
5232 /* WINED3DSIH_M4x4 */ shader_hw_mnxn
,
5233 /* WINED3DSIH_MAD */ shader_hw_map2gl
,
5234 /* WINED3DSIH_MAX */ shader_hw_map2gl
,
5235 /* WINED3DSIH_MIN */ shader_hw_map2gl
,
5236 /* WINED3DSIH_MOV */ shader_hw_mov
,
5237 /* WINED3DSIH_MOVA */ shader_hw_mov
,
5238 /* WINED3DSIH_MOVC */ NULL
,
5239 /* WINED3DSIH_MUL */ shader_hw_map2gl
,
5240 /* WINED3DSIH_NOP */ shader_hw_nop
,
5241 /* WINED3DSIH_NRM */ shader_hw_nrm
,
5242 /* WINED3DSIH_PHASE */ shader_hw_nop
,
5243 /* WINED3DSIH_POW */ shader_hw_pow
,
5244 /* WINED3DSIH_RCP */ shader_hw_scalar_op
,
5245 /* WINED3DSIH_REP */ shader_hw_rep
,
5246 /* WINED3DSIH_RET */ shader_hw_ret
,
5247 /* WINED3DSIH_ROUND_NI */ NULL
,
5248 /* WINED3DSIH_RSQ */ shader_hw_scalar_op
,
5249 /* WINED3DSIH_SAMPLE */ NULL
,
5250 /* WINED3DSIH_SAMPLE_GRAD */ NULL
,
5251 /* WINED3DSIH_SAMPLE_LOD */ NULL
,
5252 /* WINED3DSIH_SETP */ NULL
,
5253 /* WINED3DSIH_SGE */ shader_hw_map2gl
,
5254 /* WINED3DSIH_SGN */ shader_hw_sgn
,
5255 /* WINED3DSIH_SINCOS */ shader_hw_sincos
,
5256 /* WINED3DSIH_SLT */ shader_hw_map2gl
,
5257 /* WINED3DSIH_SQRT */ NULL
,
5258 /* WINED3DSIH_SUB */ shader_hw_map2gl
,
5259 /* WINED3DSIH_TEX */ pshader_hw_tex
,
5260 /* WINED3DSIH_TEXBEM */ pshader_hw_texbem
,
5261 /* WINED3DSIH_TEXBEML */ pshader_hw_texbem
,
5262 /* WINED3DSIH_TEXCOORD */ pshader_hw_texcoord
,
5263 /* WINED3DSIH_TEXDEPTH */ pshader_hw_texdepth
,
5264 /* WINED3DSIH_TEXDP3 */ pshader_hw_texdp3
,
5265 /* WINED3DSIH_TEXDP3TEX */ pshader_hw_texdp3tex
,
5266 /* WINED3DSIH_TEXKILL */ pshader_hw_texkill
,
5267 /* WINED3DSIH_TEXLDD */ shader_hw_texldd
,
5268 /* WINED3DSIH_TEXLDL */ shader_hw_texldl
,
5269 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_hw_texm3x2depth
,
5270 /* WINED3DSIH_TEXM3x2PAD */ pshader_hw_texm3x2pad
,
5271 /* WINED3DSIH_TEXM3x2TEX */ pshader_hw_texm3x2tex
,
5272 /* WINED3DSIH_TEXM3x3 */ pshader_hw_texm3x3
,
5273 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
5274 /* WINED3DSIH_TEXM3x3PAD */ pshader_hw_texm3x3pad
,
5275 /* WINED3DSIH_TEXM3x3SPEC */ pshader_hw_texm3x3spec
,
5276 /* WINED3DSIH_TEXM3x3TEX */ pshader_hw_texm3x3tex
,
5277 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_hw_texm3x3vspec
,
5278 /* WINED3DSIH_TEXREG2AR */ pshader_hw_texreg2ar
,
5279 /* WINED3DSIH_TEXREG2GB */ pshader_hw_texreg2gb
,
5280 /* WINED3DSIH_TEXREG2RGB */ pshader_hw_texreg2rgb
,
5281 /* WINED3DSIH_UDIV */ NULL
,
5282 /* WINED3DSIH_USHR */ NULL
,
5283 /* WINED3DSIH_UTOF */ NULL
,
5284 /* WINED3DSIH_XOR */ NULL
,
5287 static BOOL
get_bool_const(const struct wined3d_shader_instruction
*ins
,
5288 const struct wined3d_shader
*shader
, DWORD idx
)
5290 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
5291 BOOL vshader
= shader_is_vshader_version(reg_maps
->shader_version
.type
);
5292 const struct wined3d_shader_lconst
*constant
;
5294 WORD flag
= (1 << idx
);
5295 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5297 if (reg_maps
->local_bool_consts
& flag
)
5299 /* What good is an if(bool) with a hardcoded local constant? I don't know, but handle it */
5300 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsB
, struct wined3d_shader_lconst
, entry
)
5302 if (constant
->idx
== idx
)
5304 return constant
->value
[0];
5307 ERR("Local constant not found\n");
5312 if(vshader
) bools
= priv
->cur_vs_args
->clip
.boolclip
.bools
;
5313 else bools
= priv
->cur_ps_args
->bools
;
5314 return bools
& flag
;
5318 static void get_loop_control_const(const struct wined3d_shader_instruction
*ins
,
5319 const struct wined3d_shader
*shader
, UINT idx
, struct wined3d_shader_loop_control
*loop_control
)
5321 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
5322 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5324 /* Integer constants can either be a local constant, or they can be stored in the shader
5325 * type specific compile args. */
5326 if (reg_maps
->local_int_consts
& (1 << idx
))
5328 const struct wined3d_shader_lconst
*constant
;
5330 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsI
, struct wined3d_shader_lconst
, entry
)
5332 if (constant
->idx
== idx
)
5334 loop_control
->count
= constant
->value
[0];
5335 loop_control
->start
= constant
->value
[1];
5336 /* Step is signed. */
5337 loop_control
->step
= (int)constant
->value
[2];
5341 /* If this happens the flag was set incorrectly */
5342 ERR("Local constant not found\n");
5343 loop_control
->count
= 0;
5344 loop_control
->start
= 0;
5345 loop_control
->step
= 0;
5349 switch (reg_maps
->shader_version
.type
)
5351 case WINED3D_SHADER_TYPE_VERTEX
:
5352 /* Count and aL start value are unsigned */
5353 loop_control
->count
= priv
->cur_vs_args
->loop_ctrl
[idx
][0];
5354 loop_control
->start
= priv
->cur_vs_args
->loop_ctrl
[idx
][1];
5355 /* Step is signed. */
5356 loop_control
->step
= ((char)priv
->cur_vs_args
->loop_ctrl
[idx
][2]);
5359 case WINED3D_SHADER_TYPE_PIXEL
:
5360 loop_control
->count
= priv
->cur_ps_args
->loop_ctrl
[idx
][0];
5361 loop_control
->start
= priv
->cur_ps_args
->loop_ctrl
[idx
][1];
5362 loop_control
->step
= ((char)priv
->cur_ps_args
->loop_ctrl
[idx
][2]);
5366 FIXME("Unhandled shader type %#x.\n", reg_maps
->shader_version
.type
);
5371 static void record_instruction(struct list
*list
, const struct wined3d_shader_instruction
*ins
)
5374 struct wined3d_shader_dst_param
*dst_param
= NULL
;
5375 struct wined3d_shader_src_param
*src_param
= NULL
, *rel_addr
= NULL
;
5376 struct recorded_instruction
*rec
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*rec
));
5379 ERR("Out of memory\n");
5384 dst_param
= HeapAlloc(GetProcessHeap(), 0, sizeof(*dst_param
));
5385 if(!dst_param
) goto free
;
5386 *dst_param
= *ins
->dst
;
5387 if (ins
->dst
->reg
.idx
[0].rel_addr
)
5389 rel_addr
= HeapAlloc(GetProcessHeap(), 0, sizeof(*rel_addr
));
5392 *rel_addr
= *ins
->dst
->reg
.idx
[0].rel_addr
;
5393 dst_param
->reg
.idx
[0].rel_addr
= rel_addr
;
5395 rec
->ins
.dst
= dst_param
;
5397 src_param
= HeapAlloc(GetProcessHeap(), 0, sizeof(*src_param
) * ins
->src_count
);
5400 for (i
= 0; i
< ins
->src_count
; ++i
)
5402 src_param
[i
] = ins
->src
[i
];
5403 if (ins
->src
[i
].reg
.idx
[0].rel_addr
)
5405 rel_addr
= HeapAlloc(GetProcessHeap(), 0, sizeof(*rel_addr
));
5408 *rel_addr
= *ins
->src
[i
].reg
.idx
[0].rel_addr
;
5409 src_param
[i
].reg
.idx
[0].rel_addr
= rel_addr
;
5412 rec
->ins
.src
= src_param
;
5413 list_add_tail(list
, &rec
->entry
);
5417 ERR("Out of memory\n");
5420 HeapFree(GetProcessHeap(), 0, (void *)dst_param
->reg
.idx
[0].rel_addr
);
5421 HeapFree(GetProcessHeap(), 0, dst_param
);
5425 for(i
= 0; i
< ins
->src_count
; i
++)
5427 HeapFree(GetProcessHeap(), 0, (void *)src_param
[i
].reg
.idx
[0].rel_addr
);
5429 HeapFree(GetProcessHeap(), 0, src_param
);
5431 HeapFree(GetProcessHeap(), 0, rec
);
5434 static void free_recorded_instruction(struct list
*list
)
5436 struct recorded_instruction
*rec_ins
, *entry2
;
5439 LIST_FOR_EACH_ENTRY_SAFE(rec_ins
, entry2
, list
, struct recorded_instruction
, entry
)
5441 list_remove(&rec_ins
->entry
);
5442 if (rec_ins
->ins
.dst
)
5444 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.dst
->reg
.idx
[0].rel_addr
);
5445 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.dst
);
5447 if (rec_ins
->ins
.src
)
5449 for (i
= 0; i
< rec_ins
->ins
.src_count
; ++i
)
5451 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.src
[i
].reg
.idx
[0].rel_addr
);
5453 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.src
);
5455 HeapFree(GetProcessHeap(), 0, rec_ins
);
5459 static void pop_control_frame(const struct wined3d_shader_instruction
*ins
)
5461 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5462 struct control_frame
*control_frame
;
5464 if (ins
->handler_idx
== WINED3DSIH_ENDLOOP
|| ins
->handler_idx
== WINED3DSIH_ENDREP
)
5466 struct list
*e
= list_head(&priv
->control_frames
);
5467 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5468 list_remove(&control_frame
->entry
);
5469 HeapFree(GetProcessHeap(), 0, control_frame
);
5472 else if (ins
->handler_idx
== WINED3DSIH_ENDIF
)
5474 /* Non-ifc ENDIFs were already handled previously. */
5475 struct list
*e
= list_head(&priv
->control_frames
);
5476 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5477 list_remove(&control_frame
->entry
);
5478 HeapFree(GetProcessHeap(), 0, control_frame
);
5482 static void shader_arb_handle_instruction(const struct wined3d_shader_instruction
*ins
) {
5483 SHADER_HANDLER hw_fct
;
5484 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5485 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
5486 struct control_frame
*control_frame
;
5487 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
5490 if(ins
->handler_idx
== WINED3DSIH_LOOP
|| ins
->handler_idx
== WINED3DSIH_REP
)
5492 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
5493 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5495 if(ins
->handler_idx
== WINED3DSIH_LOOP
) control_frame
->type
= LOOP
;
5496 if(ins
->handler_idx
== WINED3DSIH_REP
) control_frame
->type
= REP
;
5498 if(priv
->target_version
>= NV2
)
5500 control_frame
->no
.loop
= priv
->num_loops
++;
5505 /* Don't bother recording when we're in a not used if branch */
5511 if(!priv
->recording
)
5513 list_init(&priv
->record
);
5514 priv
->recording
= TRUE
;
5515 control_frame
->outer_loop
= TRUE
;
5516 get_loop_control_const(ins
, shader
, ins
->src
[0].reg
.idx
[0].offset
, &control_frame
->loop_control
);
5517 return; /* Instruction is handled */
5519 /* Record this loop in the outer loop's recording */
5522 else if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
|| ins
->handler_idx
== WINED3DSIH_ENDREP
)
5524 if(priv
->target_version
>= NV2
)
5526 /* Nothing to do. The control frame is popped after the HW instr handler */
5530 struct list
*e
= list_head(&priv
->control_frames
);
5531 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5532 list_remove(&control_frame
->entry
);
5534 if(control_frame
->outer_loop
)
5536 unsigned int iteration
;
5540 /* Turn off recording before playback */
5541 priv
->recording
= FALSE
;
5543 /* Move the recorded instructions to a separate list and get them out of the private data
5544 * structure. If there are nested loops, the shader_arb_handle_instruction below will
5545 * be recorded again, thus priv->record might be overwritten
5548 list_move_tail(©
, &priv
->record
);
5549 list_init(&priv
->record
);
5551 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5553 shader_addline(buffer
, "#unrolling loop: %u iterations, aL=%u, inc %d\n",
5554 control_frame
->loop_control
.count
, control_frame
->loop_control
.start
,
5555 control_frame
->loop_control
.step
);
5556 aL
= control_frame
->loop_control
.start
;
5560 shader_addline(buffer
, "#unrolling rep: %u iterations\n", control_frame
->loop_control
.count
);
5563 for (iteration
= 0; iteration
< control_frame
->loop_control
.count
; ++iteration
)
5565 struct recorded_instruction
*rec_ins
;
5566 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5569 shader_addline(buffer
, "#Iteration %u, aL=%d\n", iteration
, aL
);
5573 shader_addline(buffer
, "#Iteration %u\n", iteration
);
5576 LIST_FOR_EACH_ENTRY(rec_ins
, ©
, struct recorded_instruction
, entry
)
5578 shader_arb_handle_instruction(&rec_ins
->ins
);
5581 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5583 aL
+= control_frame
->loop_control
.step
;
5586 shader_addline(buffer
, "#end loop/rep\n");
5588 free_recorded_instruction(©
);
5589 HeapFree(GetProcessHeap(), 0, control_frame
);
5590 return; /* Instruction is handled */
5594 /* This is a nested loop. Proceed to the normal recording function */
5595 HeapFree(GetProcessHeap(), 0, control_frame
);
5602 record_instruction(&priv
->record
, ins
);
5607 if(ins
->handler_idx
== WINED3DSIH_IF
)
5609 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
5610 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5611 control_frame
->type
= IF
;
5613 bool_const
= get_bool_const(ins
, shader
, ins
->src
[0].reg
.idx
[0].offset
);
5614 if (ins
->src
[0].modifiers
== WINED3DSPSM_NOT
)
5615 bool_const
= !bool_const
;
5616 if (!priv
->muted
&& !bool_const
)
5618 shader_addline(buffer
, "#if(FALSE){\n");
5620 control_frame
->muting
= TRUE
;
5622 else shader_addline(buffer
, "#if(TRUE) {\n");
5624 return; /* Instruction is handled */
5626 else if(ins
->handler_idx
== WINED3DSIH_IFC
)
5628 /* IF(bool) and if_cond(a, b) use the same ELSE and ENDIF tokens */
5629 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
5630 control_frame
->type
= IFC
;
5631 control_frame
->no
.ifc
= priv
->num_ifcs
++;
5632 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5634 else if(ins
->handler_idx
== WINED3DSIH_ELSE
)
5636 struct list
*e
= list_head(&priv
->control_frames
);
5637 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5639 if(control_frame
->type
== IF
)
5641 shader_addline(buffer
, "#} else {\n");
5642 if(!priv
->muted
&& !control_frame
->muting
)
5645 control_frame
->muting
= TRUE
;
5647 else if(control_frame
->muting
) priv
->muted
= FALSE
;
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");
5654 else if(ins
->handler_idx
== WINED3DSIH_ENDIF
)
5656 struct list
*e
= list_head(&priv
->control_frames
);
5657 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5659 if(control_frame
->type
== IF
)
5661 shader_addline(buffer
, "#} endif\n");
5662 if(control_frame
->muting
) priv
->muted
= FALSE
;
5663 list_remove(&control_frame
->entry
);
5664 HeapFree(GetProcessHeap(), 0, control_frame
);
5665 return; /* Instruction is handled */
5667 /* In case of an ifc, generate a HW shader instruction */
5668 if (control_frame
->type
!= IFC
)
5669 ERR("Control frame does not match.\n");
5674 pop_control_frame(ins
);
5678 /* Select handler */
5679 hw_fct
= shader_arb_instruction_handler_table
[ins
->handler_idx
];
5681 /* Unhandled opcode */
5684 FIXME("Backend can't handle opcode %#x\n", ins
->handler_idx
);
5689 pop_control_frame(ins
);
5691 shader_arb_add_instruction_modifiers(ins
);
5694 static BOOL
shader_arb_has_ffp_proj_control(void *shader_priv
)
5696 struct shader_arb_priv
*priv
= shader_priv
;
5698 return priv
->ffp_proj_control
;
5701 const struct wined3d_shader_backend_ops arb_program_shader_backend
=
5703 shader_arb_handle_instruction
,
5706 shader_arb_select_depth_blt
,
5707 shader_arb_deselect_depth_blt
,
5708 shader_arb_update_float_vertex_constants
,
5709 shader_arb_update_float_pixel_constants
,
5710 shader_arb_load_constants
,
5714 shader_arb_allocate_context_data
,
5715 shader_arb_free_context_data
,
5716 shader_arb_get_caps
,
5717 shader_arb_color_fixup_supported
,
5718 shader_arb_has_ffp_proj_control
,
5721 /* ARB_fragment_program fixed function pipeline replacement definitions */
5722 #define ARB_FFP_CONST_TFACTOR 0
5723 #define ARB_FFP_CONST_SPECULAR_ENABLE ((ARB_FFP_CONST_TFACTOR) + 1)
5724 #define ARB_FFP_CONST_CONSTANT(i) ((ARB_FFP_CONST_SPECULAR_ENABLE) + 1 + i)
5725 #define ARB_FFP_CONST_BUMPMAT(i) ((ARB_FFP_CONST_CONSTANT(7)) + 1 + i)
5726 #define ARB_FFP_CONST_LUMINANCE(i) ((ARB_FFP_CONST_BUMPMAT(7)) + 1 + i)
5728 struct arbfp_ffp_desc
5730 struct ffp_frag_desc parent
;
5734 /* Context activation is done by the caller. */
5735 static void arbfp_enable(const struct wined3d_gl_info
*gl_info
, BOOL enable
)
5739 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
5740 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
5744 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
5745 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
5749 static void *arbfp_alloc(const struct wined3d_shader_backend_ops
*shader_backend
, void *shader_priv
)
5751 struct shader_arb_priv
*priv
;
5753 /* Share private data between the shader backend and the pipeline
5754 * replacement, if both are the arb implementation. This is needed to
5755 * figure out whether ARBfp should be disabled if no pixel shader is bound
5757 if (shader_backend
== &arb_program_shader_backend
)
5759 else if (!(priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*priv
))))
5762 if (wine_rb_init(&priv
->fragment_shaders
, &wined3d_ffp_frag_program_rb_functions
) == -1)
5764 ERR("Failed to initialize rbtree.\n");
5765 if (priv
!= shader_priv
)
5766 HeapFree(GetProcessHeap(), 0, priv
);
5769 priv
->use_arbfp_fixed_func
= TRUE
;
5774 /* Context activation is done by the caller. */
5775 static void arbfp_free_ffpshader(struct wine_rb_entry
*entry
, void *context
)
5777 const struct wined3d_gl_info
*gl_info
= context
;
5778 struct arbfp_ffp_desc
*entry_arb
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_ffp_desc
, parent
.entry
);
5780 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb
->shader
));
5781 checkGLcall("glDeleteProgramsARB(1, &entry_arb->shader)");
5782 HeapFree(GetProcessHeap(), 0, entry_arb
);
5785 /* Context activation is done by the caller. */
5786 static void arbfp_free(struct wined3d_device
*device
)
5788 struct shader_arb_priv
*priv
= device
->fragment_priv
;
5790 wine_rb_destroy(&priv
->fragment_shaders
, arbfp_free_ffpshader
, &device
->adapter
->gl_info
);
5791 priv
->use_arbfp_fixed_func
= FALSE
;
5793 if (device
->shader_backend
!= &arb_program_shader_backend
)
5795 HeapFree(GetProcessHeap(), 0, device
->fragment_priv
);
5799 static void arbfp_get_caps(const struct wined3d_gl_info
*gl_info
, struct fragment_caps
*caps
)
5801 caps
->wined3d_caps
= WINED3D_FRAGMENT_CAP_PROJ_CONTROL
5802 | WINED3D_FRAGMENT_CAP_SRGB_WRITE
;
5803 caps
->PrimitiveMiscCaps
= WINED3DPMISCCAPS_TSSARGTEMP
;
5804 caps
->TextureOpCaps
= WINED3DTEXOPCAPS_DISABLE
|
5805 WINED3DTEXOPCAPS_SELECTARG1
|
5806 WINED3DTEXOPCAPS_SELECTARG2
|
5807 WINED3DTEXOPCAPS_MODULATE4X
|
5808 WINED3DTEXOPCAPS_MODULATE2X
|
5809 WINED3DTEXOPCAPS_MODULATE
|
5810 WINED3DTEXOPCAPS_ADDSIGNED2X
|
5811 WINED3DTEXOPCAPS_ADDSIGNED
|
5812 WINED3DTEXOPCAPS_ADD
|
5813 WINED3DTEXOPCAPS_SUBTRACT
|
5814 WINED3DTEXOPCAPS_ADDSMOOTH
|
5815 WINED3DTEXOPCAPS_BLENDCURRENTALPHA
|
5816 WINED3DTEXOPCAPS_BLENDFACTORALPHA
|
5817 WINED3DTEXOPCAPS_BLENDTEXTUREALPHA
|
5818 WINED3DTEXOPCAPS_BLENDDIFFUSEALPHA
|
5819 WINED3DTEXOPCAPS_BLENDTEXTUREALPHAPM
|
5820 WINED3DTEXOPCAPS_MODULATEALPHA_ADDCOLOR
|
5821 WINED3DTEXOPCAPS_MODULATECOLOR_ADDALPHA
|
5822 WINED3DTEXOPCAPS_MODULATEINVCOLOR_ADDALPHA
|
5823 WINED3DTEXOPCAPS_MODULATEINVALPHA_ADDCOLOR
|
5824 WINED3DTEXOPCAPS_DOTPRODUCT3
|
5825 WINED3DTEXOPCAPS_MULTIPLYADD
|
5826 WINED3DTEXOPCAPS_LERP
|
5827 WINED3DTEXOPCAPS_BUMPENVMAP
|
5828 WINED3DTEXOPCAPS_BUMPENVMAPLUMINANCE
;
5830 /* TODO: Implement WINED3DTEXOPCAPS_PREMODULATE */
5832 caps
->MaxTextureBlendStages
= 8;
5833 caps
->MaxSimultaneousTextures
= min(gl_info
->limits
.fragment_samplers
, 8);
5836 static void state_texfactor_arbfp(struct wined3d_context
*context
,
5837 const struct wined3d_state
*state
, DWORD state_id
)
5839 struct wined3d_device
*device
= context
->swapchain
->device
;
5840 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5843 /* Don't load the parameter if we're using an arbfp pixel shader,
5844 * otherwise we'll overwrite application provided constants. */
5845 if (device
->shader_backend
== &arb_program_shader_backend
)
5847 struct shader_arb_priv
*priv
;
5849 if (use_ps(state
)) return;
5851 priv
= device
->shader_priv
;
5852 priv
->pshader_const_dirty
[ARB_FFP_CONST_TFACTOR
] = 1;
5853 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_TFACTOR
+ 1);
5856 D3DCOLORTOGLFLOAT4(state
->render_states
[WINED3D_RS_TEXTUREFACTOR
], col
);
5857 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_TFACTOR
, col
));
5858 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_TFACTOR, col)");
5861 static void state_arb_specularenable(struct wined3d_context
*context
,
5862 const struct wined3d_state
*state
, DWORD state_id
)
5864 struct wined3d_device
*device
= context
->swapchain
->device
;
5865 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5868 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
5869 * application provided constants
5871 if (device
->shader_backend
== &arb_program_shader_backend
)
5873 struct shader_arb_priv
*priv
;
5875 if (use_ps(state
)) return;
5877 priv
= device
->shader_priv
;
5878 priv
->pshader_const_dirty
[ARB_FFP_CONST_SPECULAR_ENABLE
] = 1;
5879 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_SPECULAR_ENABLE
+ 1);
5882 if (state
->render_states
[WINED3D_RS_SPECULARENABLE
])
5884 /* The specular color has no alpha */
5885 col
[0] = 1.0f
; col
[1] = 1.0f
;
5886 col
[2] = 1.0f
; col
[3] = 0.0f
;
5888 col
[0] = 0.0f
; col
[1] = 0.0f
;
5889 col
[2] = 0.0f
; col
[3] = 0.0f
;
5891 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_SPECULAR_ENABLE
, col
));
5892 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_SPECULAR_ENABLE, col)");
5895 static void set_bumpmat_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
5897 DWORD stage
= (state_id
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
5898 struct wined3d_device
*device
= context
->swapchain
->device
;
5899 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5902 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_BUMP_ENV
;
5904 if (device
->shader_backend
== &arb_program_shader_backend
)
5906 struct shader_arb_priv
*priv
= device
->shader_priv
;
5908 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants. */
5912 priv
->pshader_const_dirty
[ARB_FFP_CONST_BUMPMAT(stage
)] = 1;
5913 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_BUMPMAT(stage
) + 1);
5916 mat
[0][0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT00
]);
5917 mat
[0][1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT01
]);
5918 mat
[1][0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT10
]);
5919 mat
[1][1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT11
]);
5921 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_BUMPMAT(stage
), &mat
[0][0]));
5922 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_BUMPMAT(stage), &mat[0][0])");
5925 static void tex_bumpenvlum_arbfp(struct wined3d_context
*context
,
5926 const struct wined3d_state
*state
, DWORD state_id
)
5928 DWORD stage
= (state_id
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
5929 struct wined3d_device
*device
= context
->swapchain
->device
;
5930 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5933 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_BUMP_ENV
;
5935 if (device
->shader_backend
== &arb_program_shader_backend
)
5937 struct shader_arb_priv
*priv
= device
->shader_priv
;
5939 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants. */
5943 priv
->pshader_const_dirty
[ARB_FFP_CONST_LUMINANCE(stage
)] = 1;
5944 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_LUMINANCE(stage
) + 1);
5947 param
[0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_LSCALE
]);
5948 param
[1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_LOFFSET
]);
5952 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_LUMINANCE(stage
), param
));
5953 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_LUMINANCE(stage), param)");
5956 static const char *get_argreg(struct wined3d_shader_buffer
*buffer
, DWORD argnum
, unsigned int stage
, DWORD arg
)
5960 if(arg
== ARG_UNUSED
) return "unused"; /* This is the marker for unused registers */
5962 switch(arg
& WINED3DTA_SELECTMASK
) {
5963 case WINED3DTA_DIFFUSE
:
5964 ret
= "fragment.color.primary"; break;
5966 case WINED3DTA_CURRENT
:
5967 if (!stage
) ret
= "fragment.color.primary";
5971 case WINED3DTA_TEXTURE
:
5973 case 0: ret
= "tex0"; break;
5974 case 1: ret
= "tex1"; break;
5975 case 2: ret
= "tex2"; break;
5976 case 3: ret
= "tex3"; break;
5977 case 4: ret
= "tex4"; break;
5978 case 5: ret
= "tex5"; break;
5979 case 6: ret
= "tex6"; break;
5980 case 7: ret
= "tex7"; break;
5981 default: ret
= "unknown texture";
5985 case WINED3DTA_TFACTOR
:
5986 ret
= "tfactor"; break;
5988 case WINED3DTA_SPECULAR
:
5989 ret
= "fragment.color.secondary"; break;
5991 case WINED3DTA_TEMP
:
5992 ret
= "tempreg"; break;
5994 case WINED3DTA_CONSTANT
:
5995 FIXME("Implement perstage constants\n");
5997 case 0: ret
= "const0"; break;
5998 case 1: ret
= "const1"; break;
5999 case 2: ret
= "const2"; break;
6000 case 3: ret
= "const3"; break;
6001 case 4: ret
= "const4"; break;
6002 case 5: ret
= "const5"; break;
6003 case 6: ret
= "const6"; break;
6004 case 7: ret
= "const7"; break;
6005 default: ret
= "unknown constant";
6013 if(arg
& WINED3DTA_COMPLEMENT
) {
6014 shader_addline(buffer
, "SUB arg%u, const.x, %s;\n", argnum
, ret
);
6015 if(argnum
== 0) ret
= "arg0";
6016 if(argnum
== 1) ret
= "arg1";
6017 if(argnum
== 2) ret
= "arg2";
6019 if(arg
& WINED3DTA_ALPHAREPLICATE
) {
6020 shader_addline(buffer
, "MOV arg%u, %s.w;\n", argnum
, ret
);
6021 if(argnum
== 0) ret
= "arg0";
6022 if(argnum
== 1) ret
= "arg1";
6023 if(argnum
== 2) ret
= "arg2";
6028 static void gen_ffp_instr(struct wined3d_shader_buffer
*buffer
, unsigned int stage
, BOOL color
,
6029 BOOL alpha
, DWORD dst
, DWORD op
, DWORD dw_arg0
, DWORD dw_arg1
, DWORD dw_arg2
)
6031 const char *dstmask
, *dstreg
, *arg0
, *arg1
, *arg2
;
6032 unsigned int mul
= 1;
6034 if(color
&& alpha
) dstmask
= "";
6035 else if(color
) dstmask
= ".xyz";
6036 else dstmask
= ".w";
6038 if(dst
== tempreg
) dstreg
= "tempreg";
6039 else dstreg
= "ret";
6041 arg0
= get_argreg(buffer
, 0, stage
, dw_arg0
);
6042 arg1
= get_argreg(buffer
, 1, stage
, dw_arg1
);
6043 arg2
= get_argreg(buffer
, 2, stage
, dw_arg2
);
6047 case WINED3D_TOP_DISABLE
:
6049 shader_addline(buffer
, "MOV %s%s, fragment.color.primary;\n", dstreg
, dstmask
);
6052 case WINED3D_TOP_SELECT_ARG2
:
6055 case WINED3D_TOP_SELECT_ARG1
:
6056 shader_addline(buffer
, "MOV %s%s, %s;\n", dstreg
, dstmask
, arg1
);
6059 case WINED3D_TOP_MODULATE_4X
:
6062 case WINED3D_TOP_MODULATE_2X
:
6065 case WINED3D_TOP_MODULATE
:
6066 shader_addline(buffer
, "MUL %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6069 case WINED3D_TOP_ADD_SIGNED_2X
:
6072 case WINED3D_TOP_ADD_SIGNED
:
6073 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
6076 case WINED3D_TOP_ADD
:
6077 shader_addline(buffer
, "ADD_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6080 case WINED3D_TOP_SUBTRACT
:
6081 shader_addline(buffer
, "SUB_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6084 case WINED3D_TOP_ADD_SMOOTH
:
6085 shader_addline(buffer
, "SUB arg1, const.x, %s;\n", arg1
);
6086 shader_addline(buffer
, "MAD_SAT %s%s, arg1, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6089 case WINED3D_TOP_BLEND_CURRENT_ALPHA
:
6090 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_CURRENT
);
6091 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6093 case WINED3D_TOP_BLEND_FACTOR_ALPHA
:
6094 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TFACTOR
);
6095 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6097 case WINED3D_TOP_BLEND_TEXTURE_ALPHA
:
6098 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
6099 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6101 case WINED3D_TOP_BLEND_DIFFUSE_ALPHA
:
6102 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_DIFFUSE
);
6103 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6106 case WINED3D_TOP_BLEND_TEXTURE_ALPHA_PM
:
6107 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
6108 shader_addline(buffer
, "SUB arg0.w, const.x, %s.w;\n", arg0
);
6109 shader_addline(buffer
, "MAD_SAT %s%s, %s, arg0.w, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6112 /* D3DTOP_PREMODULATE ???? */
6114 case WINED3D_TOP_MODULATE_INVALPHA_ADD_COLOR
:
6115 shader_addline(buffer
, "SUB arg0.w, const.x, %s;\n", arg1
);
6116 shader_addline(buffer
, "MAD_SAT %s%s, arg0.w, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6118 case WINED3D_TOP_MODULATE_ALPHA_ADD_COLOR
:
6119 shader_addline(buffer
, "MAD_SAT %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
6121 case WINED3D_TOP_MODULATE_INVCOLOR_ADD_ALPHA
:
6122 shader_addline(buffer
, "SUB arg0, const.x, %s;\n", arg1
);
6123 shader_addline(buffer
, "MAD_SAT %s%s, arg0, %s, %s.w;\n", dstreg
, dstmask
, arg2
, arg1
);
6125 case WINED3D_TOP_MODULATE_COLOR_ADD_ALPHA
:
6126 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s.w;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
6129 case WINED3D_TOP_DOTPRODUCT3
:
6131 shader_addline(buffer
, "SUB arg1, %s, const.w;\n", arg1
);
6132 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
6133 shader_addline(buffer
, "DP3_SAT %s%s, arg1, arg2;\n", dstreg
, dstmask
);
6136 case WINED3D_TOP_MULTIPLY_ADD
:
6137 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg0
);
6140 case WINED3D_TOP_LERP
:
6141 /* The msdn is not quite right here */
6142 shader_addline(buffer
, "LRP %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6145 case WINED3D_TOP_BUMPENVMAP
:
6146 case WINED3D_TOP_BUMPENVMAP_LUMINANCE
:
6147 /* Those are handled in the first pass of the shader(generation pass 1 and 2) already */
6151 FIXME("Unhandled texture op %08x\n", op
);
6155 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.y;\n", dstreg
, dstmask
, dstreg
);
6157 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.z;\n", dstreg
, dstmask
, dstreg
);
6160 static GLuint
gen_arbfp_ffp_shader(const struct ffp_frag_settings
*settings
, const struct wined3d_gl_info
*gl_info
)
6163 struct wined3d_shader_buffer buffer
;
6164 BOOL tex_read
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
6165 BOOL bump_used
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
6166 BOOL luminance_used
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
6167 UINT lowest_disabled_stage
;
6168 const char *textype
;
6170 char colorcor_dst
[8];
6172 DWORD arg0
, arg1
, arg2
;
6173 BOOL tempreg_used
= FALSE
, tfactor_used
= FALSE
;
6175 const char *final_combiner_src
= "ret";
6177 BOOL custom_linear_fog
= FALSE
;
6179 /* Find out which textures are read */
6180 for (stage
= 0; stage
< MAX_TEXTURES
; ++stage
)
6182 if (settings
->op
[stage
].cop
== WINED3D_TOP_DISABLE
)
6184 arg0
= settings
->op
[stage
].carg0
& WINED3DTA_SELECTMASK
;
6185 arg1
= settings
->op
[stage
].carg1
& WINED3DTA_SELECTMASK
;
6186 arg2
= settings
->op
[stage
].carg2
& WINED3DTA_SELECTMASK
;
6187 if(arg0
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6188 if(arg1
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6189 if(arg2
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6191 if (settings
->op
[stage
].cop
== WINED3D_TOP_BLEND_TEXTURE_ALPHA
)
6192 tex_read
[stage
] = TRUE
;
6193 if (settings
->op
[stage
].cop
== WINED3D_TOP_BLEND_TEXTURE_ALPHA_PM
)
6194 tex_read
[stage
] = TRUE
;
6195 if (settings
->op
[stage
].cop
== WINED3D_TOP_BUMPENVMAP
)
6197 bump_used
[stage
] = TRUE
;
6198 tex_read
[stage
] = TRUE
;
6200 if (settings
->op
[stage
].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
)
6202 bump_used
[stage
] = TRUE
;
6203 tex_read
[stage
] = TRUE
;
6204 luminance_used
[stage
] = TRUE
;
6206 else if (settings
->op
[stage
].cop
== WINED3D_TOP_BLEND_FACTOR_ALPHA
)
6208 tfactor_used
= TRUE
;
6211 if(arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
) {
6212 tfactor_used
= TRUE
;
6215 if(settings
->op
[stage
].dst
== tempreg
) tempreg_used
= TRUE
;
6216 if(arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
) {
6217 tempreg_used
= TRUE
;
6220 if (settings
->op
[stage
].aop
== WINED3D_TOP_DISABLE
)
6222 arg0
= settings
->op
[stage
].aarg0
& WINED3DTA_SELECTMASK
;
6223 arg1
= settings
->op
[stage
].aarg1
& WINED3DTA_SELECTMASK
;
6224 arg2
= settings
->op
[stage
].aarg2
& WINED3DTA_SELECTMASK
;
6225 if(arg0
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6226 if(arg1
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6227 if(arg2
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6229 if(arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
) {
6230 tempreg_used
= TRUE
;
6232 if(arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
) {
6233 tfactor_used
= TRUE
;
6236 lowest_disabled_stage
= stage
;
6239 if (!shader_buffer_init(&buffer
))
6241 ERR("Failed to initialize shader buffer.\n");
6245 shader_addline(&buffer
, "!!ARBfp1.0\n");
6247 switch (settings
->fog
)
6249 case WINED3D_FFP_PS_FOG_OFF
: break;
6250 case WINED3D_FFP_PS_FOG_LINEAR
:
6251 if (gl_info
->quirks
& WINED3D_QUIRK_BROKEN_ARB_FOG
)
6253 custom_linear_fog
= TRUE
;
6256 shader_addline(&buffer
, "OPTION ARB_fog_linear;\n");
6259 case WINED3D_FFP_PS_FOG_EXP
: shader_addline(&buffer
, "OPTION ARB_fog_exp;\n"); break;
6260 case WINED3D_FFP_PS_FOG_EXP2
: shader_addline(&buffer
, "OPTION ARB_fog_exp2;\n"); break;
6261 default: FIXME("Unexpected fog setting %d\n", settings
->fog
);
6264 shader_addline(&buffer
, "PARAM const = {1, 2, 4, 0.5};\n");
6265 shader_addline(&buffer
, "TEMP TMP;\n");
6266 shader_addline(&buffer
, "TEMP ret;\n");
6267 if(tempreg_used
|| settings
->sRGB_write
) shader_addline(&buffer
, "TEMP tempreg;\n");
6268 shader_addline(&buffer
, "TEMP arg0;\n");
6269 shader_addline(&buffer
, "TEMP arg1;\n");
6270 shader_addline(&buffer
, "TEMP arg2;\n");
6271 for(stage
= 0; stage
< MAX_TEXTURES
; stage
++) {
6272 if(!tex_read
[stage
]) continue;
6273 shader_addline(&buffer
, "TEMP tex%u;\n", stage
);
6274 if(!bump_used
[stage
]) continue;
6275 shader_addline(&buffer
, "PARAM bumpmat%u = program.env[%u];\n", stage
, ARB_FFP_CONST_BUMPMAT(stage
));
6276 if(!luminance_used
[stage
]) continue;
6277 shader_addline(&buffer
, "PARAM luminance%u = program.env[%u];\n", stage
, ARB_FFP_CONST_LUMINANCE(stage
));
6280 shader_addline(&buffer
, "PARAM tfactor = program.env[%u];\n", ARB_FFP_CONST_TFACTOR
);
6282 shader_addline(&buffer
, "PARAM specular_enable = program.env[%u];\n", ARB_FFP_CONST_SPECULAR_ENABLE
);
6284 if (settings
->sRGB_write
)
6286 shader_addline(&buffer
, "PARAM srgb_consts0 = ");
6287 shader_arb_append_imm_vec4(&buffer
, wined3d_srgb_const0
);
6288 shader_addline(&buffer
, ";\n");
6289 shader_addline(&buffer
, "PARAM srgb_consts1 = ");
6290 shader_arb_append_imm_vec4(&buffer
, wined3d_srgb_const1
);
6291 shader_addline(&buffer
, ";\n");
6294 if (lowest_disabled_stage
< 7 && settings
->emul_clipplanes
)
6295 shader_addline(&buffer
, "KIL fragment.texcoord[7];\n");
6297 /* Generate texture sampling instructions) */
6298 for (stage
= 0; stage
< MAX_TEXTURES
&& settings
->op
[stage
].cop
!= WINED3D_TOP_DISABLE
; ++stage
)
6300 if (!tex_read
[stage
])
6303 switch(settings
->op
[stage
].tex_type
) {
6304 case tex_1d
: textype
= "1D"; break;
6305 case tex_2d
: textype
= "2D"; break;
6306 case tex_3d
: textype
= "3D"; break;
6307 case tex_cube
: textype
= "CUBE"; break;
6308 case tex_rect
: textype
= "RECT"; break;
6309 default: textype
= "unexpected_textype"; break;
6312 if(settings
->op
[stage
].projected
== proj_none
) {
6314 } else if(settings
->op
[stage
].projected
== proj_count4
||
6315 settings
->op
[stage
].projected
== proj_count3
) {
6318 FIXME("Unexpected projection mode %d\n", settings
->op
[stage
].projected
);
6323 && (settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP
6324 || settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
))
6326 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, x, z, 0, 0;\n", stage
- 1);
6327 shader_addline(&buffer
, "DP3 ret.x, arg1, tex%u;\n", stage
- 1);
6328 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, y, w, 0, 0;\n", stage
- 1);
6329 shader_addline(&buffer
, "DP3 ret.y, arg1, tex%u;\n", stage
- 1);
6331 /* with projective textures, texbem only divides the static texture coord, not the displacement,
6332 * so multiply the displacement with the dividing parameter before passing it to TXP
6334 if (settings
->op
[stage
].projected
!= proj_none
) {
6335 if(settings
->op
[stage
].projected
== proj_count4
) {
6336 shader_addline(&buffer
, "MOV ret.w, fragment.texcoord[%u].w;\n", stage
);
6337 shader_addline(&buffer
, "MUL ret.xyz, ret, fragment.texcoord[%u].w, fragment.texcoord[%u];\n", stage
, stage
);
6339 shader_addline(&buffer
, "MOV ret.w, fragment.texcoord[%u].z;\n", stage
);
6340 shader_addline(&buffer
, "MAD ret.xyz, ret, fragment.texcoord[%u].z, fragment.texcoord[%u];\n", stage
, stage
);
6343 shader_addline(&buffer
, "ADD ret, ret, fragment.texcoord[%u];\n", stage
);
6346 shader_addline(&buffer
, "%s tex%u, ret, texture[%u], %s;\n",
6347 instr
, stage
, stage
, textype
);
6348 if (settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
)
6350 shader_addline(&buffer
, "MAD_SAT ret.x, tex%u.z, luminance%u.x, luminance%u.y;\n",
6351 stage
- 1, stage
- 1, stage
- 1);
6352 shader_addline(&buffer
, "MUL tex%u, tex%u, ret.x;\n", stage
, stage
);
6354 } else if(settings
->op
[stage
].projected
== proj_count3
) {
6355 shader_addline(&buffer
, "MOV ret, fragment.texcoord[%u];\n", stage
);
6356 shader_addline(&buffer
, "MOV ret.w, ret.z;\n");
6357 shader_addline(&buffer
, "%s tex%u, ret, texture[%u], %s;\n",
6358 instr
, stage
, stage
, textype
);
6360 shader_addline(&buffer
, "%s tex%u, fragment.texcoord[%u], texture[%u], %s;\n",
6361 instr
, stage
, stage
, stage
, textype
);
6364 sprintf(colorcor_dst
, "tex%u", stage
);
6365 gen_color_correction(&buffer
, colorcor_dst
, WINED3DSP_WRITEMASK_ALL
, "const.x", "const.y",
6366 settings
->op
[stage
].color_fixup
);
6369 /* Generate the main shader */
6370 for (stage
= 0; stage
< MAX_TEXTURES
; ++stage
)
6372 if (settings
->op
[stage
].cop
== WINED3D_TOP_DISABLE
)
6375 final_combiner_src
= "fragment.color.primary";
6379 if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG1
6380 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG1
)
6381 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
;
6382 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG1
6383 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG2
)
6384 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg2
;
6385 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG2
6386 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG1
)
6387 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg1
;
6388 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG2
6389 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG2
)
6390 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
6392 op_equal
= settings
->op
[stage
].aop
== settings
->op
[stage
].cop
6393 && settings
->op
[stage
].carg0
== settings
->op
[stage
].aarg0
6394 && settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
6395 && settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
6397 if (settings
->op
[stage
].aop
== WINED3D_TOP_DISABLE
)
6399 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].dst
,
6400 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6401 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6403 shader_addline(&buffer
, "MOV ret.w, fragment.color.primary.w;\n");
6407 gen_ffp_instr(&buffer
, stage
, TRUE
, TRUE
, settings
->op
[stage
].dst
,
6408 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6409 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6411 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].dst
,
6412 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6413 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6414 gen_ffp_instr(&buffer
, stage
, FALSE
, TRUE
, settings
->op
[stage
].dst
,
6415 settings
->op
[stage
].aop
, settings
->op
[stage
].aarg0
,
6416 settings
->op
[stage
].aarg1
, settings
->op
[stage
].aarg2
);
6420 if (settings
->sRGB_write
|| custom_linear_fog
)
6422 shader_addline(&buffer
, "MAD ret, fragment.color.secondary, specular_enable, %s;\n", final_combiner_src
);
6423 if (settings
->sRGB_write
)
6424 arbfp_add_sRGB_correction(&buffer
, "ret", "arg0", "arg1", "arg2", "tempreg", FALSE
);
6425 if (custom_linear_fog
)
6426 arbfp_add_linear_fog(&buffer
, "ret", "arg0");
6427 shader_addline(&buffer
, "MOV result.color, ret;\n");
6431 shader_addline(&buffer
, "MAD result.color, fragment.color.secondary, specular_enable, %s;\n",
6432 final_combiner_src
);
6436 shader_addline(&buffer
, "END\n");
6438 /* Generate the shader */
6439 GL_EXTCALL(glGenProgramsARB(1, &ret
));
6440 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, ret
));
6441 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
6442 strlen(buffer
.buffer
), buffer
.buffer
));
6443 checkGLcall("glProgramStringARB()");
6445 gl_info
->gl_ops
.gl
.p_glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
6448 FIXME("Fragment program error at position %d: %s\n\n", pos
,
6449 debugstr_a((const char *)gl_info
->gl_ops
.gl
.p_glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
6450 shader_arb_dump_program_source(buffer
.buffer
);
6456 GL_EXTCALL(glGetProgramivARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
6457 checkGLcall("glGetProgramivARB()");
6458 if (!native
) WARN("Program exceeds native resource limits.\n");
6461 shader_buffer_free(&buffer
);
6465 static void fragment_prog_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6467 const struct wined3d_device
*device
= context
->swapchain
->device
;
6468 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
6469 struct shader_arb_priv
*priv
= device
->fragment_priv
;
6470 BOOL use_pshader
= use_ps(state
);
6471 struct ffp_frag_settings settings
;
6472 const struct arbfp_ffp_desc
*desc
;
6475 TRACE("context %p, state %p, state_id %#x.\n", context
, state
, state_id
);
6477 if (isStateDirty(context
, STATE_RENDER(WINED3D_RS_FOGENABLE
)))
6479 if (!use_pshader
&& device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
)
6481 /* Reload fixed function constants since they collide with the
6482 * pixel shader constants. */
6483 for (i
= 0; i
< MAX_TEXTURES
; ++i
)
6485 set_bumpmat_arbfp(context
, state
, STATE_TEXTURESTAGE(i
, WINED3D_TSS_BUMPENV_MAT00
));
6487 state_texfactor_arbfp(context
, state
, STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
));
6488 state_arb_specularenable(context
, state
, STATE_RENDER(WINED3D_RS_SPECULARENABLE
));
6490 else if (use_pshader
)
6492 context
->shader_update_mask
|= 1 << WINED3D_SHADER_TYPE_PIXEL
;
6499 /* Find or create a shader implementing the fixed function pipeline
6500 * settings, then activate it. */
6501 gen_ffp_frag_op(context
, state
, &settings
, FALSE
);
6502 desc
= (const struct arbfp_ffp_desc
*)find_ffp_frag_shader(&priv
->fragment_shaders
, &settings
);
6504 struct arbfp_ffp_desc
*new_desc
= HeapAlloc(GetProcessHeap(), 0, sizeof(*new_desc
));
6507 ERR("Out of memory\n");
6511 new_desc
->parent
.settings
= settings
;
6512 new_desc
->shader
= gen_arbfp_ffp_shader(&settings
, gl_info
);
6513 add_ffp_frag_shader(&priv
->fragment_shaders
, &new_desc
->parent
);
6514 TRACE("Allocated fixed function replacement shader descriptor %p\n", new_desc
);
6518 /* Now activate the replacement program. GL_FRAGMENT_PROGRAM_ARB is already active (however, note the
6519 * comment above the shader_select call below). If e.g. GLSL is active, the shader_select call will
6522 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, desc
->shader
));
6523 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, desc->shader)");
6524 priv
->current_fprogram_id
= desc
->shader
;
6526 if (device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
)
6528 /* Reload fixed function constants since they collide with the
6529 * pixel shader constants. */
6530 for (i
= 0; i
< MAX_TEXTURES
; ++i
)
6532 set_bumpmat_arbfp(context
, state
, STATE_TEXTURESTAGE(i
, WINED3D_TSS_BUMPENV_MAT00
));
6534 state_texfactor_arbfp(context
, state
, STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
));
6535 state_arb_specularenable(context
, state
, STATE_RENDER(WINED3D_RS_SPECULARENABLE
));
6537 context
->last_was_pshader
= FALSE
;
6539 context
->last_was_pshader
= TRUE
;
6542 context
->shader_update_mask
|= 1 << WINED3D_SHADER_TYPE_PIXEL
;
6545 /* We can't link the fog states to the fragment state directly since the
6546 * vertex pipeline links them to FOGENABLE. A different linking in different
6547 * pipeline parts can't be expressed in the combined state table, so we need
6548 * to handle that with a forwarding function. The other invisible side effect
6549 * is that changing the fog start and fog end (which links to FOGENABLE in
6550 * vertex) results in the fragment_prog_arbfp function being called because
6551 * FOGENABLE is dirty, which calls this function here. */
6552 static void state_arbfp_fog(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6554 enum fogsource new_source
;
6555 DWORD fogstart
= state
->render_states
[WINED3D_RS_FOGSTART
];
6556 DWORD fogend
= state
->render_states
[WINED3D_RS_FOGEND
];
6558 TRACE("context %p, state %p, state_id %#x.\n", context
, state
, state_id
);
6560 if (!isStateDirty(context
, STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
)))
6561 fragment_prog_arbfp(context
, state
, state_id
);
6563 if (!state
->render_states
[WINED3D_RS_FOGENABLE
])
6566 if (state
->render_states
[WINED3D_RS_FOGTABLEMODE
] == WINED3D_FOG_NONE
)
6570 new_source
= FOGSOURCE_VS
;
6574 if (state
->render_states
[WINED3D_RS_FOGVERTEXMODE
] == WINED3D_FOG_NONE
|| context
->last_was_rhw
)
6575 new_source
= FOGSOURCE_COORD
;
6577 new_source
= FOGSOURCE_FFP
;
6582 new_source
= FOGSOURCE_FFP
;
6585 if (new_source
!= context
->fog_source
|| fogstart
== fogend
)
6587 context
->fog_source
= new_source
;
6588 state_fogstartend(context
, state
, STATE_RENDER(WINED3D_RS_FOGSTART
));
6592 static void textransform(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6594 if (!isStateDirty(context
, STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
)))
6595 fragment_prog_arbfp(context
, state
, state_id
);
6598 static const struct StateEntryTemplate arbfp_fragmentstate_template
[] =
6600 {STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
), { STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
), state_texfactor_arbfp
}, WINED3D_GL_EXT_NONE
},
6601 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6602 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6603 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6604 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6605 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6606 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6607 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6608 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6609 {STATE_TEXTURESTAGE(0, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6610 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6611 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6612 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6613 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6614 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6615 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6616 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6617 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6618 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6619 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6620 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6621 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6622 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6623 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6624 {STATE_TEXTURESTAGE(1, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6625 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6626 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6627 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6628 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6629 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6630 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6631 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6632 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6633 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6634 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6635 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6636 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6637 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6638 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6639 {STATE_TEXTURESTAGE(2, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6640 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6641 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6642 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6643 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6644 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6645 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6646 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6647 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6648 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6649 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6650 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6651 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6652 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6653 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6654 {STATE_TEXTURESTAGE(3, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6655 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6656 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6657 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6658 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6659 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6660 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6661 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6662 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6663 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6664 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6665 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6666 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6667 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6668 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6669 {STATE_TEXTURESTAGE(4, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6670 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6671 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6672 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6673 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6674 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6675 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6676 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6677 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6678 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6679 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6680 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6681 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6682 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6683 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6684 {STATE_TEXTURESTAGE(5, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6685 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6686 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6687 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6688 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6689 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6690 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6691 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6692 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6693 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6694 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6695 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6696 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6697 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6698 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6699 {STATE_TEXTURESTAGE(6, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6700 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6701 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6702 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6703 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6704 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6705 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6706 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6707 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6708 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6709 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6710 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6711 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6712 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6713 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6714 {STATE_TEXTURESTAGE(7, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6715 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6716 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6717 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6718 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6719 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6720 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6721 {STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
6722 {STATE_RENDER(WINED3D_RS_FOGENABLE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), state_arbfp_fog
}, WINED3D_GL_EXT_NONE
},
6723 {STATE_RENDER(WINED3D_RS_FOGTABLEMODE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6724 {STATE_RENDER(WINED3D_RS_FOGVERTEXMODE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6725 {STATE_RENDER(WINED3D_RS_FOGSTART
), { STATE_RENDER(WINED3D_RS_FOGSTART
), state_fogstartend
}, WINED3D_GL_EXT_NONE
},
6726 {STATE_RENDER(WINED3D_RS_FOGEND
), { STATE_RENDER(WINED3D_RS_FOGSTART
), NULL
}, WINED3D_GL_EXT_NONE
},
6727 {STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), { STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), state_srgbwrite
}, ARB_FRAMEBUFFER_SRGB
},
6728 {STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6729 {STATE_RENDER(WINED3D_RS_FOGCOLOR
), { STATE_RENDER(WINED3D_RS_FOGCOLOR
), state_fogcolor
}, WINED3D_GL_EXT_NONE
},
6730 {STATE_RENDER(WINED3D_RS_FOGDENSITY
), { STATE_RENDER(WINED3D_RS_FOGDENSITY
), state_fogdensity
}, WINED3D_GL_EXT_NONE
},
6731 {STATE_TEXTURESTAGE(0,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(0, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6732 {STATE_TEXTURESTAGE(1,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(1, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6733 {STATE_TEXTURESTAGE(2,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(2, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6734 {STATE_TEXTURESTAGE(3,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(3, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6735 {STATE_TEXTURESTAGE(4,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(4, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6736 {STATE_TEXTURESTAGE(5,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(5, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6737 {STATE_TEXTURESTAGE(6,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(6, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6738 {STATE_TEXTURESTAGE(7,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(7, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6739 {STATE_RENDER(WINED3D_RS_SPECULARENABLE
), { STATE_RENDER(WINED3D_RS_SPECULARENABLE
), state_arb_specularenable
}, WINED3D_GL_EXT_NONE
},
6740 {0 /* Terminate */, { 0, 0 }, WINED3D_GL_EXT_NONE
},
6743 const struct fragment_pipeline arbfp_fragment_pipeline
= {
6748 shader_arb_color_fixup_supported
,
6749 arbfp_fragmentstate_template
,
6752 struct arbfp_blit_type
6754 enum complex_fixup fixup
;
6758 struct arbfp_blit_desc
6761 struct arbfp_blit_type type
;
6762 struct wine_rb_entry entry
;
6765 struct arbfp_blit_priv
6767 struct wine_rb_tree shaders
;
6768 GLuint palette_texture
;
6771 static int arbfp_blit_type_compare(const void *key
, const struct wine_rb_entry
*entry
)
6773 const struct arbfp_blit_type
*ka
= key
;
6774 const struct arbfp_blit_type
*kb
= &WINE_RB_ENTRY_VALUE(entry
, const struct arbfp_blit_desc
, entry
)->type
;
6776 if (ka
->fixup
!= kb
->fixup
)
6777 return ka
->fixup
< kb
->fixup
? -1 : 1;
6778 if (ka
->textype
!= kb
->textype
)
6779 return ka
->textype
< kb
->textype
? -1 : 1;
6783 /* Context activation is done by the caller. */
6784 static void arbfp_free_blit_shader(struct wine_rb_entry
*entry
, void *context
)
6786 const struct wined3d_gl_info
*gl_info
= context
;
6787 struct arbfp_blit_desc
*entry_arb
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_blit_desc
, entry
);
6789 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb
->shader
));
6790 checkGLcall("glDeleteProgramsARB(1, &entry_arb->shader)");
6791 HeapFree(GetProcessHeap(), 0, entry_arb
);
6794 static const struct wine_rb_functions wined3d_arbfp_blit_rb_functions
=
6799 arbfp_blit_type_compare
,
6802 static HRESULT
arbfp_blit_alloc(struct wined3d_device
*device
)
6804 struct arbfp_blit_priv
*priv
;
6806 if (!(priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*priv
))))
6807 return E_OUTOFMEMORY
;
6809 if (wine_rb_init(&priv
->shaders
, &wined3d_arbfp_blit_rb_functions
) == -1)
6811 ERR("Failed to initialize rbtree.\n");
6812 HeapFree(GetProcessHeap(), 0, priv
);
6813 return E_OUTOFMEMORY
;
6816 device
->blit_priv
= priv
;
6821 /* Context activation is done by the caller. */
6822 static void arbfp_blit_free(struct wined3d_device
*device
)
6824 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
6825 struct arbfp_blit_priv
*priv
= device
->blit_priv
;
6827 wine_rb_destroy(&priv
->shaders
, arbfp_free_blit_shader
, &device
->adapter
->gl_info
);
6828 checkGLcall("Delete blit programs");
6830 if (priv
->palette_texture
)
6831 gl_info
->gl_ops
.gl
.p_glDeleteTextures(1, &priv
->palette_texture
);
6833 HeapFree(GetProcessHeap(), 0, device
->blit_priv
);
6834 device
->blit_priv
= NULL
;
6837 static BOOL
gen_planar_yuv_read(struct wined3d_shader_buffer
*buffer
, enum complex_fixup fixup
,
6838 GLenum textype
, char *luminance
)
6841 const char *tex
, *texinstr
;
6843 if (fixup
== COMPLEX_FIXUP_UYVY
) {
6851 case GL_TEXTURE_2D
: tex
= "2D"; texinstr
= "TXP"; break;
6852 case GL_TEXTURE_RECTANGLE_ARB
: tex
= "RECT"; texinstr
= "TEX"; break;
6854 /* This is more tricky than just replacing the texture type - we have to navigate
6855 * properly in the texture to find the correct chroma values
6857 FIXME("Implement yuv correction for non-2d, non-rect textures\n");
6861 /* First we have to read the chroma values. This means we need at least two pixels(no filtering),
6862 * or 4 pixels(with filtering). To get the unmodified chromas, we have to rid ourselves of the
6863 * filtering when we sample the texture.
6865 * These are the rules for reading the chroma:
6871 * So we have to get the sampling x position in non-normalized coordinates in integers
6873 if(textype
!= GL_TEXTURE_RECTANGLE_ARB
) {
6874 shader_addline(buffer
, "MUL texcrd.xy, fragment.texcoord[0], size.x;\n");
6875 shader_addline(buffer
, "MOV texcrd.w, size.x;\n");
6877 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
6879 /* We must not allow filtering between pixel x and x+1, this would mix U and V
6880 * Vertical filtering is ok. However, bear in mind that the pixel center is at
6883 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
6884 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, coef.y;\n");
6886 /* Divide the x coordinate by 0.5 and get the fraction. This gives 0.25 and 0.75 for the
6887 * even and odd pixels respectively
6889 shader_addline(buffer
, "MUL texcrd2, texcrd, coef.y;\n");
6890 shader_addline(buffer
, "FRC texcrd2, texcrd2;\n");
6892 /* Sample Pixel 1 */
6893 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
6895 /* Put the value into either of the chroma values */
6896 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
6897 shader_addline(buffer
, "MUL chroma.x, luminance.%c, temp.x;\n", chroma
);
6898 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
6899 shader_addline(buffer
, "MUL chroma.y, luminance.%c, temp.x;\n", chroma
);
6901 /* Sample pixel 2. If we read an even pixel(SLT above returned 1), sample
6902 * the pixel right to the current one. Otherwise, sample the left pixel.
6903 * Bias and scale the SLT result to -1;1 and add it to the texcrd.x.
6905 shader_addline(buffer
, "MAD temp.x, temp.x, coef.z, -coef.x;\n");
6906 shader_addline(buffer
, "ADD texcrd.x, texcrd, temp.x;\n");
6907 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
6909 /* Put the value into the other chroma */
6910 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
6911 shader_addline(buffer
, "MAD chroma.y, luminance.%c, temp.x, chroma.y;\n", chroma
);
6912 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
6913 shader_addline(buffer
, "MAD chroma.x, luminance.%c, temp.x, chroma.x;\n", chroma
);
6915 /* TODO: If filtering is enabled, sample a 2nd pair of pixels left or right of
6916 * the current one and lerp the two U and V values
6919 /* This gives the correctly filtered luminance value */
6920 shader_addline(buffer
, "TEX luminance, fragment.texcoord[0], texture[0], %s;\n", tex
);
6925 static BOOL
gen_yv12_read(struct wined3d_shader_buffer
*buffer
, GLenum textype
, char *luminance
)
6928 static const float yv12_coef
[]
6929 = {2.0f
/ 3.0f
, 1.0f
/ 6.0f
, (2.0f
/ 3.0f
) + (1.0f
/ 6.0f
), 1.0f
/ 3.0f
};
6932 case GL_TEXTURE_2D
: tex
= "2D"; break;
6933 case GL_TEXTURE_RECTANGLE_ARB
: tex
= "RECT"; break;
6935 FIXME("Implement yv12 correction for non-2d, non-rect textures\n");
6939 /* YV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
6940 * V and a (W/2)x(H/2) U plane, each with 8 bit per pixel. So the effective
6941 * bitdepth is 12 bits per pixel. Since the U and V planes have only half the
6942 * pitch of the luminance plane, the packing into the gl texture is a bit
6943 * unfortunate. If the whole texture is interpreted as luminance data it looks
6944 * approximately like this:
6946 * +----------------------------------+----
6958 * +----------------+-----------------+----
6960 * | V even rows | V odd rows |
6962 * +----------------+------------------ -
6964 * | U even rows | U odd rows |
6966 * +----------------+-----------------+----
6970 * So it appears as if there are 4 chroma images, but in fact the odd rows
6971 * in the chroma images are in the same row as the even ones. So its is
6972 * kinda tricky to read
6974 * When reading from rectangle textures, keep in mind that the input y coordinates
6975 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height
6977 shader_addline(buffer
, "PARAM yv12_coef = ");
6978 shader_arb_append_imm_vec4(buffer
, yv12_coef
);
6979 shader_addline(buffer
, ";\n");
6981 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
6982 /* the chroma planes have only half the width */
6983 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, coef.y;\n");
6985 /* The first value is between 2/3 and 5/6th of the texture's height, so scale+bias
6986 * the coordinate. Also read the right side of the image when reading odd lines
6988 * Don't forget to clamp the y values in into the range, otherwise we'll get filtering
6991 if(textype
== GL_TEXTURE_2D
) {
6993 shader_addline(buffer
, "RCP chroma.w, size.y;\n");
6995 shader_addline(buffer
, "MUL texcrd2.y, texcrd.y, size.y;\n");
6997 shader_addline(buffer
, "FLR texcrd2.y, texcrd2.y;\n");
6998 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, yv12_coef.y, yv12_coef.x;\n");
7000 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
7001 shader_addline(buffer
, "ADD texcrd2.x, texcrd2.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
7002 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
7003 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
7004 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
7006 /* clamp, keep the half pixel origin in mind */
7007 shader_addline(buffer
, "MAD temp.y, coef.y, chroma.w, yv12_coef.x;\n");
7008 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7009 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.z;\n");
7010 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7012 /* Read from [size - size+size/4] */
7013 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
7014 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, coef.w, size.y;\n");
7016 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
7017 shader_addline(buffer
, "ADD texcrd2.x, texcrd.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
7018 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
7019 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
7020 shader_addline(buffer
, "MUL texcrd2.x, texcrd2.x, size.x;\n");
7021 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
7023 /* Make sure to read exactly from the pixel center */
7024 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
7025 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, coef.y;\n");
7028 shader_addline(buffer
, "MAD temp.y, size.y, coef.w, size.y;\n");
7029 shader_addline(buffer
, "ADD temp.y, temp.y, -coef.y;\n");
7030 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7031 shader_addline(buffer
, "ADD temp.y, size.y, coef.y;\n");
7032 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7034 /* Read the texture, put the result into the output register */
7035 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7036 shader_addline(buffer
, "MOV chroma.x, temp.w;\n");
7038 /* The other chroma value is 1/6th of the texture lower, from 5/6th to 6/6th
7039 * No need to clamp because we're just reusing the already clamped value from above
7041 if(textype
== GL_TEXTURE_2D
) {
7042 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, yv12_coef.y;\n");
7044 shader_addline(buffer
, "MAD texcrd.y, size.y, coef.w, texcrd.y;\n");
7046 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7047 shader_addline(buffer
, "MOV chroma.y, temp.w;\n");
7049 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
7050 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
7051 * values due to filtering
7053 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7054 if(textype
== GL_TEXTURE_2D
) {
7055 /* Multiply the y coordinate by 2/3 and clamp it */
7056 shader_addline(buffer
, "MUL texcrd.y, texcrd.y, yv12_coef.x;\n");
7057 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.x;\n");
7058 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7059 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7061 /* Reading from texture_rectangles is pretty straightforward, just use the unmodified
7062 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
7065 shader_addline(buffer
, "ADD temp.x, size.y, -coef.y;\n");
7066 shader_addline(buffer
, "MIN texcrd.y, texcrd.y, size.x;\n");
7067 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7074 static BOOL
gen_nv12_read(struct wined3d_shader_buffer
*buffer
, GLenum textype
,
7078 static const float nv12_coef
[]
7079 = {2.0f
/ 3.0f
, 1.0f
/ 3.0f
, 1.0f
, 1.0f
};
7086 case GL_TEXTURE_RECTANGLE_ARB
:
7090 FIXME("Implement nv12 correction for non-2d, non-rect textures\n");
7094 /* NV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
7095 * sized plane where each component is an UV pair. So the effective
7096 * bitdepth is 12 bits per pixel If the whole texture is interpreted as luminance
7097 * data it looks approximately like this:
7099 * +----------------------------------+----
7111 * +----------------------------------+----
7112 * |UVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUV|
7113 * |UVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUV|
7119 * +----------------------------------+----
7121 * When reading from rectangle textures, keep in mind that the input y coordinates
7122 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height. */
7124 shader_addline(buffer
, "PARAM nv12_coef = ");
7125 shader_arb_append_imm_vec4(buffer
, nv12_coef
);
7126 shader_addline(buffer
, ";\n");
7128 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7129 /* We only have half the number of chroma pixels. */
7130 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, coef.y;\n");
7132 if (textype
== GL_TEXTURE_2D
)
7134 shader_addline(buffer
, "RCP chroma.w, size.x;\n");
7135 shader_addline(buffer
, "RCP chroma.z, size.y;\n");
7137 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, nv12_coef.y, nv12_coef.x;\n");
7139 /* We must not allow filtering horizontally, this would mix U and V.
7140 * Vertical filtering is ok. However, bear in mind that the pixel center is at
7141 * 0.5, so add 0.5. */
7143 /* Convert to non-normalized coordinates so we can find the
7144 * individual pixel. */
7145 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, size.x;\n");
7146 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
7147 /* Multiply by 2 since chroma components are stored in UV pixel pairs,
7148 * add 0.5 to hit the center of the pixel. */
7149 shader_addline(buffer
, "MAD texcrd.x, texcrd.x, coef.z, coef.y;\n");
7151 /* Convert back to normalized coordinates. */
7152 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, chroma.w;\n");
7154 /* Clamp, keep the half pixel origin in mind. */
7155 shader_addline(buffer
, "MAD temp.y, coef.y, chroma.z, nv12_coef.x;\n");
7156 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7157 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.z, nv12_coef.z;\n");
7158 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7162 /* Read from [size - size+size/2] */
7163 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, coef.y, size.y;\n");
7165 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
7166 /* Multiply by 2 since chroma components are stored in UV pixel pairs,
7167 * add 0.5 to hit the center of the pixel. */
7168 shader_addline(buffer
, "MAD texcrd.x, texcrd.x, coef.z, coef.y;\n");
7171 shader_addline(buffer
, "MAD temp.y, size.y, coef.y, size.y;\n");
7172 shader_addline(buffer
, "ADD temp.y, temp.y, -coef.y;\n");
7173 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7174 shader_addline(buffer
, "ADD temp.y, size.y, coef.y;\n");
7175 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7177 /* Read the texture, put the result into the output register. */
7178 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7179 shader_addline(buffer
, "MOV chroma.y, temp.w;\n");
7181 if (textype
== GL_TEXTURE_2D
)
7184 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, chroma.w;\n");
7189 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, coef.x;\n");
7191 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7192 shader_addline(buffer
, "MOV chroma.x, temp.w;\n");
7194 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
7195 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
7196 * values due to filtering. */
7197 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7198 if (textype
== GL_TEXTURE_2D
)
7200 /* Multiply the y coordinate by 2/3 and clamp it */
7201 shader_addline(buffer
, "MUL texcrd.y, texcrd.y, nv12_coef.x;\n");
7202 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, nv12_coef.x;\n");
7203 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7204 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7208 /* Reading from texture_rectangles is pretty straightforward, just use the unmodified
7209 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
7212 shader_addline(buffer
, "ADD temp.x, size.y, -coef.y;\n");
7213 shader_addline(buffer
, "MIN texcrd.y, texcrd.y, size.x;\n");
7214 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7221 /* Context activation is done by the caller. */
7222 static GLuint
gen_p8_shader(struct arbfp_blit_priv
*priv
,
7223 const struct wined3d_gl_info
*gl_info
, GLenum textype
)
7226 struct wined3d_shader_buffer buffer
;
7230 if (!shader_buffer_init(&buffer
))
7232 ERR("Failed to initialize shader buffer.\n");
7236 GL_EXTCALL(glGenProgramsARB(1, &shader
));
7237 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7240 shader_buffer_free(&buffer
);
7244 shader_addline(&buffer
, "!!ARBfp1.0\n");
7245 shader_addline(&buffer
, "TEMP index;\n");
7247 /* { 255/256, 0.5/255*255/256, 0, 0 } */
7248 shader_addline(&buffer
, "PARAM constants = { 0.996, 0.00195, 0, 0 };\n");
7250 /* The alpha-component contains the palette index */
7251 if(textype
== GL_TEXTURE_RECTANGLE_ARB
)
7252 shader_addline(&buffer
, "TXP index, fragment.texcoord[0], texture[0], RECT;\n");
7254 shader_addline(&buffer
, "TEX index, fragment.texcoord[0], texture[0], 2D;\n");
7256 /* Scale the index by 255/256 and add a bias of '0.5' in order to sample in the middle */
7257 shader_addline(&buffer
, "MAD index.a, index.a, constants.x, constants.y;\n");
7259 /* Use the alpha-component as an index in the palette to get the final color */
7260 shader_addline(&buffer
, "TEX result.color, index.a, texture[1], 1D;\n");
7261 shader_addline(&buffer
, "END\n");
7263 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
7264 strlen(buffer
.buffer
), buffer
.buffer
));
7265 checkGLcall("glProgramStringARB()");
7267 gl_info
->gl_ops
.gl
.p_glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
7270 FIXME("Fragment program error at position %d: %s\n\n", pos
,
7271 debugstr_a((const char *)gl_info
->gl_ops
.gl
.p_glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
7272 shader_arb_dump_program_source(buffer
.buffer
);
7275 shader_buffer_free(&buffer
);
7280 /* Context activation is done by the caller. */
7281 static void upload_palette(const struct wined3d_surface
*surface
, struct wined3d_context
*context
)
7283 const struct wined3d_palette
*palette
= surface
->container
->swapchain
? surface
->container
->swapchain
->palette
: NULL
;
7284 struct wined3d_device
*device
= surface
->resource
.device
;
7285 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
7286 struct arbfp_blit_priv
*priv
= device
->blit_priv
;
7288 if (!priv
->palette_texture
)
7289 gl_info
->gl_ops
.gl
.p_glGenTextures(1, &priv
->palette_texture
);
7291 GL_EXTCALL(glActiveTextureARB(GL_TEXTURE1
));
7292 gl_info
->gl_ops
.gl
.p_glBindTexture(GL_TEXTURE_1D
, priv
->palette_texture
);
7294 gl_info
->gl_ops
.gl
.p_glTexEnvi(GL_TEXTURE_ENV
, GL_TEXTURE_ENV_MODE
, GL_REPLACE
);
7296 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_MIN_FILTER
, GL_NEAREST
);
7297 /* Make sure we have discrete color levels. */
7298 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_MAG_FILTER
, GL_NEAREST
);
7299 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_WRAP_S
, GL_CLAMP_TO_EDGE
);
7300 /* TODO: avoid unneeded uploads in the future by adding some SFLAG_PALETTE_DIRTY mechanism */
7303 gl_info
->gl_ops
.gl
.p_glTexImage1D(GL_TEXTURE_1D
, 0, GL_RGB
, 256, 0, GL_BGRA
,
7304 GL_UNSIGNED_INT_8_8_8_8_REV
, palette
->colors
);
7308 static const DWORD black
;
7309 FIXME("P8 surface loaded without a palette.\n");
7310 gl_info
->gl_ops
.gl
.p_glTexImage1D(GL_TEXTURE_1D
, 0, GL_RGB
, 1, 0, GL_BGRA
,
7311 GL_UNSIGNED_INT_8_8_8_8_REV
, &black
);
7314 /* Switch back to unit 0 in which the 2D texture will be stored. */
7315 context_active_texture(context
, gl_info
, 0);
7318 /* Context activation is done by the caller. */
7319 static GLuint
gen_yuv_shader(struct arbfp_blit_priv
*priv
, const struct wined3d_gl_info
*gl_info
,
7320 enum complex_fixup yuv_fixup
, GLenum textype
)
7323 struct wined3d_shader_buffer buffer
;
7324 char luminance_component
;
7328 if (!shader_buffer_init(&buffer
))
7330 ERR("Failed to initialize shader buffer.\n");
7334 GL_EXTCALL(glGenProgramsARB(1, &shader
));
7335 checkGLcall("GL_EXTCALL(glGenProgramsARB(1, &shader))");
7336 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7337 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
7340 shader_buffer_free(&buffer
);
7344 /* The YUY2 and UYVY formats contain two pixels packed into a 32 bit macropixel,
7345 * giving effectively 16 bit per pixel. The color consists of a luminance(Y) and
7346 * two chroma(U and V) values. Each macropixel has two luminance values, one for
7347 * each single pixel it contains, and one U and one V value shared between both
7350 * The data is loaded into an A8L8 texture. With YUY2, the luminance component
7351 * contains the luminance and alpha the chroma. With UYVY it is vice versa. Thus
7352 * take the format into account when generating the read swizzles
7354 * Reading the Y value is straightforward - just sample the texture. The hardware
7355 * takes care of filtering in the horizontal and vertical direction.
7357 * Reading the U and V values is harder. We have to avoid filtering horizontally,
7358 * because that would mix the U and V values of one pixel or two adjacent pixels.
7359 * Thus floor the texture coordinate and add 0.5 to get an unfiltered read,
7360 * regardless of the filtering setting. Vertical filtering works automatically
7361 * though - the U and V values of two rows are mixed nicely.
7363 * Apart of avoiding filtering issues, the code has to know which value it just
7364 * read, and where it can find the other one. To determine this, it checks if
7365 * it sampled an even or odd pixel, and shifts the 2nd read accordingly.
7367 * Handling horizontal filtering of U and V values requires reading a 2nd pair
7368 * of pixels, extracting U and V and mixing them. This is not implemented yet.
7370 * An alternative implementation idea is to load the texture as A8R8G8B8 texture,
7371 * with width / 2. This way one read gives all 3 values, finding U and V is easy
7372 * in an unfiltered situation. Finding the luminance on the other hand requires
7373 * finding out if it is an odd or even pixel. The real drawback of this approach
7374 * is filtering. This would have to be emulated completely in the shader, reading
7375 * up two 2 packed pixels in up to 2 rows and interpolating both horizontally and
7376 * vertically. Beyond that it would require adjustments to the texture handling
7377 * code to deal with the width scaling
7379 shader_addline(&buffer
, "!!ARBfp1.0\n");
7380 shader_addline(&buffer
, "TEMP luminance;\n");
7381 shader_addline(&buffer
, "TEMP temp;\n");
7382 shader_addline(&buffer
, "TEMP chroma;\n");
7383 shader_addline(&buffer
, "TEMP texcrd;\n");
7384 shader_addline(&buffer
, "TEMP texcrd2;\n");
7385 shader_addline(&buffer
, "PARAM coef = {1.0, 0.5, 2.0, 0.25};\n");
7386 shader_addline(&buffer
, "PARAM yuv_coef = {1.403, 0.344, 0.714, 1.770};\n");
7387 shader_addline(&buffer
, "PARAM size = program.local[0];\n");
7391 case COMPLEX_FIXUP_UYVY
:
7392 case COMPLEX_FIXUP_YUY2
:
7393 if (!gen_planar_yuv_read(&buffer
, yuv_fixup
, textype
, &luminance_component
))
7395 shader_buffer_free(&buffer
);
7400 case COMPLEX_FIXUP_YV12
:
7401 if (!gen_yv12_read(&buffer
, textype
, &luminance_component
))
7403 shader_buffer_free(&buffer
);
7408 case COMPLEX_FIXUP_NV12
:
7409 if (!gen_nv12_read(&buffer
, textype
, &luminance_component
))
7411 shader_buffer_free(&buffer
);
7417 FIXME("Unsupported YUV fixup %#x\n", yuv_fixup
);
7418 shader_buffer_free(&buffer
);
7422 /* Calculate the final result. Formula is taken from
7423 * http://www.fourcc.org/fccyvrgb.php. Note that the chroma
7424 * ranges from -0.5 to 0.5
7426 shader_addline(&buffer
, "SUB chroma.xy, chroma, coef.y;\n");
7428 shader_addline(&buffer
, "MAD result.color.x, chroma.x, yuv_coef.x, luminance.%c;\n", luminance_component
);
7429 shader_addline(&buffer
, "MAD temp.x, -chroma.y, yuv_coef.y, luminance.%c;\n", luminance_component
);
7430 shader_addline(&buffer
, "MAD result.color.y, -chroma.x, yuv_coef.z, temp.x;\n");
7431 shader_addline(&buffer
, "MAD result.color.z, chroma.y, yuv_coef.w, luminance.%c;\n", luminance_component
);
7432 shader_addline(&buffer
, "END\n");
7434 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
7435 strlen(buffer
.buffer
), buffer
.buffer
));
7436 checkGLcall("glProgramStringARB()");
7438 gl_info
->gl_ops
.gl
.p_glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
7441 FIXME("Fragment program error at position %d: %s\n\n", pos
,
7442 debugstr_a((const char *)gl_info
->gl_ops
.gl
.p_glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
7443 shader_arb_dump_program_source(buffer
.buffer
);
7449 GL_EXTCALL(glGetProgramivARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
7450 checkGLcall("glGetProgramivARB()");
7451 if (!native
) WARN("Program exceeds native resource limits.\n");
7454 shader_buffer_free(&buffer
);
7459 /* Context activation is done by the caller. */
7460 static HRESULT
arbfp_blit_set(void *blit_priv
, struct wined3d_context
*context
, const struct wined3d_surface
*surface
)
7463 float size
[4] = {(float) surface
->pow2Width
, (float) surface
->pow2Height
, 1.0f
, 1.0f
};
7464 struct arbfp_blit_priv
*priv
= blit_priv
;
7465 enum complex_fixup fixup
;
7466 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
7467 GLenum textype
= surface
->container
->target
;
7468 struct wine_rb_entry
*entry
;
7469 struct arbfp_blit_type type
;
7470 struct arbfp_blit_desc
*desc
;
7472 if (surface
->flags
& SFLAG_CONVERTED
)
7474 gl_info
->gl_ops
.gl
.p_glEnable(textype
);
7475 checkGLcall("glEnable(textype)");
7479 if (!is_complex_fixup(surface
->resource
.format
->color_fixup
))
7482 dump_color_fixup_desc(surface
->resource
.format
->color_fixup
);
7483 /* Don't bother setting up a shader for unconverted formats */
7484 gl_info
->gl_ops
.gl
.p_glEnable(textype
);
7485 checkGLcall("glEnable(textype)");
7489 fixup
= get_complex_fixup(surface
->resource
.format
->color_fixup
);
7492 type
.textype
= textype
;
7493 entry
= wine_rb_get(&priv
->shaders
, &type
);
7496 desc
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_blit_desc
, entry
);
7497 shader
= desc
->shader
;
7503 case COMPLEX_FIXUP_P8
:
7504 shader
= gen_p8_shader(priv
, gl_info
, textype
);
7508 shader
= gen_yuv_shader(priv
, gl_info
, fixup
, textype
);
7514 FIXME("Unsupported complex fixup %#x, not setting a shader\n", fixup
);
7515 gl_info
->gl_ops
.gl
.p_glEnable(textype
);
7516 checkGLcall("glEnable(textype)");
7520 desc
= HeapAlloc(GetProcessHeap(), 0, sizeof(*desc
));
7524 desc
->type
.textype
= textype
;
7525 desc
->type
.fixup
= fixup
;
7526 desc
->shader
= shader
;
7527 if (wine_rb_put(&priv
->shaders
, &desc
->type
, &desc
->entry
) == -1)
7530 ERR("Out of memory\n");
7531 GL_EXTCALL(glDeleteProgramsARB(1, &shader
));
7532 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader))");
7533 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, 0));
7534 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, 0)");
7535 HeapFree(GetProcessHeap(), 0, desc
);
7536 return E_OUTOFMEMORY
;
7540 if (fixup
== COMPLEX_FIXUP_P8
)
7541 upload_palette(surface
, context
);
7543 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
7544 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
7545 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7546 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
7547 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, 0, size
));
7548 checkGLcall("glProgramLocalParameter4fvARB");
7553 /* Context activation is done by the caller. */
7554 static void arbfp_blit_unset(const struct wined3d_gl_info
*gl_info
)
7556 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
7557 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
7558 gl_info
->gl_ops
.gl
.p_glDisable(GL_TEXTURE_2D
);
7559 checkGLcall("glDisable(GL_TEXTURE_2D)");
7560 if (gl_info
->supported
[ARB_TEXTURE_CUBE_MAP
])
7562 gl_info
->gl_ops
.gl
.p_glDisable(GL_TEXTURE_CUBE_MAP_ARB
);
7563 checkGLcall("glDisable(GL_TEXTURE_CUBE_MAP_ARB)");
7565 if (gl_info
->supported
[ARB_TEXTURE_RECTANGLE
])
7567 gl_info
->gl_ops
.gl
.p_glDisable(GL_TEXTURE_RECTANGLE_ARB
);
7568 checkGLcall("glDisable(GL_TEXTURE_RECTANGLE_ARB)");
7572 static BOOL
arbfp_blit_supported(const struct wined3d_gl_info
*gl_info
, enum wined3d_blit_op blit_op
,
7573 const RECT
*src_rect
, DWORD src_usage
, enum wined3d_pool src_pool
, const struct wined3d_format
*src_format
,
7574 const RECT
*dst_rect
, DWORD dst_usage
, enum wined3d_pool dst_pool
, const struct wined3d_format
*dst_format
)
7576 enum complex_fixup src_fixup
;
7578 if (!gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
7581 if (blit_op
!= WINED3D_BLIT_OP_COLOR_BLIT
)
7583 TRACE("Unsupported blit_op=%d\n", blit_op
);
7587 if (src_pool
== WINED3D_POOL_SYSTEM_MEM
|| dst_pool
== WINED3D_POOL_SYSTEM_MEM
)
7590 src_fixup
= get_complex_fixup(src_format
->color_fixup
);
7591 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
7593 TRACE("Checking support for fixup:\n");
7594 dump_color_fixup_desc(src_format
->color_fixup
);
7597 if (!is_identity_fixup(dst_format
->color_fixup
))
7599 TRACE("Destination fixups are not supported\n");
7603 if (is_identity_fixup(src_format
->color_fixup
))
7609 /* We only support YUV conversions. */
7610 if (!is_complex_fixup(src_format
->color_fixup
))
7612 TRACE("[FAILED]\n");
7618 case COMPLEX_FIXUP_YUY2
:
7619 case COMPLEX_FIXUP_UYVY
:
7620 case COMPLEX_FIXUP_YV12
:
7621 case COMPLEX_FIXUP_NV12
:
7622 case COMPLEX_FIXUP_P8
:
7627 FIXME("Unsupported YUV fixup %#x\n", src_fixup
);
7628 TRACE("[FAILED]\n");
7633 HRESULT
arbfp_blit_surface(struct wined3d_device
*device
, DWORD filter
,
7634 struct wined3d_surface
*src_surface
, const RECT
*src_rect_in
,
7635 struct wined3d_surface
*dst_surface
, const RECT
*dst_rect_in
)
7637 struct wined3d_context
*context
;
7638 RECT src_rect
= *src_rect_in
;
7639 RECT dst_rect
= *dst_rect_in
;
7641 /* Activate the destination context, set it up for blitting */
7642 context
= context_acquire(device
, dst_surface
);
7644 /* Now load the surface */
7645 if (wined3d_settings
.offscreen_rendering_mode
!= ORM_FBO
7646 && (src_surface
->locations
& (WINED3D_LOCATION_TEXTURE_RGB
| WINED3D_LOCATION_DRAWABLE
))
7647 == WINED3D_LOCATION_DRAWABLE
7648 && !wined3d_resource_is_offscreen(&src_surface
->resource
))
7650 /* Without FBO blits transferring from the drawable to the texture is
7651 * expensive, because we have to flip the data in sysmem. Since we can
7652 * flip in the blitter, we don't actually need that flip anyway. So we
7653 * use the surface's texture as scratch texture, and flip the source
7654 * rectangle instead. */
7655 surface_load_fb_texture(src_surface
, FALSE
);
7657 src_rect
.top
= src_surface
->resource
.height
- src_rect
.top
;
7658 src_rect
.bottom
= src_surface
->resource
.height
- src_rect
.bottom
;
7661 wined3d_texture_load(src_surface
->container
, context
, FALSE
);
7663 context_apply_blit_state(context
, device
);
7665 if (!wined3d_resource_is_offscreen(&dst_surface
->resource
))
7666 surface_translate_drawable_coords(dst_surface
, context
->win_handle
, &dst_rect
);
7668 arbfp_blit_set(device
->blit_priv
, context
, src_surface
);
7670 /* Draw a textured quad */
7671 draw_textured_quad(src_surface
, context
, &src_rect
, &dst_rect
, filter
);
7673 /* Leave the opengl state valid for blitting */
7674 arbfp_blit_unset(context
->gl_info
);
7676 if (wined3d_settings
.strict_draw_ordering
7677 || (dst_surface
->container
->swapchain
7678 && (dst_surface
->container
->swapchain
->front_buffer
== dst_surface
)))
7679 context
->gl_info
->gl_ops
.gl
.p_glFlush(); /* Flush to ensure ordering across contexts. */
7681 context_release(context
);
7683 surface_validate_location(dst_surface
, dst_surface
->container
->resource
.draw_binding
);
7684 surface_invalidate_location(dst_surface
, ~dst_surface
->container
->resource
.draw_binding
);
7689 static HRESULT
arbfp_blit_color_fill(struct wined3d_device
*device
, struct wined3d_surface
*dst_surface
,
7690 const RECT
*dst_rect
, const struct wined3d_color
*color
)
7692 FIXME("Color filling not implemented by arbfp_blit\n");
7693 return WINED3DERR_INVALIDCALL
;
7696 static HRESULT
arbfp_blit_depth_fill(struct wined3d_device
*device
,
7697 struct wined3d_surface
*surface
, const RECT
*rect
, float depth
)
7699 FIXME("Depth filling not implemented by arbfp_blit.\n");
7700 return WINED3DERR_INVALIDCALL
;
7703 const struct blit_shader arbfp_blit
= {
7708 arbfp_blit_supported
,
7709 arbfp_blit_color_fill
,
7710 arbfp_blit_depth_fill
,