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 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_state
*state
)
149 return state
->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 a 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
;
275 unsigned int vs_clipplanes
;
279 /* For 3.0 vertex shaders */
280 const char *vs_output
[MAX_REG_OUTPUT
];
281 /* For 2.x and earlier vertex shaders */
282 const char *texcrd_output
[8], *color_output
[2], *fog_output
;
284 /* 3.0 pshader input for compatibility with fixed function */
285 const char *ps_input
[MAX_REG_INPUT
];
290 struct wined3d_shader_signature_element
*sig
;
292 struct wine_rb_entry entry
;
295 struct arb_pshader_private
{
296 struct arb_ps_compiled_shader
*gl_shaders
;
297 UINT num_gl_shaders
, shader_array_size
;
298 DWORD input_signature_idx
;
299 DWORD clipplane_emulation
;
303 struct arb_vshader_private
{
304 struct arb_vs_compiled_shader
*gl_shaders
;
305 UINT num_gl_shaders
, shader_array_size
;
309 struct shader_arb_priv
311 GLuint current_vprogram_id
;
312 GLuint current_fprogram_id
;
313 const struct arb_ps_compiled_shader
*compiled_fprog
;
314 const struct arb_vs_compiled_shader
*compiled_vprog
;
315 GLuint depth_blt_vprogram_id
;
316 GLuint depth_blt_fprogram_id_full
[tex_type_count
];
317 GLuint depth_blt_fprogram_id_masked
[tex_type_count
];
318 BOOL use_arbfp_fixed_func
;
319 struct wine_rb_tree fragment_shaders
;
320 BOOL last_ps_const_clamped
;
321 BOOL last_vs_color_unclamp
;
323 struct wine_rb_tree signature_tree
;
326 unsigned int highest_dirty_ps_const
, highest_dirty_vs_const
;
327 char *vshader_const_dirty
, *pshader_const_dirty
;
328 const struct wined3d_context
*last_context
;
330 const struct wined3d_vertex_pipe_ops
*vertex_pipe
;
331 const struct fragment_pipeline
*fragment_pipe
;
332 BOOL ffp_proj_control
;
335 /* Context activation for state handlers is done by the caller. */
337 static BOOL
need_rel_addr_const(const struct arb_vshader_private
*shader_data
,
338 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
)
340 if (shader_data
->rel_offset
) return TRUE
;
341 if (!reg_maps
->usesmova
) return FALSE
;
342 return !gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
];
345 /* Returns TRUE if result.clip from GL_NV_vertex_program2 should be used and FALSE otherwise */
346 static inline BOOL
use_nv_clip(const struct wined3d_gl_info
*gl_info
)
348 return gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]
349 && !(gl_info
->quirks
& WINED3D_QUIRK_NV_CLIP_BROKEN
);
352 static BOOL
need_helper_const(const struct arb_vshader_private
*shader_data
,
353 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
)
355 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
)) return TRUE
;
356 if (!gl_info
->supported
[NV_VERTEX_PROGRAM
]) return TRUE
; /* Need to init colors. */
357 if (gl_info
->quirks
& WINED3D_QUIRK_ARB_VS_OFFSET_LIMIT
) return TRUE
; /* Load the immval offset. */
358 if (gl_info
->quirks
& WINED3D_QUIRK_SET_TEXCOORD_W
) return TRUE
; /* Have to init texcoords. */
359 if (!use_nv_clip(gl_info
)) return TRUE
; /* Init the clip texcoord */
360 if (reg_maps
->usesnrm
) return TRUE
; /* 0.0 */
361 if (reg_maps
->usespow
) return TRUE
; /* EPS, 0.0 and 1.0 */
362 if (reg_maps
->fog
) return TRUE
; /* Clamping fog coord, 0.0 and 1.0 */
366 static unsigned int reserved_vs_const(const struct arb_vshader_private
*shader_data
,
367 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
)
369 unsigned int ret
= 1;
370 /* We use one PARAM for the pos fixup, and in some cases one to load
371 * some immediate values into the shader. */
372 if (need_helper_const(shader_data
, reg_maps
, gl_info
)) ++ret
;
373 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
)) ++ret
;
377 /* Loads floating point constants into the currently set ARB_vertex/fragment_program.
378 * When constant_list == NULL, it will load all the constants.
380 * @target_type should be either GL_VERTEX_PROGRAM_ARB (for vertex shaders)
381 * or GL_FRAGMENT_PROGRAM_ARB (for pixel shaders)
383 /* Context activation is done by the caller. */
384 static unsigned int shader_arb_load_constantsF(const struct wined3d_shader
*shader
,
385 const struct wined3d_gl_info
*gl_info
, GLuint target_type
, unsigned int max_constants
,
386 const float *constants
, char *dirty_consts
)
388 struct wined3d_shader_lconst
*lconst
;
392 if (TRACE_ON(d3d_constants
))
394 for(i
= 0; i
< max_constants
; i
++) {
395 if(!dirty_consts
[i
]) continue;
396 TRACE_(d3d_constants
)("Loading constants %i: %f, %f, %f, %f\n", i
,
397 constants
[i
* 4 + 0], constants
[i
* 4 + 1],
398 constants
[i
* 4 + 2], constants
[i
* 4 + 3]);
404 /* In 1.X pixel shaders constants are implicitly clamped in the range [-1;1] */
405 if (target_type
== GL_FRAGMENT_PROGRAM_ARB
&& shader
->reg_maps
.shader_version
.major
== 1)
408 /* ps 1.x supports only 8 constants, clamp only those. When switching between 1.x and higher
409 * shaders, the first 8 constants are marked dirty for reload
411 for(; i
< min(8, max_constants
); i
++) {
412 if(!dirty_consts
[i
]) continue;
416 if (constants
[j
+ 0] > 1.0f
) lcl_const
[0] = 1.0f
;
417 else if (constants
[j
+ 0] < -1.0f
) lcl_const
[0] = -1.0f
;
418 else lcl_const
[0] = constants
[j
+ 0];
420 if (constants
[j
+ 1] > 1.0f
) lcl_const
[1] = 1.0f
;
421 else if (constants
[j
+ 1] < -1.0f
) lcl_const
[1] = -1.0f
;
422 else lcl_const
[1] = constants
[j
+ 1];
424 if (constants
[j
+ 2] > 1.0f
) lcl_const
[2] = 1.0f
;
425 else if (constants
[j
+ 2] < -1.0f
) lcl_const
[2] = -1.0f
;
426 else lcl_const
[2] = constants
[j
+ 2];
428 if (constants
[j
+ 3] > 1.0f
) lcl_const
[3] = 1.0f
;
429 else if (constants
[j
+ 3] < -1.0f
) lcl_const
[3] = -1.0f
;
430 else lcl_const
[3] = constants
[j
+ 3];
432 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, i
, lcl_const
));
435 /* If further constants are dirty, reload them without clamping.
437 * The alternative is not to touch them, but then we cannot reset the dirty constant count
438 * to zero. That's bad for apps that only use PS 1.x shaders, because in that case the code
439 * above would always re-check the first 8 constants since max_constant remains at the init
444 if (gl_info
->supported
[EXT_GPU_PROGRAM_PARAMETERS
])
446 /* TODO: Benchmark if we're better of with finding the dirty constants ourselves,
447 * or just reloading *all* constants at once
449 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type, i, max_constants, constants + (i * 4)));
451 for(; i
< max_constants
; i
++) {
452 if(!dirty_consts
[i
]) continue;
454 /* Find the next block of dirty constants */
457 for(i
++; (i
< max_constants
) && dirty_consts
[i
]; i
++) {
461 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type
, j
, i
- j
, constants
+ (j
* 4)));
464 for(; i
< max_constants
; i
++) {
465 if(dirty_consts
[i
]) {
467 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, i
, constants
+ (i
* 4)));
471 checkGLcall("glProgramEnvParameter4fvARB()");
473 /* Load immediate constants */
474 if (shader
->load_local_constsF
)
476 if (TRACE_ON(d3d_shader
))
478 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
480 GLfloat
* values
= (GLfloat
*)lconst
->value
;
481 TRACE_(d3d_constants
)("Loading local constants %i: %f, %f, %f, %f\n", lconst
->idx
,
482 values
[0], values
[1], values
[2], values
[3]);
485 /* Immediate constants are clamped for 1.X shaders at loading times */
487 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
489 dirty_consts
[lconst
->idx
] = 1; /* Dirtify so the non-immediate constant overwrites it next time */
490 ret
= max(ret
, lconst
->idx
+ 1);
491 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, lconst
->idx
, (GLfloat
*)lconst
->value
));
493 checkGLcall("glProgramEnvParameter4fvARB()");
494 return ret
; /* The loaded immediate constants need reloading for the next shader */
496 return 0; /* No constants are dirty now */
501 * Loads the texture dimensions for NP2 fixup into the currently set ARB_[vertex/fragment]_programs.
503 static void shader_arb_load_np2fixup_constants(void *shader_priv
,
504 const struct wined3d_gl_info
*gl_info
, const struct wined3d_state
*state
)
506 const struct shader_arb_priv
* priv
= shader_priv
;
508 /* NP2 texcoord fixup is (currently) only done for pixelshaders. */
509 if (!use_ps(state
)) return;
511 if (priv
->compiled_fprog
&& priv
->compiled_fprog
->np2fixup_info
.super
.active
) {
512 const struct arb_ps_np2fixup_info
* const fixup
= &priv
->compiled_fprog
->np2fixup_info
;
514 WORD active
= fixup
->super
.active
;
515 GLfloat np2fixup_constants
[4 * MAX_FRAGMENT_SAMPLERS
];
517 for (i
= 0; active
; active
>>= 1, ++i
)
519 const struct wined3d_texture
*tex
= state
->textures
[i
];
520 const unsigned char idx
= fixup
->super
.idx
[i
];
521 GLfloat
*tex_dim
= &np2fixup_constants
[(idx
>> 1) * 4];
523 if (!(active
& 1)) continue;
526 FIXME("Nonexistent texture is flagged for NP2 texcoord fixup\n");
532 tex_dim
[2] = tex
->pow2_matrix
[0];
533 tex_dim
[3] = tex
->pow2_matrix
[5];
537 tex_dim
[0] = tex
->pow2_matrix
[0];
538 tex_dim
[1] = tex
->pow2_matrix
[5];
542 for (i
= 0; i
< fixup
->super
.num_consts
; ++i
) {
543 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
544 fixup
->offset
+ i
, &np2fixup_constants
[i
* 4]));
549 /* Context activation is done by the caller. */
550 static void shader_arb_ps_local_constants(const struct arb_ps_compiled_shader
*gl_shader
,
551 const struct wined3d_context
*context
, const struct wined3d_state
*state
, UINT rt_height
)
553 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
556 for(i
= 0; i
< gl_shader
->numbumpenvmatconsts
; i
++)
558 int texunit
= gl_shader
->bumpenvmatconst
[i
].texunit
;
560 /* The state manager takes care that this function is always called if the bump env matrix changes */
561 const float *data
= (const float *)&state
->texture_states
[texunit
][WINED3D_TSS_BUMPENV_MAT00
];
562 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
563 gl_shader
->bumpenvmatconst
[i
].const_num
, data
));
565 if (gl_shader
->luminanceconst
[i
].const_num
!= WINED3D_CONST_NUM_UNUSED
)
567 /* WINED3D_TSS_BUMPENVLSCALE and WINED3D_TSS_BUMPENVLOFFSET are next to each other.
568 * point gl to the scale, and load 4 floats. x = scale, y = offset, z and w are junk, we
569 * don't care about them. The pointers are valid for sure because the stateblock is bigger.
570 * (they're WINED3D_TSS_TEXTURETRANSFORMFLAGS and WINED3D_TSS_ADDRESSW, so most likely 0 or NaN
572 const float *scale
= (const float *)&state
->texture_states
[texunit
][WINED3D_TSS_BUMPENV_LSCALE
];
573 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
574 gl_shader
->luminanceconst
[i
].const_num
, scale
));
577 checkGLcall("Load bumpmap consts");
579 if(gl_shader
->ycorrection
!= WINED3D_CONST_NUM_UNUSED
)
581 /* ycorrection.x: Backbuffer height(onscreen) or 0(offscreen).
582 * ycorrection.y: -1.0(onscreen), 1.0(offscreen)
587 val
[0] = context
->render_offscreen
? 0.0f
: (float) rt_height
;
588 val
[1] = context
->render_offscreen
? 1.0f
: -1.0f
;
591 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, gl_shader
->ycorrection
, val
));
592 checkGLcall("y correction loading");
595 if (!gl_shader
->num_int_consts
) return;
597 for(i
= 0; i
< MAX_CONST_I
; i
++)
599 if(gl_shader
->int_consts
[i
] != WINED3D_CONST_NUM_UNUSED
)
602 val
[0] = (float)state
->ps_consts_i
[4 * i
];
603 val
[1] = (float)state
->ps_consts_i
[4 * i
+ 1];
604 val
[2] = (float)state
->ps_consts_i
[4 * i
+ 2];
607 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, gl_shader
->int_consts
[i
], val
));
610 checkGLcall("Load ps int consts");
613 /* Context activation is done by the caller. */
614 static void shader_arb_vs_local_constants(const struct arb_vs_compiled_shader
*gl_shader
,
615 const struct wined3d_context
*context
, const struct wined3d_state
*state
)
617 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
618 float position_fixup
[4];
621 /* Upload the position fixup */
622 shader_get_position_fixup(context
, state
, position_fixup
);
623 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, gl_shader
->pos_fixup
, position_fixup
));
625 if (!gl_shader
->num_int_consts
) return;
627 for(i
= 0; i
< MAX_CONST_I
; i
++)
629 if(gl_shader
->int_consts
[i
] != WINED3D_CONST_NUM_UNUSED
)
632 val
[0] = (float)state
->vs_consts_i
[4 * i
];
633 val
[1] = (float)state
->vs_consts_i
[4 * i
+ 1];
634 val
[2] = (float)state
->vs_consts_i
[4 * i
+ 2];
637 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, gl_shader
->int_consts
[i
], val
));
640 checkGLcall("Load vs int consts");
643 static void shader_arb_select(const struct wined3d_context
*context
, enum wined3d_shader_mode vertex_mode
,
644 enum wined3d_shader_mode fragment_mode
);
647 * Loads the app-supplied constants into the currently set ARB_[vertex/fragment]_programs.
649 * We only support float constants in ARB at the moment, so don't
650 * worry about the Integers or Booleans
652 /* Context activation is done by the caller (state handler). */
653 static void shader_arb_load_constants_internal(struct shader_arb_priv
*priv
,
654 const struct wined3d_context
*context
, const struct wined3d_state
*state
,
655 BOOL usePixelShader
, BOOL useVertexShader
, BOOL from_shader_select
)
657 const struct wined3d_d3d_info
*d3d_info
= context
->d3d_info
;
658 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
660 if (!from_shader_select
)
662 const struct wined3d_shader
*vshader
= state
->vertex_shader
, *pshader
= state
->pixel_shader
;
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(context
,
670 useVertexShader
? WINED3D_SHADER_MODE_SHADER
: WINED3D_SHADER_MODE_FFP
,
671 usePixelShader
? WINED3D_SHADER_MODE_SHADER
: WINED3D_SHADER_MODE_FFP
);
674 && (pshader
->reg_maps
.boolean_constants
675 || (!gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
]
676 && (pshader
->reg_maps
.integer_constants
& ~pshader
->reg_maps
.local_int_consts
))))
678 TRACE("bool/integer pixel shader constants potentially modified, forcing shader reselection.\n");
679 shader_arb_select(context
,
680 useVertexShader
? WINED3D_SHADER_MODE_SHADER
: WINED3D_SHADER_MODE_FFP
,
681 usePixelShader
? WINED3D_SHADER_MODE_SHADER
: WINED3D_SHADER_MODE_FFP
);
685 if (context
!= priv
->last_context
)
687 memset(priv
->vshader_const_dirty
, 1,
688 sizeof(*priv
->vshader_const_dirty
) * d3d_info
->limits
.vs_uniform_count
);
689 priv
->highest_dirty_vs_const
= d3d_info
->limits
.vs_uniform_count
;
691 memset(priv
->pshader_const_dirty
, 1,
692 sizeof(*priv
->pshader_const_dirty
) * d3d_info
->limits
.ps_uniform_count
);
693 priv
->highest_dirty_ps_const
= d3d_info
->limits
.ps_uniform_count
;
695 priv
->last_context
= context
;
700 struct wined3d_shader
*vshader
= state
->vertex_shader
;
701 const struct arb_vs_compiled_shader
*gl_shader
= priv
->compiled_vprog
;
703 /* Load DirectX 9 float constants for vertex shader */
704 priv
->highest_dirty_vs_const
= shader_arb_load_constantsF(vshader
, gl_info
, GL_VERTEX_PROGRAM_ARB
,
705 priv
->highest_dirty_vs_const
, state
->vs_consts_f
, priv
->vshader_const_dirty
);
706 shader_arb_vs_local_constants(gl_shader
, context
, state
);
711 struct wined3d_shader
*pshader
= state
->pixel_shader
;
712 const struct arb_ps_compiled_shader
*gl_shader
= priv
->compiled_fprog
;
713 UINT rt_height
= state
->fb
->render_targets
[0]->resource
.height
;
715 /* Load DirectX 9 float constants for pixel shader */
716 priv
->highest_dirty_ps_const
= shader_arb_load_constantsF(pshader
, gl_info
, GL_FRAGMENT_PROGRAM_ARB
,
717 priv
->highest_dirty_ps_const
, state
->ps_consts_f
, priv
->pshader_const_dirty
);
718 shader_arb_ps_local_constants(gl_shader
, context
, state
, rt_height
);
722 static void shader_arb_load_constants(void *shader_priv
, const struct wined3d_context
*context
,
723 const struct wined3d_state
*state
)
725 BOOL vs
= use_vs(state
);
726 BOOL ps
= use_ps(state
);
728 shader_arb_load_constants_internal(shader_priv
, context
, state
, ps
, vs
, FALSE
);
731 static void shader_arb_update_float_vertex_constants(struct wined3d_device
*device
, UINT start
, UINT count
)
733 struct wined3d_context
*context
= context_get_current();
734 struct shader_arb_priv
*priv
= device
->shader_priv
;
736 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
737 * context. On a context switch the old context will be fully dirtified */
738 if (!context
|| context
->swapchain
->device
!= device
) return;
740 memset(priv
->vshader_const_dirty
+ start
, 1, sizeof(*priv
->vshader_const_dirty
) * count
);
741 priv
->highest_dirty_vs_const
= max(priv
->highest_dirty_vs_const
, start
+ count
);
744 static void shader_arb_update_float_pixel_constants(struct wined3d_device
*device
, UINT start
, UINT count
)
746 struct wined3d_context
*context
= context_get_current();
747 struct shader_arb_priv
*priv
= device
->shader_priv
;
749 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
750 * context. On a context switch the old context will be fully dirtified */
751 if (!context
|| context
->swapchain
->device
!= device
) return;
753 memset(priv
->pshader_const_dirty
+ start
, 1, sizeof(*priv
->pshader_const_dirty
) * count
);
754 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, start
+ count
);
757 /* Generate the variable & register declarations for the ARB_vertex_program output target */
758 static void shader_generate_arb_declarations(const struct wined3d_shader
*shader
,
759 const struct wined3d_shader_reg_maps
*reg_maps
, struct wined3d_shader_buffer
*buffer
,
760 const struct wined3d_gl_info
*gl_info
, DWORD
*num_clipplanes
,
761 const struct shader_arb_ctx_priv
*ctx
)
764 char pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
765 const struct wined3d_shader_lconst
*lconst
;
766 unsigned max_constantsF
;
769 /* In pixel shaders, all private constants are program local, we don't need anything
770 * from program.env. Thus we can advertise the full set of constants in pixel shaders.
771 * If we need a private constant the GL implementation will squeeze it in somewhere
773 * With vertex shaders we need the posFixup and on some GL implementations 4 helper
774 * immediate values. The posFixup is loaded using program.env for now, so always
775 * subtract one from the number of constants. If the shader uses indirect addressing,
776 * account for the helper const too because we have to declare all available d3d constants
777 * and don't know which are actually used.
781 max_constantsF
= gl_info
->limits
.arb_ps_native_constants
;
782 /* 24 is the minimum MAX_PROGRAM_ENV_PARAMETERS_ARB value. */
783 if (max_constantsF
< 24)
784 max_constantsF
= gl_info
->limits
.arb_ps_float_constants
;
788 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
789 max_constantsF
= gl_info
->limits
.arb_vs_native_constants
;
790 /* 96 is the minimum MAX_PROGRAM_ENV_PARAMETERS_ARB value.
791 * Also prevents max_constantsF from becoming less than 0 and
793 if (max_constantsF
< 96)
794 max_constantsF
= gl_info
->limits
.arb_vs_float_constants
;
796 if (reg_maps
->usesrelconstF
)
798 DWORD highest_constf
= 0, clip_limit
;
800 max_constantsF
-= reserved_vs_const(shader_data
, reg_maps
, gl_info
);
801 max_constantsF
-= count_bits(reg_maps
->integer_constants
);
802 max_constantsF
-= gl_info
->reserved_arb_constants
;
804 for (i
= 0; i
< shader
->limits
.constant_float
; ++i
)
807 DWORD shift
= i
& 0x1f;
808 if(reg_maps
->constf
[idx
] & (1 << shift
)) highest_constf
= i
;
811 if(use_nv_clip(gl_info
) && ctx
->target_version
>= NV2
)
813 if(ctx
->cur_vs_args
->super
.clip_enabled
)
814 clip_limit
= gl_info
->limits
.clipplanes
;
820 unsigned int mask
= ctx
->cur_vs_args
->clip
.boolclip
.clipplane_mask
;
821 clip_limit
= min(count_bits(mask
), 4);
823 *num_clipplanes
= min(clip_limit
, max_constantsF
- highest_constf
- 1);
824 max_constantsF
-= *num_clipplanes
;
825 if(*num_clipplanes
< clip_limit
)
827 WARN("Only %u clipplanes out of %u enabled\n", *num_clipplanes
, gl_info
->limits
.clipplanes
);
832 if (ctx
->target_version
>= NV2
) *num_clipplanes
= gl_info
->limits
.clipplanes
;
833 else *num_clipplanes
= min(gl_info
->limits
.clipplanes
, 4);
837 for (i
= 0, map
= reg_maps
->temporary
; map
; map
>>= 1, ++i
)
839 if (map
& 1) shader_addline(buffer
, "TEMP R%u;\n", i
);
842 for (i
= 0, map
= reg_maps
->address
; map
; map
>>= 1, ++i
)
844 if (map
& 1) shader_addline(buffer
, "ADDRESS A%u;\n", i
);
847 if (pshader
&& reg_maps
->shader_version
.major
== 1 && reg_maps
->shader_version
.minor
<= 3)
849 for (i
= 0, map
= reg_maps
->texcoord
; map
; map
>>= 1, ++i
)
851 if (map
& 1) shader_addline(buffer
, "TEMP T%u;\n", i
);
855 if (!shader
->load_local_constsF
)
857 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
860 value
= (const float *)lconst
->value
;
861 shader_addline(buffer
, "PARAM C%u = {%.8e, %.8e, %.8e, %.8e};\n", lconst
->idx
,
862 value
[0], value
[1], value
[2], value
[3]);
866 /* After subtracting privately used constants from the hardware limit(they are loaded as
867 * local constants), make sure the shader doesn't violate the env constant limit
871 max_constantsF
= min(max_constantsF
, gl_info
->limits
.arb_ps_float_constants
);
875 max_constantsF
= min(max_constantsF
, gl_info
->limits
.arb_vs_float_constants
);
878 /* Avoid declaring more constants than needed */
879 max_constantsF
= min(max_constantsF
, shader
->limits
.constant_float
);
881 /* we use the array-based constants array if the local constants are marked for loading,
882 * because then we use indirect addressing, or when the local constant list is empty,
883 * because then we don't know if we're using indirect addressing or not. If we're hardcoding
884 * local constants do not declare the loaded constants as an array because ARB compilers usually
885 * do not optimize unused constants away
887 if (reg_maps
->usesrelconstF
)
889 /* Need to PARAM the environment parameters (constants) so we can use relative addressing */
890 shader_addline(buffer
, "PARAM C[%d] = { program.env[0..%d] };\n",
891 max_constantsF
, max_constantsF
- 1);
893 for(i
= 0; i
< max_constantsF
; i
++) {
896 mask
= 1 << (i
& 0x1f);
897 if (!shader_constant_is_local(shader
, i
) && (reg_maps
->constf
[idx
] & mask
))
899 shader_addline(buffer
, "PARAM C%d = program.env[%d];\n",i
, i
);
905 static const char * const shift_tab
[] = {
906 "dummy", /* 0 (none) */
907 "coefmul.x", /* 1 (x2) */
908 "coefmul.y", /* 2 (x4) */
909 "coefmul.z", /* 3 (x8) */
910 "coefmul.w", /* 4 (x16) */
911 "dummy", /* 5 (x32) */
912 "dummy", /* 6 (x64) */
913 "dummy", /* 7 (x128) */
914 "dummy", /* 8 (d256) */
915 "dummy", /* 9 (d128) */
916 "dummy", /* 10 (d64) */
917 "dummy", /* 11 (d32) */
918 "coefdiv.w", /* 12 (d16) */
919 "coefdiv.z", /* 13 (d8) */
920 "coefdiv.y", /* 14 (d4) */
921 "coefdiv.x" /* 15 (d2) */
924 static void shader_arb_get_write_mask(const struct wined3d_shader_instruction
*ins
,
925 const struct wined3d_shader_dst_param
*dst
, char *write_mask
)
927 char *ptr
= write_mask
;
929 if (dst
->write_mask
!= WINED3DSP_WRITEMASK_ALL
)
932 if (dst
->write_mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
933 if (dst
->write_mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
934 if (dst
->write_mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
935 if (dst
->write_mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
941 static void shader_arb_get_swizzle(const struct wined3d_shader_src_param
*param
, BOOL fixup
, char *swizzle_str
)
943 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
944 * but addressed as "rgba". To fix this we need to swap the register's x
945 * and z components. */
946 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
947 char *ptr
= swizzle_str
;
949 /* swizzle bits fields: wwzzyyxx */
950 DWORD swizzle
= param
->swizzle
;
951 DWORD swizzle_x
= swizzle
& 0x03;
952 DWORD swizzle_y
= (swizzle
>> 2) & 0x03;
953 DWORD swizzle_z
= (swizzle
>> 4) & 0x03;
954 DWORD swizzle_w
= (swizzle
>> 6) & 0x03;
956 /* If the swizzle is the default swizzle (ie, "xyzw"), we don't need to
957 * generate a swizzle string. Unless we need to our own swizzling. */
958 if (swizzle
!= WINED3DSP_NOSWIZZLE
|| fixup
)
961 if (swizzle_x
== swizzle_y
&& swizzle_x
== swizzle_z
&& swizzle_x
== swizzle_w
) {
962 *ptr
++ = swizzle_chars
[swizzle_x
];
964 *ptr
++ = swizzle_chars
[swizzle_x
];
965 *ptr
++ = swizzle_chars
[swizzle_y
];
966 *ptr
++ = swizzle_chars
[swizzle_z
];
967 *ptr
++ = swizzle_chars
[swizzle_w
];
974 static void shader_arb_request_a0(const struct wined3d_shader_instruction
*ins
, const char *src
)
976 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
977 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
979 if (!strcmp(priv
->addr_reg
, src
)) return;
981 strcpy(priv
->addr_reg
, src
);
982 shader_addline(buffer
, "ARL A0.x, %s;\n", src
);
985 static void shader_arb_get_src_param(const struct wined3d_shader_instruction
*ins
,
986 const struct wined3d_shader_src_param
*src
, unsigned int tmpreg
, char *outregstr
);
988 static void shader_arb_get_register_name(const struct wined3d_shader_instruction
*ins
,
989 const struct wined3d_shader_register
*reg
, char *register_name
, BOOL
*is_color
)
991 /* oPos, oFog and oPts in D3D */
992 static const char * const rastout_reg_names
[] = {"TMP_OUT", "TMP_FOGCOORD", "result.pointsize"};
993 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
994 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
995 BOOL pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
996 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1002 case WINED3DSPR_TEMP
:
1003 sprintf(register_name
, "R%u", reg
->idx
[0].offset
);
1006 case WINED3DSPR_INPUT
:
1009 if (reg_maps
->shader_version
.major
< 3)
1011 if (!reg
->idx
[0].offset
)
1012 strcpy(register_name
, "fragment.color.primary");
1014 strcpy(register_name
, "fragment.color.secondary");
1018 if (reg
->idx
[0].rel_addr
)
1021 shader_arb_get_src_param(ins
, reg
->idx
[0].rel_addr
, 0, rel_reg
);
1023 if (!strcmp(rel_reg
, "**aL_emul**"))
1025 DWORD idx
= ctx
->aL
+ reg
->idx
[0].offset
;
1026 if(idx
< MAX_REG_INPUT
)
1028 strcpy(register_name
, ctx
->ps_input
[idx
]);
1032 ERR("Pixel shader input register out of bounds: %u\n", idx
);
1033 sprintf(register_name
, "out_of_bounds_%u", idx
);
1036 else if (reg_maps
->input_registers
& 0x0300)
1038 /* There are two ways basically:
1040 * 1) Use the unrolling code that is used for loop emulation and unroll the loop.
1041 * That means trouble if the loop also contains a breakc or if the control values
1042 * aren't local constants.
1043 * 2) Generate an if block that checks if aL.y < 8, == 8 or == 9 and selects the
1044 * source dynamically. The trouble is that we cannot simply read aL.y because it
1045 * is an ADDRESS register. We could however push it, load .zw with a value and use
1046 * ADAC to load the condition code register and pop it again afterwards
1048 FIXME("Relative input register addressing with more than 8 registers\n");
1050 /* This is better than nothing for now */
1051 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1053 else if(ctx
->cur_ps_args
->super
.vp_mode
!= vertexshader
)
1055 /* This is problematic because we'd have to consult the ctx->ps_input strings
1056 * for where to find the varying. Some may be "0.0", others can be texcoords or
1057 * colors. This needs either a pipeline replacement to make the vertex shader feed
1058 * proper varyings, or loop unrolling
1060 * For now use the texcoords and hope for the best
1062 FIXME("Non-vertex shader varying input with indirect addressing\n");
1063 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1067 /* D3D supports indirect addressing only with aL in loop registers. The loop instruction
1068 * pulls GL_NV_fragment_program2 in
1070 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1075 if (reg
->idx
[0].offset
< MAX_REG_INPUT
)
1077 strcpy(register_name
, ctx
->ps_input
[reg
->idx
[0].offset
]);
1081 ERR("Pixel shader input register out of bounds: %u\n", reg
->idx
[0].offset
);
1082 sprintf(register_name
, "out_of_bounds_%u", reg
->idx
[0].offset
);
1089 if (ctx
->cur_vs_args
->super
.swizzle_map
& (1 << reg
->idx
[0].offset
))
1091 sprintf(register_name
, "vertex.attrib[%u]", reg
->idx
[0].offset
);
1095 case WINED3DSPR_CONST
:
1096 if (!pshader
&& reg
->idx
[0].rel_addr
)
1098 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1099 UINT rel_offset
= shader_data
->rel_offset
;
1102 if (reg_maps
->shader_version
.major
< 2)
1104 sprintf(rel_reg
, "A0.x");
1108 shader_arb_get_src_param(ins
, reg
->idx
[0].rel_addr
, 0, rel_reg
);
1109 if (ctx
->target_version
== ARB
)
1111 if (!strcmp(rel_reg
, "**aL_emul**"))
1115 shader_arb_request_a0(ins
, rel_reg
);
1116 sprintf(rel_reg
, "A0.x");
1121 sprintf(register_name
, "C[%u]", ctx
->aL
+ reg
->idx
[0].offset
);
1122 else if (reg
->idx
[0].offset
>= rel_offset
)
1123 sprintf(register_name
, "C[%s + %u]", rel_reg
, reg
->idx
[0].offset
- rel_offset
);
1125 sprintf(register_name
, "C[%s - %u]", rel_reg
, rel_offset
- reg
->idx
[0].offset
);
1129 if (reg_maps
->usesrelconstF
)
1130 sprintf(register_name
, "C[%u]", reg
->idx
[0].offset
);
1132 sprintf(register_name
, "C%u", reg
->idx
[0].offset
);
1136 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
1139 if (reg_maps
->shader_version
.major
== 1
1140 && reg_maps
->shader_version
.minor
<= 3)
1141 /* In ps <= 1.3, Tx is a temporary register as destination
1142 * to all instructions, and as source to most instructions.
1143 * For some instructions it is the texcoord input. Those
1144 * instructions know about the special use. */
1145 sprintf(register_name
, "T%u", reg
->idx
[0].offset
);
1147 /* In ps 1.4 and 2.x Tx is always a (read-only) varying. */
1148 sprintf(register_name
, "fragment.texcoord[%u]", reg
->idx
[0].offset
);
1152 if (reg_maps
->shader_version
.major
== 1 || ctx
->target_version
>= NV2
)
1153 sprintf(register_name
, "A%u", reg
->idx
[0].offset
);
1155 sprintf(register_name
, "A%u_SHADOW", reg
->idx
[0].offset
);
1159 case WINED3DSPR_COLOROUT
:
1160 if (ctx
->cur_ps_args
->super
.srgb_correction
&& !reg
->idx
[0].offset
)
1162 strcpy(register_name
, "TMP_COLOR");
1166 if (ctx
->cur_ps_args
->super
.srgb_correction
)
1167 FIXME("sRGB correction on higher render targets.\n");
1168 if (reg_maps
->rt_mask
> 1)
1169 sprintf(register_name
, "result.color[%u]", reg
->idx
[0].offset
);
1171 strcpy(register_name
, "result.color");
1175 case WINED3DSPR_RASTOUT
:
1176 if (reg
->idx
[0].offset
== 1)
1177 sprintf(register_name
, "%s", ctx
->fog_output
);
1179 sprintf(register_name
, "%s", rastout_reg_names
[reg
->idx
[0].offset
]);
1182 case WINED3DSPR_DEPTHOUT
:
1183 strcpy(register_name
, "result.depth");
1186 case WINED3DSPR_ATTROUT
:
1187 /* case WINED3DSPR_OUTPUT: */
1189 sprintf(register_name
, "oD[%u]", reg
->idx
[0].offset
);
1191 strcpy(register_name
, ctx
->color_output
[reg
->idx
[0].offset
]);
1194 case WINED3DSPR_TEXCRDOUT
:
1196 sprintf(register_name
, "oT[%u]", reg
->idx
[0].offset
);
1197 else if (reg_maps
->shader_version
.major
< 3)
1198 strcpy(register_name
, ctx
->texcrd_output
[reg
->idx
[0].offset
]);
1200 strcpy(register_name
, ctx
->vs_output
[reg
->idx
[0].offset
]);
1203 case WINED3DSPR_LOOP
:
1204 if(ctx
->target_version
>= NV2
)
1206 /* Pshader has an implicitly declared loop index counter A0.x that cannot be renamed */
1207 if(pshader
) sprintf(register_name
, "A0.x");
1208 else sprintf(register_name
, "aL.y");
1212 /* Unfortunately this code cannot return the value of ctx->aL here. An immediate value
1213 * would be valid, but if aL is used for indexing(its only use), there's likely an offset,
1214 * thus the result would be something like C[15 + 30], which is not valid in the ARB program
1215 * grammar. So return a marker for the emulated aL and intercept it in constant and varying
1218 sprintf(register_name
, "**aL_emul**");
1223 case WINED3DSPR_CONSTINT
:
1224 sprintf(register_name
, "I%u", reg
->idx
[0].offset
);
1227 case WINED3DSPR_MISCTYPE
:
1228 if (!reg
->idx
[0].offset
)
1229 sprintf(register_name
, "vpos");
1230 else if (reg
->idx
[0].offset
== 1)
1231 sprintf(register_name
, "fragment.facing.x");
1233 FIXME("Unknown MISCTYPE register index %u.\n", reg
->idx
[0].offset
);
1237 FIXME("Unhandled register type %#x[%u].\n", reg
->type
, reg
->idx
[0].offset
);
1238 sprintf(register_name
, "unrecognized_register[%u]", reg
->idx
[0].offset
);
1243 static void shader_arb_get_dst_param(const struct wined3d_shader_instruction
*ins
,
1244 const struct wined3d_shader_dst_param
*wined3d_dst
, char *str
)
1246 char register_name
[255];
1250 shader_arb_get_register_name(ins
, &wined3d_dst
->reg
, register_name
, &is_color
);
1251 strcpy(str
, register_name
);
1253 shader_arb_get_write_mask(ins
, wined3d_dst
, write_mask
);
1254 strcat(str
, write_mask
);
1257 static const char *shader_arb_get_fixup_swizzle(enum fixup_channel_source channel_source
)
1259 switch(channel_source
)
1261 case CHANNEL_SOURCE_ZERO
: return "0";
1262 case CHANNEL_SOURCE_ONE
: return "1";
1263 case CHANNEL_SOURCE_X
: return "x";
1264 case CHANNEL_SOURCE_Y
: return "y";
1265 case CHANNEL_SOURCE_Z
: return "z";
1266 case CHANNEL_SOURCE_W
: return "w";
1268 FIXME("Unhandled channel source %#x\n", channel_source
);
1273 static void gen_color_correction(struct wined3d_shader_buffer
*buffer
, const char *reg
,
1274 DWORD dst_mask
, const char *one
, const char *two
, struct color_fixup_desc fixup
)
1278 if (is_complex_fixup(fixup
))
1280 enum complex_fixup complex_fixup
= get_complex_fixup(fixup
);
1281 FIXME("Complex fixup (%#x) not supported\n", complex_fixup
);
1286 if (fixup
.x_source
!= CHANNEL_SOURCE_X
) mask
|= WINED3DSP_WRITEMASK_0
;
1287 if (fixup
.y_source
!= CHANNEL_SOURCE_Y
) mask
|= WINED3DSP_WRITEMASK_1
;
1288 if (fixup
.z_source
!= CHANNEL_SOURCE_Z
) mask
|= WINED3DSP_WRITEMASK_2
;
1289 if (fixup
.w_source
!= CHANNEL_SOURCE_W
) mask
|= WINED3DSP_WRITEMASK_3
;
1294 shader_addline(buffer
, "SWZ %s, %s, %s, %s, %s, %s;\n", reg
, reg
,
1295 shader_arb_get_fixup_swizzle(fixup
.x_source
), shader_arb_get_fixup_swizzle(fixup
.y_source
),
1296 shader_arb_get_fixup_swizzle(fixup
.z_source
), shader_arb_get_fixup_swizzle(fixup
.w_source
));
1300 if (fixup
.x_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_0
;
1301 if (fixup
.y_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_1
;
1302 if (fixup
.z_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_2
;
1303 if (fixup
.w_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_3
;
1309 char *ptr
= reg_mask
;
1311 if (mask
!= WINED3DSP_WRITEMASK_ALL
)
1314 if (mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
1315 if (mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
1316 if (mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
1317 if (mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
1321 shader_addline(buffer
, "MAD %s%s, %s, %s, -%s;\n", reg
, reg_mask
, reg
, two
, one
);
1325 static const char *shader_arb_get_modifier(const struct wined3d_shader_instruction
*ins
)
1328 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1329 if (!ins
->dst_count
) return "";
1331 mod
= ins
->dst
[0].modifiers
;
1333 /* Silently ignore PARTIALPRECISION if its not supported */
1334 if(priv
->target_version
== ARB
) mod
&= ~WINED3DSPDM_PARTIALPRECISION
;
1336 if(mod
& WINED3DSPDM_MSAMPCENTROID
)
1338 FIXME("Unhandled modifier WINED3DSPDM_MSAMPCENTROID\n");
1339 mod
&= ~WINED3DSPDM_MSAMPCENTROID
;
1344 case WINED3DSPDM_SATURATE
| WINED3DSPDM_PARTIALPRECISION
:
1347 case WINED3DSPDM_SATURATE
:
1350 case WINED3DSPDM_PARTIALPRECISION
:
1357 FIXME("Unknown modifiers 0x%08x\n", mod
);
1362 #define TEX_PROJ 0x1
1363 #define TEX_BIAS 0x2
1365 #define TEX_DERIV 0x10
1367 static void shader_hw_sample(const struct wined3d_shader_instruction
*ins
, DWORD sampler_idx
,
1368 const char *dst_str
, const char *coord_reg
, WORD flags
, const char *dsx
, const char *dsy
)
1370 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1371 DWORD sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[sampler_idx
];
1372 const char *tex_type
;
1373 BOOL np2_fixup
= FALSE
;
1374 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1376 BOOL pshader
= shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
1377 const struct wined3d_shader
*shader
;
1378 const struct wined3d_device
*device
;
1379 const struct wined3d_gl_info
*gl_info
;
1381 /* D3D vertex shader sampler IDs are vertex samplers(0-3), not global d3d samplers */
1382 if(!pshader
) sampler_idx
+= MAX_FRAGMENT_SAMPLERS
;
1384 switch(sampler_type
) {
1390 shader
= ins
->ctx
->shader
;
1391 device
= shader
->device
;
1392 gl_info
= &device
->adapter
->gl_info
;
1394 if (pshader
&& priv
->cur_ps_args
->super
.np2_fixup
& (1 << sampler_idx
)
1395 && gl_info
->supported
[ARB_TEXTURE_RECTANGLE
])
1399 if (shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
))
1401 if (priv
->cur_np2fixup_info
->super
.active
& (1 << sampler_idx
))
1403 if (flags
) FIXME("Only ordinary sampling from NP2 textures is supported.\n");
1404 else np2_fixup
= TRUE
;
1409 case WINED3DSTT_VOLUME
:
1413 case WINED3DSTT_CUBE
:
1418 ERR("Unexpected texture type %d\n", sampler_type
);
1422 /* TEX, TXL, TXD and TXP do not support the "H" modifier,
1423 * so don't use shader_arb_get_modifier
1425 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
) mod
= "_SAT";
1428 /* Fragment samplers always have indentity mapping */
1429 if(sampler_idx
>= MAX_FRAGMENT_SAMPLERS
)
1431 sampler_idx
= priv
->cur_vs_args
->vertex
.samplers
[sampler_idx
- MAX_FRAGMENT_SAMPLERS
];
1434 if (flags
& TEX_DERIV
)
1436 if(flags
& TEX_PROJ
) FIXME("Projected texture sampling with custom derivatives\n");
1437 if(flags
& TEX_BIAS
) FIXME("Biased texture sampling with custom derivatives\n");
1438 shader_addline(buffer
, "TXD%s %s, %s, %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
,
1439 dsx
, dsy
,sampler_idx
, tex_type
);
1441 else if(flags
& TEX_LOD
)
1443 if(flags
& TEX_PROJ
) FIXME("Projected texture sampling with explicit lod\n");
1444 if(flags
& TEX_BIAS
) FIXME("Biased texture sampling with explicit lod\n");
1445 shader_addline(buffer
, "TXL%s %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
,
1446 sampler_idx
, tex_type
);
1448 else if (flags
& TEX_BIAS
)
1450 /* Shouldn't be possible, but let's check for it */
1451 if(flags
& TEX_PROJ
) FIXME("Biased and Projected texture sampling\n");
1452 /* TXB takes the 4th component of the source vector automatically, as d3d. Nothing more to do */
1453 shader_addline(buffer
, "TXB%s %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
, sampler_idx
, tex_type
);
1455 else if (flags
& TEX_PROJ
)
1457 shader_addline(buffer
, "TXP%s %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
, sampler_idx
, tex_type
);
1463 const unsigned char idx
= priv
->cur_np2fixup_info
->super
.idx
[sampler_idx
];
1464 shader_addline(buffer
, "MUL TA, np2fixup[%u].%s, %s;\n", idx
>> 1,
1465 (idx
% 2) ? "zwxy" : "xyzw", coord_reg
);
1467 shader_addline(buffer
, "TEX%s %s, TA, texture[%u], %s;\n", mod
, dst_str
, sampler_idx
, tex_type
);
1470 shader_addline(buffer
, "TEX%s %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
, sampler_idx
, tex_type
);
1475 gen_color_correction(buffer
, dst_str
, ins
->dst
[0].write_mask
,
1476 arb_get_helper_value(WINED3D_SHADER_TYPE_PIXEL
, ARB_ONE
),
1477 arb_get_helper_value(WINED3D_SHADER_TYPE_PIXEL
, ARB_TWO
),
1478 priv
->cur_ps_args
->super
.color_fixup
[sampler_idx
]);
1482 static void shader_arb_get_src_param(const struct wined3d_shader_instruction
*ins
,
1483 const struct wined3d_shader_src_param
*src
, unsigned int tmpreg
, char *outregstr
)
1485 /* Generate a line that does the input modifier computation and return the input register to use */
1486 BOOL is_color
= FALSE
;
1490 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1491 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1492 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
1493 const char *two
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_TWO
);
1495 /* Assume a new line will be added */
1498 /* Get register name */
1499 shader_arb_get_register_name(ins
, &src
->reg
, regstr
, &is_color
);
1500 shader_arb_get_swizzle(src
, is_color
, swzstr
);
1502 switch (src
->modifiers
)
1504 case WINED3DSPSM_NONE
:
1505 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
1508 case WINED3DSPSM_NEG
:
1509 sprintf(outregstr
, "-%s%s", regstr
, swzstr
);
1512 case WINED3DSPSM_BIAS
:
1513 shader_addline(buffer
, "ADD T%c, %s, -coefdiv.x;\n", 'A' + tmpreg
, regstr
);
1515 case WINED3DSPSM_BIASNEG
:
1516 shader_addline(buffer
, "ADD T%c, -%s, coefdiv.x;\n", 'A' + tmpreg
, regstr
);
1518 case WINED3DSPSM_SIGN
:
1519 shader_addline(buffer
, "MAD T%c, %s, %s, -%s;\n", 'A' + tmpreg
, regstr
, two
, one
);
1521 case WINED3DSPSM_SIGNNEG
:
1522 shader_addline(buffer
, "MAD T%c, %s, -%s, %s;\n", 'A' + tmpreg
, regstr
, two
, one
);
1524 case WINED3DSPSM_COMP
:
1525 shader_addline(buffer
, "SUB T%c, %s, %s;\n", 'A' + tmpreg
, one
, regstr
);
1527 case WINED3DSPSM_X2
:
1528 shader_addline(buffer
, "ADD T%c, %s, %s;\n", 'A' + tmpreg
, regstr
, regstr
);
1530 case WINED3DSPSM_X2NEG
:
1531 shader_addline(buffer
, "ADD T%c, -%s, -%s;\n", 'A' + tmpreg
, regstr
, regstr
);
1533 case WINED3DSPSM_DZ
:
1534 shader_addline(buffer
, "RCP T%c, %s.z;\n", 'A' + tmpreg
, regstr
);
1535 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
1537 case WINED3DSPSM_DW
:
1538 shader_addline(buffer
, "RCP T%c, %s.w;\n", 'A' + tmpreg
, regstr
);
1539 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
1541 case WINED3DSPSM_ABS
:
1542 if(ctx
->target_version
>= NV2
) {
1543 sprintf(outregstr
, "|%s%s|", regstr
, swzstr
);
1546 shader_addline(buffer
, "ABS T%c, %s;\n", 'A' + tmpreg
, regstr
);
1549 case WINED3DSPSM_ABSNEG
:
1550 if(ctx
->target_version
>= NV2
) {
1551 sprintf(outregstr
, "-|%s%s|", regstr
, swzstr
);
1553 shader_addline(buffer
, "ABS T%c, %s;\n", 'A' + tmpreg
, regstr
);
1554 sprintf(outregstr
, "-T%c%s", 'A' + tmpreg
, swzstr
);
1559 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
1563 /* Return modified or original register, with swizzle */
1565 sprintf(outregstr
, "T%c%s", 'A' + tmpreg
, swzstr
);
1568 static void pshader_hw_bem(const struct wined3d_shader_instruction
*ins
)
1570 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1571 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1572 DWORD sampler_code
= dst
->reg
.idx
[0].offset
;
1574 char src_name
[2][50];
1576 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1578 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed
1580 * Keep in mind that src_name[1] can be "TB" and src_name[0] can be "TA" because modifiers like _x2 are valid
1581 * with bem. So delay loading the first parameter until after the perturbation calculation which needs two
1584 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1585 shader_addline(buffer
, "SWZ TA, bumpenvmat%d, x, z, 0, 0;\n", sampler_code
);
1586 shader_addline(buffer
, "DP3 TC.r, TA, %s;\n", src_name
[1]);
1587 shader_addline(buffer
, "SWZ TA, bumpenvmat%d, y, w, 0, 0;\n", sampler_code
);
1588 shader_addline(buffer
, "DP3 TC.g, TA, %s;\n", src_name
[1]);
1590 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1591 shader_addline(buffer
, "ADD %s, %s, TC;\n", dst_name
, src_name
[0]);
1594 static DWORD
negate_modifiers(DWORD mod
, char *extra_char
)
1599 case WINED3DSPSM_NONE
: return WINED3DSPSM_NEG
;
1600 case WINED3DSPSM_NEG
: return WINED3DSPSM_NONE
;
1601 case WINED3DSPSM_BIAS
: return WINED3DSPSM_BIASNEG
;
1602 case WINED3DSPSM_BIASNEG
: return WINED3DSPSM_BIAS
;
1603 case WINED3DSPSM_SIGN
: return WINED3DSPSM_SIGNNEG
;
1604 case WINED3DSPSM_SIGNNEG
: return WINED3DSPSM_SIGN
;
1605 case WINED3DSPSM_COMP
: *extra_char
= '-'; return WINED3DSPSM_COMP
;
1606 case WINED3DSPSM_X2
: return WINED3DSPSM_X2NEG
;
1607 case WINED3DSPSM_X2NEG
: return WINED3DSPSM_X2
;
1608 case WINED3DSPSM_DZ
: *extra_char
= '-'; return WINED3DSPSM_DZ
;
1609 case WINED3DSPSM_DW
: *extra_char
= '-'; return WINED3DSPSM_DW
;
1610 case WINED3DSPSM_ABS
: return WINED3DSPSM_ABSNEG
;
1611 case WINED3DSPSM_ABSNEG
: return WINED3DSPSM_ABS
;
1613 FIXME("Unknown modifier %u\n", mod
);
1617 static void pshader_hw_cnd(const struct wined3d_shader_instruction
*ins
)
1619 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1620 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1622 char src_name
[3][50];
1623 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1624 ins
->ctx
->reg_maps
->shader_version
.minor
);
1626 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1627 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1629 /* The coissue flag changes the semantic of the cnd instruction in <= 1.3 shaders */
1630 if (shader_version
<= WINED3D_SHADER_VERSION(1, 3) && ins
->coissue
)
1632 shader_addline(buffer
, "MOV%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
[1]);
1636 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
1639 /* src0 may have a negate srcmod set, so we can't blindly add "-" to the name */
1640 src0_copy
.modifiers
= negate_modifiers(src0_copy
.modifiers
, &extra_neg
);
1642 shader_arb_get_src_param(ins
, &src0_copy
, 0, src_name
[0]);
1643 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1644 shader_addline(buffer
, "ADD TA, %c%s, coefdiv.x;\n", extra_neg
, src_name
[0]);
1645 shader_addline(buffer
, "CMP%s %s, TA, %s, %s;\n", shader_arb_get_modifier(ins
),
1646 dst_name
, src_name
[1], src_name
[2]);
1650 static void pshader_hw_cmp(const struct wined3d_shader_instruction
*ins
)
1652 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1653 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1655 char src_name
[3][50];
1657 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1659 /* Generate input register names (with modifiers) */
1660 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1661 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1662 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1664 shader_addline(buffer
, "CMP%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1665 dst_name
, src_name
[0], src_name
[2], src_name
[1]);
1668 /** Process the WINED3DSIO_DP2ADD instruction in ARB.
1669 * dst = dot2(src0, src1) + src2 */
1670 static void pshader_hw_dp2add(const struct wined3d_shader_instruction
*ins
)
1672 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1673 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1675 char src_name
[3][50];
1676 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1678 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1679 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1680 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1682 if(ctx
->target_version
>= NV3
)
1684 /* GL_NV_fragment_program2 has a 1:1 matching instruction */
1685 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1686 shader_addline(buffer
, "DP2A%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1687 dst_name
, src_name
[0], src_name
[1], src_name
[2]);
1689 else if(ctx
->target_version
>= NV2
)
1691 /* dst.x = src2.?, src0.x, src1.x + src0.y * src1.y
1692 * dst.y = src2.?, src0.x, src1.z + src0.y * src1.w
1693 * dst.z = src2.?, src0.x, src1.x + src0.y * src1.y
1694 * dst.z = src2.?, src0.x, src1.z + src0.y * src1.w
1696 * Make sure that src1.zw = src1.xy, then we get a classic dp2add
1698 * .xyxy and other swizzles that we could get with this are not valid in
1699 * plain ARBfp, but luckily the NV extension grammar lifts this limitation.
1701 struct wined3d_shader_src_param tmp_param
= ins
->src
[1];
1702 DWORD swizzle
= tmp_param
.swizzle
& 0xf; /* Selects .xy */
1703 tmp_param
.swizzle
= swizzle
| (swizzle
<< 4); /* Creates .xyxy */
1705 shader_arb_get_src_param(ins
, &tmp_param
, 1, src_name
[1]);
1707 shader_addline(buffer
, "X2D%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1708 dst_name
, src_name
[2], src_name
[0], src_name
[1]);
1712 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1713 /* 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]
1714 * src_name[0] can be TA, but TA is a private temp for modifiers, so it is save to overwrite
1716 shader_addline(buffer
, "MOV TA, %s;\n", src_name
[0]);
1717 shader_addline(buffer
, "MOV TA.z, 0.0;\n");
1718 shader_addline(buffer
, "DP3 TA, TA, %s;\n", src_name
[1]);
1719 shader_addline(buffer
, "ADD%s %s, TA, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
[2]);
1723 /* Map the opcode 1-to-1 to the GL code */
1724 static void shader_hw_map2gl(const struct wined3d_shader_instruction
*ins
)
1726 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1727 const char *instruction
;
1728 char arguments
[256], dst_str
[50];
1730 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1732 switch (ins
->handler_idx
)
1734 case WINED3DSIH_ABS
: instruction
= "ABS"; break;
1735 case WINED3DSIH_ADD
: instruction
= "ADD"; break;
1736 case WINED3DSIH_CRS
: instruction
= "XPD"; break;
1737 case WINED3DSIH_DP3
: instruction
= "DP3"; break;
1738 case WINED3DSIH_DP4
: instruction
= "DP4"; break;
1739 case WINED3DSIH_DST
: instruction
= "DST"; break;
1740 case WINED3DSIH_FRC
: instruction
= "FRC"; break;
1741 case WINED3DSIH_LIT
: instruction
= "LIT"; break;
1742 case WINED3DSIH_LRP
: instruction
= "LRP"; break;
1743 case WINED3DSIH_MAD
: instruction
= "MAD"; break;
1744 case WINED3DSIH_MAX
: instruction
= "MAX"; break;
1745 case WINED3DSIH_MIN
: instruction
= "MIN"; break;
1746 case WINED3DSIH_MOV
: instruction
= "MOV"; break;
1747 case WINED3DSIH_MUL
: instruction
= "MUL"; break;
1748 case WINED3DSIH_SGE
: instruction
= "SGE"; break;
1749 case WINED3DSIH_SLT
: instruction
= "SLT"; break;
1750 case WINED3DSIH_SUB
: instruction
= "SUB"; break;
1751 case WINED3DSIH_MOVA
:instruction
= "ARR"; break;
1752 case WINED3DSIH_DSX
: instruction
= "DDX"; break;
1753 default: instruction
= "";
1754 FIXME("Unhandled opcode %#x\n", ins
->handler_idx
);
1758 /* Note that shader_arb_add_dst_param() adds spaces. */
1759 arguments
[0] = '\0';
1760 shader_arb_get_dst_param(ins
, dst
, dst_str
);
1761 for (i
= 0; i
< ins
->src_count
; ++i
)
1764 strcat(arguments
, ", ");
1765 shader_arb_get_src_param(ins
, &ins
->src
[i
], i
, operand
);
1766 strcat(arguments
, operand
);
1768 shader_addline(buffer
, "%s%s %s%s;\n", instruction
, shader_arb_get_modifier(ins
), dst_str
, arguments
);
1771 static void shader_hw_nop(const struct wined3d_shader_instruction
*ins
) {}
1773 static void shader_hw_mov(const struct wined3d_shader_instruction
*ins
)
1775 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
1776 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
1777 BOOL pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
1778 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1779 const char *zero
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ZERO
);
1780 const char *one
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ONE
);
1781 const char *two
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_TWO
);
1783 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1784 char src0_param
[256];
1786 if (ins
->handler_idx
== WINED3DSIH_MOVA
)
1788 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1790 const char *offset
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_VS_REL_OFFSET
);
1792 if(ctx
->target_version
>= NV2
) {
1793 shader_hw_map2gl(ins
);
1796 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_param
);
1797 shader_arb_get_write_mask(ins
, &ins
->dst
[0], write_mask
);
1799 /* This implements the mova formula used in GLSL. The first two instructions
1800 * prepare the sign() part. Note that it is fine to have my_sign(0.0) = 1.0
1804 * A0.x = arl(floor(abs(0.0) + 0.5) * 1.0) = floor(0.5) = 0.0 since arl does a floor
1806 * The ARL is performed when A0 is used - the requested component is read from A0_SHADOW into
1807 * A0.x. We can use the overwritten component of A0_shadow as temporary storage for the sign.
1809 shader_addline(buffer
, "SGE A0_SHADOW%s, %s, %s;\n", write_mask
, src0_param
, zero
);
1810 shader_addline(buffer
, "MAD A0_SHADOW%s, A0_SHADOW, %s, -%s;\n", write_mask
, two
, one
);
1812 shader_addline(buffer
, "ABS TA%s, %s;\n", write_mask
, src0_param
);
1813 shader_addline(buffer
, "ADD TA%s, TA, rel_addr_const.x;\n", write_mask
);
1814 shader_addline(buffer
, "FLR TA%s, TA;\n", write_mask
);
1815 if (shader_data
->rel_offset
)
1817 shader_addline(buffer
, "ADD TA%s, TA, %s;\n", write_mask
, offset
);
1819 shader_addline(buffer
, "MUL A0_SHADOW%s, TA, A0_SHADOW;\n", write_mask
);
1821 ((struct shader_arb_ctx_priv
*)ins
->ctx
->backend_data
)->addr_reg
[0] = '\0';
1823 else if (reg_maps
->shader_version
.major
== 1
1824 && !shader_is_pshader_version(reg_maps
->shader_version
.type
)
1825 && ins
->dst
[0].reg
.type
== WINED3DSPR_ADDR
)
1827 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1828 src0_param
[0] = '\0';
1830 if (shader_data
->rel_offset
)
1832 const char *offset
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_VS_REL_OFFSET
);
1833 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_param
);
1834 shader_addline(buffer
, "ADD TA.x, %s, %s;\n", src0_param
, offset
);
1835 shader_addline(buffer
, "ARL A0.x, TA.x;\n");
1839 /* Apple's ARB_vertex_program implementation does not accept an ARL source argument
1840 * with more than one component. Thus replicate the first source argument over all
1841 * 4 components. For example, .xyzw -> .x (or better: .xxxx), .zwxy -> .z, etc) */
1842 struct wined3d_shader_src_param tmp_src
= ins
->src
[0];
1843 tmp_src
.swizzle
= (tmp_src
.swizzle
& 0x3) * 0x55;
1844 shader_arb_get_src_param(ins
, &tmp_src
, 0, src0_param
);
1845 shader_addline(buffer
, "ARL A0.x, %s;\n", src0_param
);
1848 else if (ins
->dst
[0].reg
.type
== WINED3DSPR_COLOROUT
&& !ins
->dst
[0].reg
.idx
[0].offset
&& pshader
)
1850 if (ctx
->cur_ps_args
->super
.srgb_correction
&& shader
->u
.ps
.color0_mov
)
1852 shader_addline(buffer
, "#mov handled in srgb write code\n");
1855 shader_hw_map2gl(ins
);
1859 shader_hw_map2gl(ins
);
1863 static void pshader_hw_texkill(const struct wined3d_shader_instruction
*ins
)
1865 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1866 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1869 /* No swizzles are allowed in d3d's texkill. PS 1.x ignores the 4th component as documented,
1870 * but >= 2.0 honors it (undocumented, but tested by the d3d9 testsuite)
1872 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1874 if (ins
->ctx
->reg_maps
->shader_version
.major
>= 2)
1876 const char *kilsrc
= "TA";
1879 shader_arb_get_register_name(ins
, &dst
->reg
, reg_dest
, &is_color
);
1880 if(dst
->write_mask
== WINED3DSP_WRITEMASK_ALL
)
1886 /* Sigh. KIL doesn't support swizzles/writemasks. KIL passes a writemask, but ".xy" for example
1887 * is not valid as a swizzle in ARB (needs ".xyyy"). Use SWZ to load the register properly, and set
1888 * masked out components to 0(won't kill)
1890 char x
= '0', y
= '0', z
= '0', w
= '0';
1891 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
) x
= 'x';
1892 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
) y
= 'y';
1893 if(dst
->write_mask
& WINED3DSP_WRITEMASK_2
) z
= 'z';
1894 if(dst
->write_mask
& WINED3DSP_WRITEMASK_3
) w
= 'w';
1895 shader_addline(buffer
, "SWZ TA, %s, %c, %c, %c, %c;\n", reg_dest
, x
, y
, z
, w
);
1897 shader_addline(buffer
, "KIL %s;\n", kilsrc
);
1901 /* ARB fp doesn't like swizzles on the parameter of the KIL instruction. To mask the 4th component,
1902 * copy the register into our general purpose TMP variable, overwrite .w and pass TMP to KIL
1904 * ps_1_3 shaders use the texcoord incarnation of the Tx register. ps_1_4 shaders can use the same,
1905 * or pass in any temporary register(in shader phase 2)
1907 if (ins
->ctx
->reg_maps
->shader_version
.minor
<= 3)
1908 sprintf(reg_dest
, "fragment.texcoord[%u]", dst
->reg
.idx
[0].offset
);
1910 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1911 shader_addline(buffer
, "SWZ TA, %s, x, y, z, 1;\n", reg_dest
);
1912 shader_addline(buffer
, "KIL TA;\n");
1916 static void pshader_hw_tex(const struct wined3d_shader_instruction
*ins
)
1918 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1919 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1920 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1921 ins
->ctx
->reg_maps
->shader_version
.minor
);
1922 struct wined3d_shader_src_param src
;
1926 DWORD reg_sampler_code
;
1928 BOOL swizzle_coord
= FALSE
;
1930 /* All versions have a destination register */
1931 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1933 /* 1.0-1.4: Use destination register number as texture code.
1934 2.0+: Use provided sampler number as texure code. */
1935 if (shader_version
< WINED3D_SHADER_VERSION(2,0))
1936 reg_sampler_code
= dst
->reg
.idx
[0].offset
;
1938 reg_sampler_code
= ins
->src
[1].reg
.idx
[0].offset
;
1940 /* 1.0-1.3: Use the texcoord varying.
1941 1.4+: Use provided coordinate source register. */
1942 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
1943 sprintf(reg_coord
, "fragment.texcoord[%u]", reg_sampler_code
);
1945 /* TEX is the only instruction that can handle DW and DZ natively */
1947 if(src
.modifiers
== WINED3DSPSM_DW
) src
.modifiers
= WINED3DSPSM_NONE
;
1948 if(src
.modifiers
== WINED3DSPSM_DZ
) src
.modifiers
= WINED3DSPSM_NONE
;
1949 shader_arb_get_src_param(ins
, &src
, 0, reg_coord
);
1953 * 1.1, 1.2, 1.3: Use WINED3D_TSS_TEXTURETRANSFORMFLAGS
1954 * 1.4: Use WINED3DSPSM_DZ or WINED3DSPSM_DW on src[0]
1955 * 2.0+: Use WINED3DSI_TEXLD_PROJECT on the opcode
1957 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
1960 if (reg_sampler_code
< MAX_TEXTURES
)
1961 flags
= priv
->cur_ps_args
->super
.tex_transform
>> reg_sampler_code
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
;
1962 if (flags
& WINED3D_PSARGS_PROJECTED
)
1964 myflags
|= TEX_PROJ
;
1965 if ((flags
& ~WINED3D_PSARGS_PROJECTED
) == WINED3D_TTFF_COUNT3
)
1966 swizzle_coord
= TRUE
;
1969 else if (shader_version
< WINED3D_SHADER_VERSION(2,0))
1971 enum wined3d_shader_src_modifier src_mod
= ins
->src
[0].modifiers
;
1972 if (src_mod
== WINED3DSPSM_DZ
)
1974 swizzle_coord
= TRUE
;
1975 myflags
|= TEX_PROJ
;
1976 } else if(src_mod
== WINED3DSPSM_DW
) {
1977 myflags
|= TEX_PROJ
;
1980 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) myflags
|= TEX_PROJ
;
1981 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) myflags
|= TEX_BIAS
;
1986 /* TXP cannot handle DZ natively, so move the z coordinate to .w.
1987 * reg_coord is a read-only varying register, so we need a temp reg */
1988 shader_addline(ins
->ctx
->buffer
, "SWZ TA, %s, x, y, z, z;\n", reg_coord
);
1989 strcpy(reg_coord
, "TA");
1992 shader_hw_sample(ins
, reg_sampler_code
, reg_dest
, reg_coord
, myflags
, NULL
, NULL
);
1995 static void pshader_hw_texcoord(const struct wined3d_shader_instruction
*ins
)
1997 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1998 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1999 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
2000 ins
->ctx
->reg_maps
->shader_version
.minor
);
2003 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
2005 DWORD reg
= dst
->reg
.idx
[0].offset
;
2007 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2008 shader_addline(buffer
, "MOV_SAT %s, fragment.texcoord[%u];\n", dst_str
, reg
);
2012 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_src
);
2013 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2014 shader_addline(buffer
, "MOV %s, %s;\n", dst_str
, reg_src
);
2018 static void pshader_hw_texreg2ar(const struct wined3d_shader_instruction
*ins
)
2020 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2023 DWORD reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2027 /* Note that texreg2ar treats Tx as a temporary register, not as a varying */
2028 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2029 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2030 /* Move .x first in case src_str is "TA" */
2031 shader_addline(buffer
, "MOV TA.y, %s.x;\n", src_str
);
2032 shader_addline(buffer
, "MOV TA.x, %s.w;\n", src_str
);
2033 if (reg1
< MAX_TEXTURES
)
2035 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2036 flags
= priv
->cur_ps_args
->super
.tex_transform
>> reg1
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
;
2038 shader_hw_sample(ins
, reg1
, dst_str
, "TA", flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2041 static void pshader_hw_texreg2gb(const struct wined3d_shader_instruction
*ins
)
2043 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2045 DWORD reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2049 /* Note that texreg2gb treats Tx as a temporary register, not as a varying */
2050 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2051 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2052 shader_addline(buffer
, "MOV TA.x, %s.y;\n", src_str
);
2053 shader_addline(buffer
, "MOV TA.y, %s.z;\n", src_str
);
2054 shader_hw_sample(ins
, reg1
, dst_str
, "TA", 0, NULL
, NULL
);
2057 static void pshader_hw_texreg2rgb(const struct wined3d_shader_instruction
*ins
)
2059 DWORD reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2063 /* Note that texreg2rg treats Tx as a temporary register, not as a varying */
2064 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2065 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2066 shader_hw_sample(ins
, reg1
, dst_str
, src_str
, 0, NULL
, NULL
);
2069 static void pshader_hw_texbem(const struct wined3d_shader_instruction
*ins
)
2071 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2072 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2073 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2074 char reg_coord
[40], dst_reg
[50], src_reg
[50];
2075 DWORD reg_dest_code
;
2077 /* All versions have a destination register. The Tx where the texture coordinates come
2078 * from is the varying incarnation of the texture register
2080 reg_dest_code
= dst
->reg
.idx
[0].offset
;
2081 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_reg
);
2082 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_reg
);
2083 sprintf(reg_coord
, "fragment.texcoord[%u]", reg_dest_code
);
2085 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed
2086 * The Tx in which the perturbation map is stored is the tempreg incarnation of the texture register
2088 * GL_NV_fragment_program_option could handle this in one instruction via X2D:
2089 * X2D TA.xy, fragment.texcoord, T%u, bumpenvmat%u.xzyw
2091 * However, the NV extensions are never enabled for <= 2.0 shaders because of the performance penalty that
2092 * comes with it, and texbem is an 1.x only instruction. No 1.x instruction forces us to enable the NV
2095 shader_addline(buffer
, "SWZ TB, bumpenvmat%d, x, z, 0, 0;\n", reg_dest_code
);
2096 shader_addline(buffer
, "DP3 TA.x, TB, %s;\n", src_reg
);
2097 shader_addline(buffer
, "SWZ TB, bumpenvmat%d, y, w, 0, 0;\n", reg_dest_code
);
2098 shader_addline(buffer
, "DP3 TA.y, TB, %s;\n", src_reg
);
2100 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2101 * so we can't let the GL handle this.
2103 if ((priv
->cur_ps_args
->super
.tex_transform
>> reg_dest_code
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
)
2104 & WINED3D_PSARGS_PROJECTED
)
2106 shader_addline(buffer
, "RCP TB.w, %s.w;\n", reg_coord
);
2107 shader_addline(buffer
, "MUL TB.xy, %s, TB.w;\n", reg_coord
);
2108 shader_addline(buffer
, "ADD TA.xy, TA, TB;\n");
2110 shader_addline(buffer
, "ADD TA.xy, TA, %s;\n", reg_coord
);
2113 shader_hw_sample(ins
, reg_dest_code
, dst_reg
, "TA", 0, NULL
, NULL
);
2115 if (ins
->handler_idx
== WINED3DSIH_TEXBEML
)
2117 /* No src swizzles are allowed, so this is ok */
2118 shader_addline(buffer
, "MAD TA, %s.z, luminance%d.x, luminance%d.y;\n",
2119 src_reg
, reg_dest_code
, reg_dest_code
);
2120 shader_addline(buffer
, "MUL %s, %s, TA;\n", dst_reg
, dst_reg
);
2124 static void pshader_hw_texm3x2pad(const struct wined3d_shader_instruction
*ins
)
2126 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2127 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2128 char src0_name
[50], dst_name
[50];
2130 struct wined3d_shader_register tmp_reg
= ins
->dst
[0].reg
;
2132 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2133 /* The next instruction will be a texm3x2tex or texm3x2depth that writes to the uninitialized
2134 * T<reg+1> register. Use this register to store the calculated vector
2136 tmp_reg
.idx
[0].offset
= reg
+ 1;
2137 shader_arb_get_register_name(ins
, &tmp_reg
, dst_name
, &is_color
);
2138 shader_addline(buffer
, "DP3 %s.x, fragment.texcoord[%u], %s;\n", dst_name
, reg
, src0_name
);
2141 static void pshader_hw_texm3x2tex(const struct wined3d_shader_instruction
*ins
)
2143 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2145 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2146 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2152 /* We know that we're writing to the uninitialized T<reg> register, so use it for temporary storage */
2153 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2155 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2156 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2157 shader_addline(buffer
, "DP3 %s.y, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2158 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2159 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2162 static void pshader_hw_texm3x3pad(const struct wined3d_shader_instruction
*ins
)
2164 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2165 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2166 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2167 char src0_name
[50], dst_name
[50];
2168 struct wined3d_shader_register tmp_reg
= ins
->dst
[0].reg
;
2171 /* There are always 2 texm3x3pad instructions followed by one texm3x3[tex,vspec, ...] instruction, with
2172 * incrementing ins->dst[0].register_idx numbers. So the pad instruction already knows the final destination
2173 * register, and this register is uninitialized(otherwise the assembler complains that it is 'redeclared')
2175 tmp_reg
.idx
[0].offset
= reg
+ 2 - tex_mx
->current_row
;
2176 shader_arb_get_register_name(ins
, &tmp_reg
, dst_name
, &is_color
);
2178 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2179 shader_addline(buffer
, "DP3 %s.%c, fragment.texcoord[%u], %s;\n",
2180 dst_name
, 'x' + tex_mx
->current_row
, reg
, src0_name
);
2181 tex_mx
->texcoord_w
[tex_mx
->current_row
++] = reg
;
2184 static void pshader_hw_texm3x3tex(const struct wined3d_shader_instruction
*ins
)
2186 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2187 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2189 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2190 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2192 char src0_name
[50], dst_name
[50];
2195 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2196 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2197 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_name
, reg
, src0_name
);
2199 /* Sample the texture using the calculated coordinates */
2200 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2201 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2202 shader_hw_sample(ins
, reg
, dst_str
, dst_name
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2203 tex_mx
->current_row
= 0;
2206 static void pshader_hw_texm3x3vspec(const struct wined3d_shader_instruction
*ins
)
2208 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2209 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2211 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2212 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2218 /* Get the dst reg without writemask strings. We know this register is uninitialized, so we can use all
2219 * components for temporary data storage
2221 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2222 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2223 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2225 /* Construct the eye-ray vector from w coordinates */
2226 shader_addline(buffer
, "MOV TB.x, fragment.texcoord[%u].w;\n", tex_mx
->texcoord_w
[0]);
2227 shader_addline(buffer
, "MOV TB.y, fragment.texcoord[%u].w;\n", tex_mx
->texcoord_w
[1]);
2228 shader_addline(buffer
, "MOV TB.z, fragment.texcoord[%u].w;\n", reg
);
2230 /* Calculate reflection vector
2232 shader_addline(buffer
, "DP3 %s.w, %s, TB;\n", dst_reg
, dst_reg
);
2233 /* The .w is ignored when sampling, so I can use TB.w to calculate dot(N, N) */
2234 shader_addline(buffer
, "DP3 TB.w, %s, %s;\n", dst_reg
, dst_reg
);
2235 shader_addline(buffer
, "RCP TB.w, TB.w;\n");
2236 shader_addline(buffer
, "MUL %s.w, %s.w, TB.w;\n", dst_reg
, dst_reg
);
2237 shader_addline(buffer
, "MUL %s, %s.w, %s;\n", dst_reg
, dst_reg
, dst_reg
);
2238 shader_addline(buffer
, "MAD %s, coefmul.x, %s, -TB;\n", dst_reg
, dst_reg
);
2240 /* Sample the texture using the calculated coordinates */
2241 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2242 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2243 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2244 tex_mx
->current_row
= 0;
2247 static void pshader_hw_texm3x3spec(const struct wined3d_shader_instruction
*ins
)
2249 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2250 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2252 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2253 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2260 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2261 shader_arb_get_src_param(ins
, &ins
->src
[0], 1, src1_name
);
2262 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2263 /* Note: dst_reg.xy is input here, generated by two texm3x3pad instructions */
2264 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2266 /* Calculate reflection vector.
2269 * dst_reg.xyz = 2 * --------- * N - E
2272 * Which normalizes the normal vector
2274 shader_addline(buffer
, "DP3 %s.w, %s, %s;\n", dst_reg
, dst_reg
, src1_name
);
2275 shader_addline(buffer
, "DP3 TC.w, %s, %s;\n", dst_reg
, dst_reg
);
2276 shader_addline(buffer
, "RCP TC.w, TC.w;\n");
2277 shader_addline(buffer
, "MUL %s.w, %s.w, TC.w;\n", dst_reg
, dst_reg
);
2278 shader_addline(buffer
, "MUL %s, %s.w, %s;\n", dst_reg
, dst_reg
, dst_reg
);
2279 shader_addline(buffer
, "MAD %s, coefmul.x, %s, -%s;\n", dst_reg
, dst_reg
, src1_name
);
2281 /* Sample the texture using the calculated coordinates */
2282 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2283 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2284 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2285 tex_mx
->current_row
= 0;
2288 static void pshader_hw_texdepth(const struct wined3d_shader_instruction
*ins
)
2290 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2291 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2293 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2294 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2296 /* texdepth has an implicit destination, the fragment depth value. It's only parameter,
2297 * which is essentially an input, is the destination register because it is the first
2298 * parameter. According to the msdn, this must be register r5, but let's keep it more flexible
2299 * here(writemasks/swizzles are not valid on texdepth)
2301 shader_arb_get_dst_param(ins
, dst
, dst_name
);
2303 /* According to the msdn, the source register(must be r5) is unusable after
2304 * the texdepth instruction, so we're free to modify it
2306 shader_addline(buffer
, "MIN %s.y, %s.y, %s;\n", dst_name
, dst_name
, one
);
2308 /* How to deal with the special case dst_name.g == 0? if r != 0, then
2309 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
2310 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
2312 shader_addline(buffer
, "RCP %s.y, %s.y;\n", dst_name
, dst_name
);
2313 shader_addline(buffer
, "MUL TA.x, %s.x, %s.y;\n", dst_name
, dst_name
);
2314 shader_addline(buffer
, "MIN TA.x, TA.x, %s;\n", one
);
2315 shader_addline(buffer
, "MAX result.depth, TA.x, %s;\n", zero
);
2318 /** Process the WINED3DSIO_TEXDP3TEX instruction in ARB:
2319 * Take a 3-component dot product of the TexCoord[dstreg] and src,
2320 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
2321 static void pshader_hw_texdp3tex(const struct wined3d_shader_instruction
*ins
)
2323 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2324 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
[0].offset
;
2328 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2329 shader_addline(buffer
, "MOV TB, 0.0;\n");
2330 shader_addline(buffer
, "DP3 TB.x, fragment.texcoord[%u], %s;\n", sampler_idx
, src0
);
2332 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2333 shader_hw_sample(ins
, sampler_idx
, dst_str
, "TB", 0 /* Only one coord, can't be projected */, NULL
, NULL
);
2336 /** Process the WINED3DSIO_TEXDP3 instruction in ARB:
2337 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
2338 static void pshader_hw_texdp3(const struct wined3d_shader_instruction
*ins
)
2340 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2343 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2345 /* Handle output register */
2346 shader_arb_get_dst_param(ins
, dst
, dst_str
);
2347 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2348 shader_addline(buffer
, "DP3 %s, fragment.texcoord[%u], %s;\n", dst_str
, dst
->reg
.idx
[0].offset
, src0
);
2351 /** Process the WINED3DSIO_TEXM3X3 instruction in ARB
2352 * Perform the 3rd row of a 3x3 matrix multiply */
2353 static void pshader_hw_texm3x3(const struct wined3d_shader_instruction
*ins
)
2355 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2356 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2357 char dst_str
[50], dst_name
[50];
2361 shader_arb_get_dst_param(ins
, dst
, dst_str
);
2362 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2363 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2364 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_name
, dst
->reg
.idx
[0].offset
, src0
);
2365 shader_addline(buffer
, "MOV %s, %s;\n", dst_str
, dst_name
);
2368 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in ARB:
2369 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
2370 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
2371 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
2373 static void pshader_hw_texm3x2depth(const struct wined3d_shader_instruction
*ins
)
2375 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2376 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2377 char src0
[50], dst_name
[50];
2379 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2380 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2382 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2383 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2384 shader_addline(buffer
, "DP3 %s.y, fragment.texcoord[%u], %s;\n", dst_name
, dst
->reg
.idx
[0].offset
, src0
);
2386 /* How to deal with the special case dst_name.g == 0? if r != 0, then
2387 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
2388 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
2390 shader_addline(buffer
, "RCP %s.y, %s.y;\n", dst_name
, dst_name
);
2391 shader_addline(buffer
, "MUL %s.x, %s.x, %s.y;\n", dst_name
, dst_name
, dst_name
);
2392 shader_addline(buffer
, "MIN %s.x, %s.x, %s;\n", dst_name
, dst_name
, one
);
2393 shader_addline(buffer
, "MAX result.depth, %s.x, %s;\n", dst_name
, zero
);
2396 /** Handles transforming all WINED3DSIO_M?x? opcodes for
2397 Vertex/Pixel shaders to ARB_vertex_program codes */
2398 static void shader_hw_mnxn(const struct wined3d_shader_instruction
*ins
)
2401 int nComponents
= 0;
2402 struct wined3d_shader_dst_param tmp_dst
= {{0}};
2403 struct wined3d_shader_src_param tmp_src
[2] = {{{0}}};
2404 struct wined3d_shader_instruction tmp_ins
;
2406 memset(&tmp_ins
, 0, sizeof(tmp_ins
));
2408 /* Set constants for the temporary argument */
2409 tmp_ins
.ctx
= ins
->ctx
;
2410 tmp_ins
.dst_count
= 1;
2411 tmp_ins
.dst
= &tmp_dst
;
2412 tmp_ins
.src_count
= 2;
2413 tmp_ins
.src
= tmp_src
;
2415 switch(ins
->handler_idx
)
2417 case WINED3DSIH_M4x4
:
2419 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2421 case WINED3DSIH_M4x3
:
2423 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2425 case WINED3DSIH_M3x4
:
2427 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2429 case WINED3DSIH_M3x3
:
2431 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2433 case WINED3DSIH_M3x2
:
2435 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2438 FIXME("Unhandled opcode %#x\n", ins
->handler_idx
);
2442 tmp_dst
= ins
->dst
[0];
2443 tmp_src
[0] = ins
->src
[0];
2444 tmp_src
[1] = ins
->src
[1];
2445 for (i
= 0; i
< nComponents
; ++i
)
2447 tmp_dst
.write_mask
= WINED3DSP_WRITEMASK_0
<< i
;
2448 shader_hw_map2gl(&tmp_ins
);
2449 ++tmp_src
[1].reg
.idx
[0].offset
;
2453 static void shader_hw_rcp(const struct wined3d_shader_instruction
*ins
)
2455 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2460 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
); /* Destination */
2461 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src
);
2462 if (ins
->src
[0].swizzle
== WINED3DSP_NOSWIZZLE
)
2464 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
2470 shader_addline(buffer
, "RCP%s %s, %s;\n", shader_arb_get_modifier(ins
), dst
, src
);
2473 static void shader_hw_scalar_op(const struct wined3d_shader_instruction
*ins
)
2475 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2476 const char *instruction
;
2481 switch(ins
->handler_idx
)
2483 case WINED3DSIH_RSQ
: instruction
= "RSQ"; break;
2484 case WINED3DSIH_RCP
: instruction
= "RCP"; break;
2485 case WINED3DSIH_EXP
: instruction
= "EX2"; break;
2486 case WINED3DSIH_EXPP
: instruction
= "EXP"; break;
2487 default: instruction
= "";
2488 FIXME("Unhandled opcode %#x\n", ins
->handler_idx
);
2492 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
); /* Destination */
2493 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src
);
2494 if (ins
->src
[0].swizzle
== WINED3DSP_NOSWIZZLE
)
2496 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
2502 shader_addline(buffer
, "%s%s %s, %s;\n", instruction
, shader_arb_get_modifier(ins
), dst
, src
);
2505 static void shader_hw_nrm(const struct wined3d_shader_instruction
*ins
)
2507 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2510 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2511 BOOL pshader
= shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2512 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2514 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2515 shader_arb_get_src_param(ins
, &ins
->src
[0], 1 /* Use TB */, src_name
);
2517 /* In D3D, NRM of a vector with length zero returns zero. Catch this situation, as
2518 * otherwise NRM or RSQ would return NaN */
2519 if(pshader
&& priv
->target_version
>= NV3
)
2521 /* GL_NV_fragment_program2's NRM needs protection against length zero vectors too
2523 * TODO: Find out if DP3+NRM+MOV is really faster than DP3+RSQ+MUL
2525 shader_addline(buffer
, "DP3C TA, %s, %s;\n", src_name
, src_name
);
2526 shader_addline(buffer
, "NRM%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
);
2527 shader_addline(buffer
, "MOV %s (EQ), %s;\n", dst_name
, zero
);
2529 else if(priv
->target_version
>= NV2
)
2531 shader_addline(buffer
, "DP3C TA.x, %s, %s;\n", src_name
, src_name
);
2532 shader_addline(buffer
, "RSQ TA.x (NE), TA.x;\n");
2533 shader_addline(buffer
, "MUL%s %s, %s, TA.x;\n", shader_arb_get_modifier(ins
), dst_name
,
2538 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2540 shader_addline(buffer
, "DP3 TA.x, %s, %s;\n", src_name
, src_name
);
2541 /* Pass any non-zero value to RSQ if the input vector has a length of zero. The
2542 * RSQ result doesn't matter, as long as multiplying it by 0 returns 0.
2544 shader_addline(buffer
, "SGE TA.y, -TA.x, %s;\n", zero
);
2545 shader_addline(buffer
, "MAD TA.x, %s, TA.y, TA.x;\n", one
);
2547 shader_addline(buffer
, "RSQ TA.x, TA.x;\n");
2548 /* dst.w = src[0].w * 1 / (src.x^2 + src.y^2 + src.z^2)^(1/2) according to msdn*/
2549 shader_addline(buffer
, "MUL%s %s, %s, TA.x;\n", shader_arb_get_modifier(ins
), dst_name
,
2554 static void shader_hw_lrp(const struct wined3d_shader_instruction
*ins
)
2556 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2558 char src_name
[3][50];
2560 /* ARB_fragment_program has a convenient LRP instruction */
2561 if(shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)) {
2562 shader_hw_map2gl(ins
);
2566 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2567 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
2568 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
2569 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
2571 shader_addline(buffer
, "SUB TA, %s, %s;\n", src_name
[1], src_name
[2]);
2572 shader_addline(buffer
, "MAD%s %s, %s, TA, %s;\n", shader_arb_get_modifier(ins
),
2573 dst_name
, src_name
[0], src_name
[2]);
2576 static void shader_hw_sincos(const struct wined3d_shader_instruction
*ins
)
2578 /* This instruction exists in ARB, but the d3d instruction takes two extra parameters which
2579 * must contain fixed constants. So we need a separate function to filter those constants and
2582 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2583 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2584 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2586 char src_name0
[50], src_name1
[50], src_name2
[50];
2589 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
2590 if(shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)) {
2591 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2592 /* No modifiers are supported on SCS */
2593 shader_addline(buffer
, "SCS %s, %s;\n", dst_name
, src_name0
);
2595 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
)
2597 shader_arb_get_register_name(ins
, &dst
->reg
, src_name0
, &is_color
);
2598 shader_addline(buffer
, "MOV_SAT %s, %s;\n", dst_name
, src_name0
);
2600 } else if(priv
->target_version
>= NV2
) {
2601 shader_arb_get_register_name(ins
, &dst
->reg
, dst_name
, &is_color
);
2603 /* Sincos writemask must be .x, .y or .xy */
2604 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
)
2605 shader_addline(buffer
, "COS%s %s.x, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name0
);
2606 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
)
2607 shader_addline(buffer
, "SIN%s %s.y, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name0
);
2609 /* Approximate sine and cosine with a taylor series, as per math textbook. The application passes 8
2610 * helper constants(D3DSINCOSCONST1 and D3DSINCOSCONST2) in src1 and src2.
2612 * sin(x) = x - x^3/3! + x^5/5! - x^7/7! + ...
2613 * cos(x) = 1 - x^2/2! + x^4/4! - x^6/6! + ...
2615 * The constants we get are:
2617 * +1 +1, -1 -1 +1 +1 -1 -1
2618 * ---- , ---- , ---- , ----- , ----- , ----- , ------
2619 * 1!*2 2!*4 3!*8 4!*16 5!*32 6!*64 7!*128
2621 * If used with x^2, x^3, x^4 etc they calculate sin(x/2) and cos(x/2):
2625 * (x/2)^4 = x^4 / 16
2626 * (x/2)^5 = x^5 / 32
2629 * To get the final result:
2630 * sin(x) = 2 * sin(x/2) * cos(x/2)
2631 * cos(x) = cos(x/2)^2 - sin(x/2)^2
2632 * (from sin(x+y) and cos(x+y) rules)
2634 * As per MSDN, dst.z is undefined after the operation, and so is
2635 * dst.x and dst.y if they're masked out by the writemask. Ie
2636 * sincos dst.y, src1, c0, c1
2637 * returns the sine in dst.y. dst.x and dst.z are undefined, dst.w is not touched. The assembler
2638 * vsa.exe also stops with an error if the dest register is the same register as the source
2639 * register. This means we can use dest.xyz as temporary storage. The assembler vsa.exe output also
2640 * indicates that sincos consumes 8 instruction slots in vs_2_0(and, strangely, in vs_3_0).
2642 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
2643 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name2
);
2644 shader_arb_get_register_name(ins
, &dst
->reg
, dst_name
, &is_color
);
2646 shader_addline(buffer
, "MUL %s.x, %s, %s;\n", dst_name
, src_name0
, src_name0
); /* x ^ 2 */
2647 shader_addline(buffer
, "MUL TA.y, %s.x, %s;\n", dst_name
, src_name0
); /* x ^ 3 */
2648 shader_addline(buffer
, "MUL %s.y, TA.y, %s;\n", dst_name
, src_name0
); /* x ^ 4 */
2649 shader_addline(buffer
, "MUL TA.z, %s.y, %s;\n", dst_name
, src_name0
); /* x ^ 5 */
2650 shader_addline(buffer
, "MUL %s.z, TA.z, %s;\n", dst_name
, src_name0
); /* x ^ 6 */
2651 shader_addline(buffer
, "MUL TA.w, %s.z, %s;\n", dst_name
, src_name0
); /* x ^ 7 */
2655 * Unfortunately we don't get the constants in a DP4-capable form. Is there a way to
2656 * properly merge that with MULs in the code above?
2657 * The swizzles .yz and xw however fit into the .yzxw swizzle added to ps_2_0. Maybe
2658 * we can merge the sine and cosine MAD rows to calculate them together.
2660 shader_addline(buffer
, "MUL TA.x, %s, %s.w;\n", src_name0
, src_name2
); /* x^1, +1/(1!*2) */
2661 shader_addline(buffer
, "MAD TA.x, TA.y, %s.x, TA.x;\n", src_name2
); /* -1/(3!*8) */
2662 shader_addline(buffer
, "MAD TA.x, TA.z, %s.w, TA.x;\n", src_name1
); /* +1/(5!*32) */
2663 shader_addline(buffer
, "MAD TA.x, TA.w, %s.x, TA.x;\n", src_name1
); /* -1/(7!*128) */
2666 shader_addline(buffer
, "MAD TA.y, %s.x, %s.y, %s.z;\n", dst_name
, src_name2
, src_name2
); /* -1/(2!*4), +1.0 */
2667 shader_addline(buffer
, "MAD TA.y, %s.y, %s.z, TA.y;\n", dst_name
, src_name1
); /* +1/(4!*16) */
2668 shader_addline(buffer
, "MAD TA.y, %s.z, %s.y, TA.y;\n", dst_name
, src_name1
); /* -1/(6!*64) */
2670 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
) {
2672 shader_addline(buffer
, "MUL TA.z, TA.y, TA.y;\n");
2673 shader_addline(buffer
, "MAD %s.x, -TA.x, TA.x, TA.z;\n", dst_name
);
2675 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
) {
2677 shader_addline(buffer
, "MUL %s.y, TA.x, TA.y;\n", dst_name
);
2678 shader_addline(buffer
, "ADD %s.y, %s.y, %s.y;\n", dst_name
, dst_name
, dst_name
);
2683 static void shader_hw_sgn(const struct wined3d_shader_instruction
*ins
)
2685 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2688 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
2690 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2691 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
);
2693 /* SGN is only valid in vertex shaders */
2694 if(ctx
->target_version
>= NV2
) {
2695 shader_addline(buffer
, "SSG%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
);
2699 /* If SRC > 0.0, -SRC < SRC = TRUE, otherwise false.
2700 * if SRC < 0.0, SRC < -SRC = TRUE. If neither is true, src = 0.0
2702 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
) {
2703 shader_addline(buffer
, "SLT %s, -%s, %s;\n", dst_name
, src_name
, src_name
);
2705 /* src contains TA? Write to the dest first. This won't overwrite our destination.
2706 * Then use TA, and calculate the final result
2708 * Not reading from TA? Store the first result in TA to avoid overwriting the
2709 * destination if src reg = dst reg
2711 if(strstr(src_name
, "TA"))
2713 shader_addline(buffer
, "SLT %s, %s, -%s;\n", dst_name
, src_name
, src_name
);
2714 shader_addline(buffer
, "SLT TA, -%s, %s;\n", src_name
, src_name
);
2715 shader_addline(buffer
, "ADD %s, %s, -TA;\n", dst_name
, dst_name
);
2719 shader_addline(buffer
, "SLT TA, -%s, %s;\n", src_name
, src_name
);
2720 shader_addline(buffer
, "SLT %s, %s, -%s;\n", dst_name
, src_name
, src_name
);
2721 shader_addline(buffer
, "ADD %s, TA, -%s;\n", dst_name
, dst_name
);
2726 static void shader_hw_dsy(const struct wined3d_shader_instruction
*ins
)
2728 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2734 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
);
2735 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src
);
2736 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2738 shader_addline(buffer
, "DDY %s, %s;\n", dst
, src
);
2739 shader_addline(buffer
, "MUL%s %s, %s, ycorrection.y;\n", shader_arb_get_modifier(ins
), dst
, dst_name
);
2742 static DWORD
abs_modifier(DWORD mod
, BOOL
*need_abs
)
2748 case WINED3DSPSM_NONE
: return WINED3DSPSM_ABS
;
2749 case WINED3DSPSM_NEG
: return WINED3DSPSM_ABS
;
2750 case WINED3DSPSM_BIAS
: *need_abs
= TRUE
; return WINED3DSPSM_BIAS
;
2751 case WINED3DSPSM_BIASNEG
: *need_abs
= TRUE
; return WINED3DSPSM_BIASNEG
;
2752 case WINED3DSPSM_SIGN
: *need_abs
= TRUE
; return WINED3DSPSM_SIGN
;
2753 case WINED3DSPSM_SIGNNEG
: *need_abs
= TRUE
; return WINED3DSPSM_SIGNNEG
;
2754 case WINED3DSPSM_COMP
: *need_abs
= TRUE
; return WINED3DSPSM_COMP
;
2755 case WINED3DSPSM_X2
: *need_abs
= TRUE
; return WINED3DSPSM_X2
;
2756 case WINED3DSPSM_X2NEG
: *need_abs
= TRUE
; return WINED3DSPSM_X2NEG
;
2757 case WINED3DSPSM_DZ
: *need_abs
= TRUE
; return WINED3DSPSM_DZ
;
2758 case WINED3DSPSM_DW
: *need_abs
= TRUE
; return WINED3DSPSM_DW
;
2759 case WINED3DSPSM_ABS
: return WINED3DSPSM_ABS
;
2760 case WINED3DSPSM_ABSNEG
: return WINED3DSPSM_ABS
;
2762 FIXME("Unknown modifier %u\n", mod
);
2766 static void shader_hw_log(const struct wined3d_shader_instruction
*ins
)
2768 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2769 char src0
[50], dst
[50];
2770 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
2771 BOOL need_abs
= FALSE
;
2774 switch(ins
->handler_idx
)
2776 case WINED3DSIH_LOG
: instr
= "LG2"; break;
2777 case WINED3DSIH_LOGP
: instr
= "LOG"; break;
2779 ERR("Unexpected instruction %d\n", ins
->handler_idx
);
2783 /* LOG and LOGP operate on the absolute value of the input */
2784 src0_copy
.modifiers
= abs_modifier(src0_copy
.modifiers
, &need_abs
);
2786 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
);
2787 shader_arb_get_src_param(ins
, &src0_copy
, 0, src0
);
2791 shader_addline(buffer
, "ABS TA, %s;\n", src0
);
2792 shader_addline(buffer
, "%s%s %s, TA;\n", instr
, shader_arb_get_modifier(ins
), dst
);
2796 shader_addline(buffer
, "%s%s %s, %s;\n", instr
, shader_arb_get_modifier(ins
), dst
, src0
);
2800 static void shader_hw_pow(const struct wined3d_shader_instruction
*ins
)
2802 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2803 char src0
[50], src1
[50], dst
[50];
2804 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
2805 BOOL need_abs
= FALSE
;
2806 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2807 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2809 /* POW operates on the absolute value of the input */
2810 src0_copy
.modifiers
= abs_modifier(src0_copy
.modifiers
, &need_abs
);
2812 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
);
2813 shader_arb_get_src_param(ins
, &src0_copy
, 0, src0
);
2814 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src1
);
2817 shader_addline(buffer
, "ABS TA.x, %s;\n", src0
);
2819 shader_addline(buffer
, "MOV TA.x, %s;\n", src0
);
2821 if (priv
->target_version
>= NV2
)
2823 shader_addline(buffer
, "MOVC TA.y, %s;\n", src1
);
2824 shader_addline(buffer
, "POW%s %s, TA.x, TA.y;\n", shader_arb_get_modifier(ins
), dst
);
2825 shader_addline(buffer
, "MOV %s (EQ.y), %s;\n", dst
, one
);
2829 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2830 const char *flt_eps
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_EPS
);
2832 shader_addline(buffer
, "ABS TA.y, %s;\n", src1
);
2833 shader_addline(buffer
, "SGE TA.y, -TA.y, %s;\n", zero
);
2834 /* Possibly add flt_eps to avoid getting float special values */
2835 shader_addline(buffer
, "MAD TA.z, TA.y, %s, %s;\n", flt_eps
, src1
);
2836 shader_addline(buffer
, "POW%s TA.x, TA.x, TA.z;\n", shader_arb_get_modifier(ins
));
2837 shader_addline(buffer
, "MAD TA.x, -TA.x, TA.y, TA.x;\n");
2838 shader_addline(buffer
, "MAD %s, TA.y, %s, TA.x;\n", dst
, one
);
2842 static void shader_hw_loop(const struct wined3d_shader_instruction
*ins
)
2844 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2846 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2849 shader_arb_get_src_param(ins
, &ins
->src
[1], 0, src_name
);
2853 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2854 struct list
*e
= list_head(&priv
->control_frames
);
2855 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2857 if(priv
->loop_depth
> 1) shader_addline(buffer
, "PUSHA aL;\n");
2858 /* The constant loader makes sure to load -1 into iX.w */
2859 shader_addline(buffer
, "ARLC aL, %s.xywz;\n", src_name
);
2860 shader_addline(buffer
, "BRA loop_%u_end (LE.x);\n", control_frame
->no
.loop
);
2861 shader_addline(buffer
, "loop_%u_start:\n", control_frame
->no
.loop
);
2865 shader_addline(buffer
, "LOOP %s;\n", src_name
);
2869 static void shader_hw_rep(const struct wined3d_shader_instruction
*ins
)
2871 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2873 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2875 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
);
2877 /* The constant loader makes sure to load -1 into iX.w */
2880 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2881 struct list
*e
= list_head(&priv
->control_frames
);
2882 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2884 if(priv
->loop_depth
> 1) shader_addline(buffer
, "PUSHA aL;\n");
2886 shader_addline(buffer
, "ARLC aL, %s.xywz;\n", src_name
);
2887 shader_addline(buffer
, "BRA loop_%u_end (LE.x);\n", control_frame
->no
.loop
);
2888 shader_addline(buffer
, "loop_%u_start:\n", control_frame
->no
.loop
);
2892 shader_addline(buffer
, "REP %s;\n", src_name
);
2896 static void shader_hw_endloop(const struct wined3d_shader_instruction
*ins
)
2898 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2899 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2903 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2904 struct list
*e
= list_head(&priv
->control_frames
);
2905 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2907 shader_addline(buffer
, "ARAC aL.xy, aL;\n");
2908 shader_addline(buffer
, "BRA loop_%u_start (GT.x);\n", control_frame
->no
.loop
);
2909 shader_addline(buffer
, "loop_%u_end:\n", control_frame
->no
.loop
);
2911 if(priv
->loop_depth
> 1) shader_addline(buffer
, "POPA aL;\n");
2915 shader_addline(buffer
, "ENDLOOP;\n");
2919 static void shader_hw_endrep(const struct wined3d_shader_instruction
*ins
)
2921 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2922 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2926 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2927 struct list
*e
= list_head(&priv
->control_frames
);
2928 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2930 shader_addline(buffer
, "ARAC aL.xy, aL;\n");
2931 shader_addline(buffer
, "BRA loop_%u_start (GT.x);\n", control_frame
->no
.loop
);
2932 shader_addline(buffer
, "loop_%u_end:\n", control_frame
->no
.loop
);
2934 if(priv
->loop_depth
> 1) shader_addline(buffer
, "POPA aL;\n");
2938 shader_addline(buffer
, "ENDREP;\n");
2942 static const struct control_frame
*find_last_loop(const struct shader_arb_ctx_priv
*priv
)
2944 struct control_frame
*control_frame
;
2946 LIST_FOR_EACH_ENTRY(control_frame
, &priv
->control_frames
, struct control_frame
, entry
)
2948 if(control_frame
->type
== LOOP
|| control_frame
->type
== REP
) return control_frame
;
2950 ERR("Could not find loop for break\n");
2954 static void shader_hw_break(const struct wined3d_shader_instruction
*ins
)
2956 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2957 const struct control_frame
*control_frame
= find_last_loop(ins
->ctx
->backend_data
);
2958 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2962 shader_addline(buffer
, "BRA loop_%u_end;\n", control_frame
->no
.loop
);
2966 shader_addline(buffer
, "BRK;\n");
2970 static const char *get_compare(enum wined3d_shader_rel_op op
)
2974 case WINED3D_SHADER_REL_OP_GT
: return "GT";
2975 case WINED3D_SHADER_REL_OP_EQ
: return "EQ";
2976 case WINED3D_SHADER_REL_OP_GE
: return "GE";
2977 case WINED3D_SHADER_REL_OP_LT
: return "LT";
2978 case WINED3D_SHADER_REL_OP_NE
: return "NE";
2979 case WINED3D_SHADER_REL_OP_LE
: return "LE";
2981 FIXME("Unrecognized operator %#x.\n", op
);
2986 static enum wined3d_shader_rel_op
invert_compare(enum wined3d_shader_rel_op op
)
2990 case WINED3D_SHADER_REL_OP_GT
: return WINED3D_SHADER_REL_OP_LE
;
2991 case WINED3D_SHADER_REL_OP_EQ
: return WINED3D_SHADER_REL_OP_NE
;
2992 case WINED3D_SHADER_REL_OP_GE
: return WINED3D_SHADER_REL_OP_LT
;
2993 case WINED3D_SHADER_REL_OP_LT
: return WINED3D_SHADER_REL_OP_GE
;
2994 case WINED3D_SHADER_REL_OP_NE
: return WINED3D_SHADER_REL_OP_EQ
;
2995 case WINED3D_SHADER_REL_OP_LE
: return WINED3D_SHADER_REL_OP_GT
;
2997 FIXME("Unrecognized operator %#x.\n", op
);
3002 static void shader_hw_breakc(const struct wined3d_shader_instruction
*ins
)
3004 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3005 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3006 const struct control_frame
*control_frame
= find_last_loop(ins
->ctx
->backend_data
);
3009 const char *comp
= get_compare(ins
->flags
);
3011 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
3012 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
3016 /* SUBC CC, src0, src1" works only in pixel shaders, so use TA to throw
3017 * away the subtraction result
3019 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3020 shader_addline(buffer
, "BRA loop_%u_end (%s.x);\n", control_frame
->no
.loop
, comp
);
3024 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3025 shader_addline(buffer
, "BRK (%s.x);\n", comp
);
3029 static void shader_hw_ifc(const struct wined3d_shader_instruction
*ins
)
3031 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3032 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3033 struct list
*e
= list_head(&priv
->control_frames
);
3034 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3038 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3040 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
3041 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
3045 /* Invert the flag. We jump to the else label if the condition is NOT true */
3046 comp
= get_compare(invert_compare(ins
->flags
));
3047 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3048 shader_addline(buffer
, "BRA ifc_%u_else (%s.x);\n", control_frame
->no
.ifc
, comp
);
3052 comp
= get_compare(ins
->flags
);
3053 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3054 shader_addline(buffer
, "IF %s.x;\n", comp
);
3058 static void shader_hw_else(const struct wined3d_shader_instruction
*ins
)
3060 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3061 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3062 struct list
*e
= list_head(&priv
->control_frames
);
3063 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3064 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3068 shader_addline(buffer
, "BRA ifc_%u_endif;\n", control_frame
->no
.ifc
);
3069 shader_addline(buffer
, "ifc_%u_else:\n", control_frame
->no
.ifc
);
3070 control_frame
->had_else
= TRUE
;
3074 shader_addline(buffer
, "ELSE;\n");
3078 static void shader_hw_endif(const struct wined3d_shader_instruction
*ins
)
3080 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3081 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3082 struct list
*e
= list_head(&priv
->control_frames
);
3083 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3084 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3088 if(control_frame
->had_else
)
3090 shader_addline(buffer
, "ifc_%u_endif:\n", control_frame
->no
.ifc
);
3094 shader_addline(buffer
, "#No else branch. else is endif\n");
3095 shader_addline(buffer
, "ifc_%u_else:\n", control_frame
->no
.ifc
);
3100 shader_addline(buffer
, "ENDIF;\n");
3104 static void shader_hw_texldd(const struct wined3d_shader_instruction
*ins
)
3106 DWORD sampler_idx
= ins
->src
[1].reg
.idx
[0].offset
;
3108 char reg_src
[3][40];
3109 WORD flags
= TEX_DERIV
;
3111 shader_arb_get_dst_param(ins
, &ins
->dst
[0], reg_dest
);
3112 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_src
[0]);
3113 shader_arb_get_src_param(ins
, &ins
->src
[2], 1, reg_src
[1]);
3114 shader_arb_get_src_param(ins
, &ins
->src
[3], 2, reg_src
[2]);
3116 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) flags
|= TEX_PROJ
;
3117 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) flags
|= TEX_BIAS
;
3119 shader_hw_sample(ins
, sampler_idx
, reg_dest
, reg_src
[0], flags
, reg_src
[1], reg_src
[2]);
3122 static void shader_hw_texldl(const struct wined3d_shader_instruction
*ins
)
3124 DWORD sampler_idx
= ins
->src
[1].reg
.idx
[0].offset
;
3127 WORD flags
= TEX_LOD
;
3129 shader_arb_get_dst_param(ins
, &ins
->dst
[0], reg_dest
);
3130 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_coord
);
3132 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) flags
|= TEX_PROJ
;
3133 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) flags
|= TEX_BIAS
;
3135 shader_hw_sample(ins
, sampler_idx
, reg_dest
, reg_coord
, flags
, NULL
, NULL
);
3138 static void shader_hw_label(const struct wined3d_shader_instruction
*ins
)
3140 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3141 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3143 priv
->in_main_func
= FALSE
;
3144 /* Call instructions activate the NV extensions, not labels and rets. If there is an uncalled
3145 * subroutine, don't generate a label that will make GL complain
3147 if(priv
->target_version
== ARB
) return;
3149 shader_addline(buffer
, "l%u:\n", ins
->src
[0].reg
.idx
[0].offset
);
3152 static void vshader_add_footer(struct shader_arb_ctx_priv
*priv_ctx
,
3153 const struct arb_vshader_private
*shader_data
, const struct arb_vs_compile_args
*args
,
3154 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
,
3155 struct wined3d_shader_buffer
*buffer
)
3159 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
3160 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
3161 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
3162 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
3164 if (args
->super
.fog_src
== VS_FOG_Z
)
3166 shader_addline(buffer
, "MOV result.fogcoord, TMP_OUT.z;\n");
3172 /* posFixup.x is always 1.0, so we can safely use it */
3173 shader_addline(buffer
, "ADD result.fogcoord, posFixup.x, -posFixup.x;\n");
3177 /* Clamp fogcoord */
3178 const char *zero
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ZERO
);
3179 const char *one
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ONE
);
3181 shader_addline(buffer
, "MIN TMP_FOGCOORD.x, TMP_FOGCOORD.x, %s;\n", one
);
3182 shader_addline(buffer
, "MAX result.fogcoord.x, TMP_FOGCOORD.x, %s;\n", zero
);
3186 /* Clipplanes are always stored without y inversion */
3187 if (use_nv_clip(gl_info
) && priv_ctx
->target_version
>= NV2
)
3189 if (args
->super
.clip_enabled
)
3191 for (i
= 0; i
< priv_ctx
->vs_clipplanes
; i
++)
3193 shader_addline(buffer
, "DP4 result.clip[%u].x, TMP_OUT, state.clip[%u].plane;\n", i
, i
);
3197 else if (args
->clip
.boolclip
.clip_texcoord
)
3199 unsigned int cur_clip
= 0;
3200 char component
[4] = {'x', 'y', 'z', 'w'};
3201 const char *zero
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_ZERO
);
3203 for (i
= 0; i
< gl_info
->limits
.clipplanes
; ++i
)
3205 if (args
->clip
.boolclip
.clipplane_mask
& (1 << i
))
3207 shader_addline(buffer
, "DP4 TA.%c, TMP_OUT, state.clip[%u].plane;\n",
3208 component
[cur_clip
++], i
);
3214 shader_addline(buffer
, "MOV TA, %s;\n", zero
);
3217 shader_addline(buffer
, "MOV TA.yzw, %s;\n", zero
);
3220 shader_addline(buffer
, "MOV TA.zw, %s;\n", zero
);
3223 shader_addline(buffer
, "MOV TA.w, %s;\n", zero
);
3226 shader_addline(buffer
, "MOV result.texcoord[%u], TA;\n",
3227 args
->clip
.boolclip
.clip_texcoord
- 1);
3230 /* Write the final position.
3232 * OpenGL coordinates specify the center of the pixel while d3d coords specify
3233 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
3234 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
3235 * contains 1.0 to allow a mad, but arb vs swizzles are too restricted for that.
3237 shader_addline(buffer
, "MUL TA, posFixup, TMP_OUT.w;\n");
3238 shader_addline(buffer
, "ADD TMP_OUT.x, TMP_OUT.x, TA.z;\n");
3239 shader_addline(buffer
, "MAD TMP_OUT.y, TMP_OUT.y, posFixup.y, TA.w;\n");
3241 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
3242 * and the glsl equivalent
3244 if (need_helper_const(shader_data
, reg_maps
, gl_info
))
3246 const char *two
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_TWO
);
3247 shader_addline(buffer
, "MAD TMP_OUT.z, TMP_OUT.z, %s, -TMP_OUT.w;\n", two
);
3251 shader_addline(buffer
, "ADD TMP_OUT.z, TMP_OUT.z, TMP_OUT.z;\n");
3252 shader_addline(buffer
, "ADD TMP_OUT.z, TMP_OUT.z, -TMP_OUT.w;\n");
3255 shader_addline(buffer
, "MOV result.position, TMP_OUT;\n");
3257 priv_ctx
->footer_written
= TRUE
;
3260 static void shader_hw_ret(const struct wined3d_shader_instruction
*ins
)
3262 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3263 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3264 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
3265 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3267 if(priv
->target_version
== ARB
) return;
3271 if (priv
->in_main_func
) vshader_add_footer(priv
, shader
->backend_data
,
3272 priv
->cur_vs_args
, ins
->ctx
->reg_maps
, ins
->ctx
->gl_info
, buffer
);
3275 shader_addline(buffer
, "RET;\n");
3278 static void shader_hw_call(const struct wined3d_shader_instruction
*ins
)
3280 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3281 shader_addline(buffer
, "CAL l%u;\n", ins
->src
[0].reg
.idx
[0].offset
);
3284 /* Context activation is done by the caller. */
3285 static GLuint
create_arb_blt_vertex_program(const struct wined3d_gl_info
*gl_info
)
3287 GLuint program_id
= 0;
3290 const char *blt_vprogram
=
3292 "PARAM c[1] = { { 1, 0.5 } };\n"
3293 "MOV result.position, vertex.position;\n"
3294 "MOV result.color, c[0].x;\n"
3295 "MOV result.texcoord[0], vertex.texcoord[0];\n"
3298 GL_EXTCALL(glGenProgramsARB(1, &program_id
));
3299 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, program_id
));
3300 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
3301 strlen(blt_vprogram
), blt_vprogram
));
3302 checkGLcall("glProgramStringARB()");
3304 gl_info
->gl_ops
.gl
.p_glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
3307 FIXME("Vertex program error at position %d: %s\n\n", pos
,
3308 debugstr_a((const char *)gl_info
->gl_ops
.gl
.p_glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
3309 shader_arb_dump_program_source(blt_vprogram
);
3315 GL_EXTCALL(glGetProgramivARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
3316 checkGLcall("glGetProgramivARB()");
3317 if (!native
) WARN("Program exceeds native resource limits.\n");
3323 /* Context activation is done by the caller. */
3324 static GLuint
create_arb_blt_fragment_program(const struct wined3d_gl_info
*gl_info
,
3325 enum tex_types tex_type
, BOOL masked
)
3327 GLuint program_id
= 0;
3328 const char *fprogram
;
3331 static const char * const blt_fprograms_full
[tex_type_count
] =
3338 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
3339 "MOV result.depth.z, R0.x;\n"
3346 "TEX R0.x, fragment.texcoord[0], texture[0], CUBE;\n"
3347 "MOV result.depth.z, R0.x;\n"
3352 "TEX R0.x, fragment.texcoord[0], texture[0], RECT;\n"
3353 "MOV result.depth.z, R0.x;\n"
3357 static const char * const blt_fprograms_masked
[tex_type_count
] =
3363 "PARAM mask = program.local[0];\n"
3365 "SLT R0.xy, fragment.position, mask.zwzw;\n"
3366 "MUL R0.x, R0.x, R0.y;\n"
3368 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
3369 "MOV result.depth.z, R0.x;\n"
3375 "PARAM mask = program.local[0];\n"
3377 "SLT R0.xy, fragment.position, mask.zwzw;\n"
3378 "MUL R0.x, R0.x, R0.y;\n"
3380 "TEX R0.x, fragment.texcoord[0], texture[0], CUBE;\n"
3381 "MOV result.depth.z, R0.x;\n"
3385 "PARAM mask = program.local[0];\n"
3387 "SLT R0.xy, fragment.position, mask.zwzw;\n"
3388 "MUL R0.x, R0.x, R0.y;\n"
3390 "TEX R0.x, fragment.texcoord[0], texture[0], RECT;\n"
3391 "MOV result.depth.z, R0.x;\n"
3395 fprogram
= masked
? blt_fprograms_masked
[tex_type
] : blt_fprograms_full
[tex_type
];
3398 FIXME("tex_type %#x not supported, falling back to tex_2d\n", tex_type
);
3400 fprogram
= masked
? blt_fprograms_masked
[tex_type
] : blt_fprograms_full
[tex_type
];
3403 GL_EXTCALL(glGenProgramsARB(1, &program_id
));
3404 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, program_id
));
3405 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
, strlen(fprogram
), fprogram
));
3406 checkGLcall("glProgramStringARB()");
3408 gl_info
->gl_ops
.gl
.p_glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
3411 FIXME("Fragment program error at position %d: %s\n\n", pos
,
3412 debugstr_a((const char *)gl_info
->gl_ops
.gl
.p_glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
3413 shader_arb_dump_program_source(fprogram
);
3419 GL_EXTCALL(glGetProgramivARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
3420 checkGLcall("glGetProgramivARB()");
3421 if (!native
) WARN("Program exceeds native resource limits.\n");
3427 static void arbfp_add_sRGB_correction(struct wined3d_shader_buffer
*buffer
, const char *fragcolor
,
3428 const char *tmp1
, const char *tmp2
, const char *tmp3
, const char *tmp4
, BOOL condcode
)
3430 /* Perform sRGB write correction. See GLX_EXT_framebuffer_sRGB */
3434 /* Sigh. MOVC CC doesn't work, so use one of the temps as dummy dest */
3435 shader_addline(buffer
, "SUBC %s, %s.x, srgb_consts1.y;\n", tmp1
, fragcolor
);
3436 /* Calculate the > 0.0031308 case */
3437 shader_addline(buffer
, "POW %s.x (GE), %s.x, srgb_consts1.z;\n", fragcolor
, fragcolor
);
3438 shader_addline(buffer
, "POW %s.y (GE), %s.y, srgb_consts1.z;\n", fragcolor
, fragcolor
);
3439 shader_addline(buffer
, "POW %s.z (GE), %s.z, srgb_consts1.z;\n", fragcolor
, fragcolor
);
3440 shader_addline(buffer
, "MUL %s.xyz (GE), %s, srgb_consts1.w;\n", fragcolor
, fragcolor
);
3441 shader_addline(buffer
, "SUB %s.xyz (GE), %s, srgb_consts2.x;\n", fragcolor
, fragcolor
);
3442 /* Calculate the < case */
3443 shader_addline(buffer
, "MUL %s.xyz (LT), srgb_consts1.x, %s;\n", fragcolor
, fragcolor
);
3447 /* Calculate the > 0.0031308 case */
3448 shader_addline(buffer
, "POW %s.x, %s.x, srgb_consts1.z;\n", tmp1
, fragcolor
);
3449 shader_addline(buffer
, "POW %s.y, %s.y, srgb_consts1.z;\n", tmp1
, fragcolor
);
3450 shader_addline(buffer
, "POW %s.z, %s.z, srgb_consts1.z;\n", tmp1
, fragcolor
);
3451 shader_addline(buffer
, "MUL %s, %s, srgb_consts1.w;\n", tmp1
, tmp1
);
3452 shader_addline(buffer
, "SUB %s, %s, srgb_consts2.x;\n", tmp1
, tmp1
);
3453 /* Calculate the < case */
3454 shader_addline(buffer
, "MUL %s, srgb_consts1.x, %s;\n", tmp2
, fragcolor
);
3455 /* Get 1.0 / 0.0 masks for > 0.0031308 and < 0.0031308 */
3456 shader_addline(buffer
, "SLT %s, srgb_consts1.y, %s;\n", tmp3
, fragcolor
);
3457 shader_addline(buffer
, "SGE %s, srgb_consts1.y, %s;\n", tmp4
, fragcolor
);
3458 /* Store the components > 0.0031308 in the destination */
3459 shader_addline(buffer
, "MUL %s.xyz, %s, %s;\n", fragcolor
, tmp1
, tmp3
);
3460 /* Add the components that are < 0.0031308 */
3461 shader_addline(buffer
, "MAD %s.xyz, %s, %s, %s;\n", fragcolor
, tmp2
, tmp4
, fragcolor
);
3462 /* Move everything into result.color at once. Nvidia hardware cannot handle partial
3463 * result.color writes(.rgb first, then .a), or handle overwriting already written
3464 * components. The assembler uses a temporary register in this case, which is usually
3465 * not allocated from one of our registers that were used earlier.
3468 /* [0.0;1.0] clamping. Not needed, this is done implicitly */
3471 static const DWORD
*find_loop_control_values(const struct wined3d_shader
*shader
, DWORD idx
)
3473 const struct wined3d_shader_lconst
*constant
;
3475 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsI
, struct wined3d_shader_lconst
, entry
)
3477 if (constant
->idx
== idx
)
3479 return constant
->value
;
3485 static void init_ps_input(const struct wined3d_shader
*shader
,
3486 const struct arb_ps_compile_args
*args
, struct shader_arb_ctx_priv
*priv
)
3488 static const char * const texcoords
[8] =
3490 "fragment.texcoord[0]", "fragment.texcoord[1]", "fragment.texcoord[2]", "fragment.texcoord[3]",
3491 "fragment.texcoord[4]", "fragment.texcoord[5]", "fragment.texcoord[6]", "fragment.texcoord[7]"
3494 const struct wined3d_shader_signature_element
*sig
= shader
->input_signature
;
3495 const char *semantic_name
;
3498 switch(args
->super
.vp_mode
)
3500 case pretransformed
:
3502 /* The pixelshader has to collect the varyings on its own. In any case properly load
3503 * color0 and color1. In the case of pretransformed vertices also load texcoords. Set
3504 * other attribs to 0.0.
3506 * For fixedfunction this behavior is correct, according to the tests. For pretransformed
3507 * we'd either need a replacement shader that can load other attribs like BINORMAL, or
3508 * load the texcoord attrib pointers to match the pixel shader signature
3510 for(i
= 0; i
< MAX_REG_INPUT
; i
++)
3512 semantic_name
= sig
[i
].semantic_name
;
3513 semantic_idx
= sig
[i
].semantic_idx
;
3514 if (!semantic_name
) continue;
3516 if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_COLOR
))
3518 if (!semantic_idx
) priv
->ps_input
[i
] = "fragment.color.primary";
3519 else if(semantic_idx
== 1) priv
->ps_input
[i
] = "fragment.color.secondary";
3520 else priv
->ps_input
[i
] = "0.0";
3522 else if(args
->super
.vp_mode
== fixedfunction
)
3524 priv
->ps_input
[i
] = "0.0";
3526 else if(shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
3528 if(semantic_idx
< 8) priv
->ps_input
[i
] = texcoords
[semantic_idx
];
3529 else priv
->ps_input
[i
] = "0.0";
3531 else if(shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_FOG
))
3533 if (!semantic_idx
) priv
->ps_input
[i
] = "fragment.fogcoord";
3534 else priv
->ps_input
[i
] = "0.0";
3538 priv
->ps_input
[i
] = "0.0";
3541 TRACE("v%u, semantic %s%u is %s\n", i
, semantic_name
, semantic_idx
, priv
->ps_input
[i
]);
3546 /* That one is easy. The vertex shaders provide v0-v7 in fragment.texcoord and v8 and v9 in
3549 for(i
= 0; i
< 8; i
++)
3551 priv
->ps_input
[i
] = texcoords
[i
];
3553 priv
->ps_input
[8] = "fragment.color.primary";
3554 priv
->ps_input
[9] = "fragment.color.secondary";
3559 /* Context activation is done by the caller. */
3560 static GLuint
shader_arb_generate_pshader(const struct wined3d_shader
*shader
,
3561 const struct wined3d_gl_info
*gl_info
, struct wined3d_shader_buffer
*buffer
,
3562 const struct arb_ps_compile_args
*args
, struct arb_ps_compiled_shader
*compiled
)
3564 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
3565 const DWORD
*function
= shader
->function
;
3568 DWORD next_local
= 0;
3569 struct shader_arb_ctx_priv priv_ctx
;
3570 BOOL dcl_td
= FALSE
;
3571 BOOL want_nv_prog
= FALSE
;
3572 struct arb_pshader_private
*shader_priv
= shader
->backend_data
;
3577 unsigned int i
, found
= 0;
3579 for (i
= 0, map
= reg_maps
->temporary
; map
; map
>>= 1, ++i
)
3582 || (shader
->u
.ps
.color0_mov
&& i
== shader
->u
.ps
.color0_reg
)
3583 || (reg_maps
->shader_version
.major
< 2 && !i
))
3586 sprintf(srgbtmp
[found
], "R%u", i
);
3588 if (found
== 4) break;
3593 sprintf(srgbtmp
[0], "TA");
3594 sprintf(srgbtmp
[1], "TB");
3595 sprintf(srgbtmp
[2], "TC");
3596 sprintf(srgbtmp
[3], "TD");
3600 sprintf(srgbtmp
[1], "TA");
3601 sprintf(srgbtmp
[2], "TB");
3602 sprintf(srgbtmp
[3], "TC");
3605 sprintf(srgbtmp
[2], "TA");
3606 sprintf(srgbtmp
[3], "TB");
3609 sprintf(srgbtmp
[3], "TA");
3615 /* Create the hw ARB shader */
3616 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
3617 priv_ctx
.cur_ps_args
= args
;
3618 priv_ctx
.compiled_fprog
= compiled
;
3619 priv_ctx
.cur_np2fixup_info
= &compiled
->np2fixup_info
;
3620 init_ps_input(shader
, args
, &priv_ctx
);
3621 list_init(&priv_ctx
.control_frames
);
3623 /* Avoid enabling NV_fragment_program* if we do not need it.
3625 * Enabling GL_NV_fragment_program_option causes the driver to occupy a temporary register,
3626 * and it slows down the shader execution noticeably(about 5%). Usually our instruction emulation
3627 * is faster than what we gain from using higher native instructions. There are some things though
3628 * that cannot be emulated. In that case enable the extensions.
3629 * If the extension is enabled, instruction handlers that support both ways will use it.
3631 * Testing shows no performance difference between OPTION NV_fragment_program2 and NV_fragment_program.
3632 * So enable the best we can get.
3634 if(reg_maps
->usesdsx
|| reg_maps
->usesdsy
|| reg_maps
->loop_depth
> 0 || reg_maps
->usestexldd
||
3635 reg_maps
->usestexldl
|| reg_maps
->usesfacing
|| reg_maps
->usesifc
|| reg_maps
->usescall
)
3637 want_nv_prog
= TRUE
;
3640 shader_addline(buffer
, "!!ARBfp1.0\n");
3641 if (want_nv_prog
&& gl_info
->supported
[NV_FRAGMENT_PROGRAM2
])
3643 shader_addline(buffer
, "OPTION NV_fragment_program2;\n");
3644 priv_ctx
.target_version
= NV3
;
3646 else if (want_nv_prog
&& gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
])
3648 shader_addline(buffer
, "OPTION NV_fragment_program;\n");
3649 priv_ctx
.target_version
= NV2
;
3653 /* This is an error - either we're advertising the wrong shader version, or aren't enforcing some
3656 ERR("The shader requires instructions that are not available in plain GL_ARB_fragment_program\n");
3659 priv_ctx
.target_version
= ARB
;
3662 if (reg_maps
->rt_mask
> 1)
3664 shader_addline(buffer
, "OPTION ARB_draw_buffers;\n");
3667 if (reg_maps
->shader_version
.major
< 3)
3669 switch(args
->super
.fog
) {
3673 shader_addline(buffer
, "OPTION ARB_fog_linear;\n");
3676 shader_addline(buffer
, "OPTION ARB_fog_exp;\n");
3679 shader_addline(buffer
, "OPTION ARB_fog_exp2;\n");
3684 /* For now always declare the temps. At least the Nvidia assembler optimizes completely
3685 * unused temps away(but occupies them for the whole shader if they're used once). Always
3686 * declaring them avoids tricky bookkeeping work
3688 shader_addline(buffer
, "TEMP TA;\n"); /* Used for modifiers */
3689 shader_addline(buffer
, "TEMP TB;\n"); /* Used for modifiers */
3690 shader_addline(buffer
, "TEMP TC;\n"); /* Used for modifiers */
3691 if(dcl_td
) shader_addline(buffer
, "TEMP TD;\n"); /* Used for sRGB writing */
3692 shader_addline(buffer
, "PARAM coefdiv = { 0.5, 0.25, 0.125, 0.0625 };\n");
3693 shader_addline(buffer
, "PARAM coefmul = { 2, 4, 8, 16 };\n");
3694 shader_addline(buffer
, "PARAM ps_helper_const = { 0.0, 1.0, %1.10f, 0.0 };\n", eps
);
3696 if (reg_maps
->shader_version
.major
< 2)
3698 strcpy(fragcolor
, "R0");
3702 if (args
->super
.srgb_correction
)
3704 if (shader
->u
.ps
.color0_mov
)
3706 sprintf(fragcolor
, "R%u", shader
->u
.ps
.color0_reg
);
3710 shader_addline(buffer
, "TEMP TMP_COLOR;\n");
3711 strcpy(fragcolor
, "TMP_COLOR");
3714 strcpy(fragcolor
, "result.color");
3718 if(args
->super
.srgb_correction
) {
3719 shader_addline(buffer
, "PARAM srgb_consts1 = {%f, %f, %f, %f};\n",
3720 srgb_mul_low
, srgb_cmp
, srgb_pow
, srgb_mul_high
);
3721 shader_addline(buffer
, "PARAM srgb_consts2 = {%f, %f, %f, %f};\n",
3722 srgb_sub_high
, 0.0, 0.0, 0.0);
3725 /* Base Declarations */
3726 shader_generate_arb_declarations(shader
, reg_maps
, buffer
, gl_info
, NULL
, &priv_ctx
);
3728 for (i
= 0, map
= reg_maps
->bumpmat
; map
; map
>>= 1, ++i
)
3730 unsigned char bump_const
;
3732 if (!(map
& 1)) continue;
3734 bump_const
= compiled
->numbumpenvmatconsts
;
3735 compiled
->bumpenvmatconst
[bump_const
].const_num
= WINED3D_CONST_NUM_UNUSED
;
3736 compiled
->bumpenvmatconst
[bump_const
].texunit
= i
;
3737 compiled
->luminanceconst
[bump_const
].const_num
= WINED3D_CONST_NUM_UNUSED
;
3738 compiled
->luminanceconst
[bump_const
].texunit
= i
;
3740 /* We can fit the constants into the constant limit for sure because texbem, texbeml, bem and beml are only supported
3741 * in 1.x shaders, and GL_ARB_fragment_program has a constant limit of 24 constants. So in the worst case we're loading
3742 * 8 shader constants, 8 bump matrices and 8 luminance parameters and are perfectly fine. (No NP2 fixup on bumpmapped
3743 * textures due to conditional NP2 restrictions)
3745 * Use local constants to load the bump env parameters, not program.env. This avoids collisions with d3d constants of
3746 * shaders in newer shader models. Since the bump env parameters have to share their space with NP2 fixup constants,
3747 * their location is shader dependent anyway and they cannot be loaded globally.
3749 compiled
->bumpenvmatconst
[bump_const
].const_num
= next_local
++;
3750 shader_addline(buffer
, "PARAM bumpenvmat%d = program.local[%d];\n",
3751 i
, compiled
->bumpenvmatconst
[bump_const
].const_num
);
3752 compiled
->numbumpenvmatconsts
= bump_const
+ 1;
3754 if (!(reg_maps
->luminanceparams
& (1 << i
))) continue;
3756 compiled
->luminanceconst
[bump_const
].const_num
= next_local
++;
3757 shader_addline(buffer
, "PARAM luminance%d = program.local[%d];\n",
3758 i
, compiled
->luminanceconst
[bump_const
].const_num
);
3761 for(i
= 0; i
< MAX_CONST_I
; i
++)
3763 compiled
->int_consts
[i
] = WINED3D_CONST_NUM_UNUSED
;
3764 if (reg_maps
->integer_constants
& (1 << i
) && priv_ctx
.target_version
>= NV2
)
3766 const DWORD
*control_values
= find_loop_control_values(shader
, i
);
3770 shader_addline(buffer
, "PARAM I%u = {%u, %u, %u, -1};\n", i
,
3771 control_values
[0], control_values
[1], control_values
[2]);
3775 compiled
->int_consts
[i
] = next_local
;
3776 compiled
->num_int_consts
++;
3777 shader_addline(buffer
, "PARAM I%u = program.local[%u];\n", i
, next_local
++);
3782 if(reg_maps
->vpos
|| reg_maps
->usesdsy
)
3784 compiled
->ycorrection
= next_local
;
3785 shader_addline(buffer
, "PARAM ycorrection = program.local[%u];\n", next_local
++);
3789 shader_addline(buffer
, "TEMP vpos;\n");
3790 /* ycorrection.x: Backbuffer height(onscreen) or 0(offscreen).
3791 * ycorrection.y: -1.0(onscreen), 1.0(offscreen)
3792 * ycorrection.z: 1.0
3793 * ycorrection.w: 0.0
3795 shader_addline(buffer
, "MAD vpos, fragment.position, ycorrection.zyww, ycorrection.wxww;\n");
3796 shader_addline(buffer
, "FLR vpos.xy, vpos;\n");
3801 compiled
->ycorrection
= WINED3D_CONST_NUM_UNUSED
;
3804 /* Load constants to fixup NP2 texcoords if there are still free constants left:
3805 * Constants (texture dimensions) for the NP2 fixup are loaded as local program parameters. This will consume
3806 * at most 8 (MAX_FRAGMENT_SAMPLERS / 2) parameters, which is highly unlikely, since the application had to
3807 * use 16 NP2 textures at the same time. In case that we run out of constants the fixup is simply not
3808 * applied / activated. This will probably result in wrong rendering of the texture, but will save us from
3809 * shader compilation errors and the subsequent errors when drawing with this shader. */
3810 if (priv_ctx
.cur_ps_args
->super
.np2_fixup
) {
3811 unsigned char cur_fixup_sampler
= 0;
3813 struct arb_ps_np2fixup_info
* const fixup
= priv_ctx
.cur_np2fixup_info
;
3814 const WORD map
= priv_ctx
.cur_ps_args
->super
.np2_fixup
;
3815 const UINT max_lconsts
= gl_info
->limits
.arb_ps_local_constants
;
3817 fixup
->offset
= next_local
;
3818 fixup
->super
.active
= 0;
3820 for (i
= 0; i
< MAX_FRAGMENT_SAMPLERS
; ++i
) {
3821 if (!(map
& (1 << i
))) continue;
3823 if (fixup
->offset
+ (cur_fixup_sampler
>> 1) < max_lconsts
) {
3824 fixup
->super
.active
|= (1 << i
);
3825 fixup
->super
.idx
[i
] = cur_fixup_sampler
++;
3827 FIXME("No free constant found to load NP2 fixup data into shader. "
3828 "Sampling from this texture will probably look wrong.\n");
3833 fixup
->super
.num_consts
= (cur_fixup_sampler
+ 1) >> 1;
3834 if (fixup
->super
.num_consts
) {
3835 shader_addline(buffer
, "PARAM np2fixup[%u] = { program.env[%u..%u] };\n",
3836 fixup
->super
.num_consts
, fixup
->offset
, fixup
->super
.num_consts
+ fixup
->offset
- 1);
3840 if (shader_priv
->clipplane_emulation
!= ~0U && args
->clip
)
3842 shader_addline(buffer
, "KIL fragment.texcoord[%u];\n", shader_priv
->clipplane_emulation
);
3845 /* Base Shader Body */
3846 shader_generate_main(shader
, buffer
, reg_maps
, function
, &priv_ctx
);
3848 if(args
->super
.srgb_correction
) {
3849 arbfp_add_sRGB_correction(buffer
, fragcolor
, srgbtmp
[0], srgbtmp
[1], srgbtmp
[2], srgbtmp
[3],
3850 priv_ctx
.target_version
>= NV2
);
3853 if(strcmp(fragcolor
, "result.color")) {
3854 shader_addline(buffer
, "MOV result.color, %s;\n", fragcolor
);
3856 shader_addline(buffer
, "END\n");
3858 /* TODO: change to resource.glObjectHandle or something like that */
3859 GL_EXTCALL(glGenProgramsARB(1, &retval
));
3861 TRACE("Creating a hw pixel shader, prg=%d\n", retval
);
3862 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, retval
));
3864 TRACE("Created hw pixel shader, prg=%d\n", retval
);
3865 /* Create the program and check for errors */
3866 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
3867 buffer
->content_size
, buffer
->buffer
));
3868 checkGLcall("glProgramStringARB()");
3870 gl_info
->gl_ops
.gl
.p_glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &errPos
);
3873 FIXME("HW PixelShader Error at position %d: %s\n\n",
3874 errPos
, debugstr_a((const char *)gl_info
->gl_ops
.gl
.p_glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
3875 shader_arb_dump_program_source(buffer
->buffer
);
3882 GL_EXTCALL(glGetProgramivARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
3883 checkGLcall("glGetProgramivARB()");
3884 if (!native
) WARN("Program exceeds native resource limits.\n");
3890 static int compare_sig(const struct wined3d_shader_signature_element
*sig1
, const struct wined3d_shader_signature_element
*sig2
)
3895 for(i
= 0; i
< MAX_REG_INPUT
; i
++)
3897 if (!sig1
[i
].semantic_name
|| !sig2
[i
].semantic_name
)
3899 /* Compare pointers, not contents. One string is NULL(element does not exist), the other one is not NULL */
3900 if(sig1
[i
].semantic_name
!= sig2
[i
].semantic_name
) return sig1
[i
].semantic_name
< sig2
[i
].semantic_name
? -1 : 1;
3904 if ((ret
= strcmp(sig1
[i
].semantic_name
, sig2
[i
].semantic_name
))) return ret
;
3905 if(sig1
[i
].semantic_idx
!= sig2
[i
].semantic_idx
) return sig1
[i
].semantic_idx
< sig2
[i
].semantic_idx
? -1 : 1;
3906 if(sig1
[i
].sysval_semantic
!= sig2
[i
].sysval_semantic
) return sig1
[i
].sysval_semantic
< sig2
[i
].sysval_semantic
? -1 : 1;
3907 if(sig1
[i
].component_type
!= sig2
[i
].component_type
) return sig1
[i
].component_type
< sig2
[i
].component_type
? -1 : 1;
3908 if(sig1
[i
].register_idx
!= sig2
[i
].register_idx
) return sig1
[i
].register_idx
< sig2
[i
].register_idx
? -1 : 1;
3909 if(sig1
[i
].mask
!= sig2
[i
].mask
) return sig1
[i
].mask
< sig2
[i
].mask
? -1 : 1;
3914 static struct wined3d_shader_signature_element
*clone_sig(const struct wined3d_shader_signature_element
*sig
)
3916 struct wined3d_shader_signature_element
*new;
3920 new = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*new) * MAX_REG_INPUT
);
3921 for(i
= 0; i
< MAX_REG_INPUT
; i
++)
3923 if (!sig
[i
].semantic_name
) continue;
3926 /* Clone the semantic string */
3927 name
= HeapAlloc(GetProcessHeap(), 0, strlen(sig
[i
].semantic_name
) + 1);
3928 strcpy(name
, sig
[i
].semantic_name
);
3929 new[i
].semantic_name
= name
;
3934 static DWORD
find_input_signature(struct shader_arb_priv
*priv
, const struct wined3d_shader_signature_element
*sig
)
3936 struct wine_rb_entry
*entry
= wine_rb_get(&priv
->signature_tree
, sig
);
3937 struct ps_signature
*found_sig
;
3941 found_sig
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
3942 TRACE("Found existing signature %u\n", found_sig
->idx
);
3943 return found_sig
->idx
;
3945 found_sig
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*sig
));
3946 found_sig
->sig
= clone_sig(sig
);
3947 found_sig
->idx
= priv
->ps_sig_number
++;
3948 TRACE("New signature stored and assigned number %u\n", found_sig
->idx
);
3949 if(wine_rb_put(&priv
->signature_tree
, sig
, &found_sig
->entry
) == -1)
3951 ERR("Failed to insert program entry.\n");
3953 return found_sig
->idx
;
3956 static void init_output_registers(const struct wined3d_shader
*shader
,
3957 const struct wined3d_shader_signature_element
*ps_input_sig
,
3958 struct shader_arb_ctx_priv
*priv_ctx
, struct arb_vs_compiled_shader
*compiled
)
3961 static const char * const texcoords
[8] =
3963 "result.texcoord[0]", "result.texcoord[1]", "result.texcoord[2]", "result.texcoord[3]",
3964 "result.texcoord[4]", "result.texcoord[5]", "result.texcoord[6]", "result.texcoord[7]"
3966 const char *semantic_name
;
3967 DWORD semantic_idx
, reg_idx
;
3969 /* Write generic input varyings 0 to 7 to result.texcoord[], varying 8 to result.color.primary
3970 * and varying 9 to result.color.secondary
3972 static const char * const decl_idx_to_string
[MAX_REG_INPUT
] =
3974 "result.texcoord[0]", "result.texcoord[1]", "result.texcoord[2]", "result.texcoord[3]",
3975 "result.texcoord[4]", "result.texcoord[5]", "result.texcoord[6]", "result.texcoord[7]",
3976 "result.color.primary", "result.color.secondary"
3981 TRACE("Pixel shader uses builtin varyings\n");
3982 /* Map builtins to builtins */
3983 for(i
= 0; i
< 8; i
++)
3985 priv_ctx
->texcrd_output
[i
] = texcoords
[i
];
3987 priv_ctx
->color_output
[0] = "result.color.primary";
3988 priv_ctx
->color_output
[1] = "result.color.secondary";
3989 priv_ctx
->fog_output
= "TMP_FOGCOORD";
3991 /* Map declared regs to builtins. Use "TA" to /dev/null unread output */
3992 for (i
= 0; i
< (sizeof(shader
->output_signature
) / sizeof(*shader
->output_signature
)); ++i
)
3994 semantic_name
= shader
->output_signature
[i
].semantic_name
;
3995 if (!semantic_name
) continue;
3997 if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_POSITION
))
3999 TRACE("o%u is TMP_OUT\n", i
);
4000 if (!shader
->output_signature
[i
].semantic_idx
) priv_ctx
->vs_output
[i
] = "TMP_OUT";
4001 else priv_ctx
->vs_output
[i
] = "TA";
4003 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_PSIZE
))
4005 TRACE("o%u is result.pointsize\n", i
);
4006 if (!shader
->output_signature
[i
].semantic_idx
) priv_ctx
->vs_output
[i
] = "result.pointsize";
4007 else priv_ctx
->vs_output
[i
] = "TA";
4009 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_COLOR
))
4011 TRACE("o%u is result.color.?, idx %u\n", i
, shader
->output_signature
[i
].semantic_idx
);
4012 if (!shader
->output_signature
[i
].semantic_idx
)
4013 priv_ctx
->vs_output
[i
] = "result.color.primary";
4014 else if (shader
->output_signature
[i
].semantic_idx
== 1)
4015 priv_ctx
->vs_output
[i
] = "result.color.secondary";
4016 else priv_ctx
->vs_output
[i
] = "TA";
4018 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
4020 TRACE("o%u is %s\n", i
, texcoords
[shader
->output_signature
[i
].semantic_idx
]);
4021 if (shader
->output_signature
[i
].semantic_idx
>= 8) priv_ctx
->vs_output
[i
] = "TA";
4022 else priv_ctx
->vs_output
[i
] = texcoords
[shader
->output_signature
[i
].semantic_idx
];
4024 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_FOG
))
4026 TRACE("o%u is result.fogcoord\n", i
);
4027 if (shader
->output_signature
[i
].semantic_idx
> 0) priv_ctx
->vs_output
[i
] = "TA";
4028 else priv_ctx
->vs_output
[i
] = "result.fogcoord";
4032 priv_ctx
->vs_output
[i
] = "TA";
4038 TRACE("Pixel shader uses declared varyings\n");
4040 /* Map builtin to declared. /dev/null the results by default to the TA temp reg */
4041 for(i
= 0; i
< 8; i
++)
4043 priv_ctx
->texcrd_output
[i
] = "TA";
4045 priv_ctx
->color_output
[0] = "TA";
4046 priv_ctx
->color_output
[1] = "TA";
4047 priv_ctx
->fog_output
= "TA";
4049 for(i
= 0; i
< MAX_REG_INPUT
; i
++)
4051 semantic_name
= ps_input_sig
[i
].semantic_name
;
4052 semantic_idx
= ps_input_sig
[i
].semantic_idx
;
4053 reg_idx
= ps_input_sig
[i
].register_idx
;
4054 if (!semantic_name
) continue;
4056 /* If a declared input register is not written by builtin arguments, don't write to it.
4057 * GL_NV_vertex_program makes sure the input defaults to 0.0, which is correct with D3D
4059 * Don't care about POSITION and PSIZE here - this is a builtin vertex shader, position goes
4060 * to TMP_OUT in any case
4062 if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
4064 if (semantic_idx
< 8)
4065 priv_ctx
->texcrd_output
[semantic_idx
] = decl_idx_to_string
[reg_idx
];
4067 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_COLOR
))
4069 if (semantic_idx
< 2)
4070 priv_ctx
->color_output
[semantic_idx
] = decl_idx_to_string
[reg_idx
];
4072 else if(shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_FOG
))
4075 priv_ctx
->fog_output
= decl_idx_to_string
[reg_idx
];
4082 if (!strcmp(decl_idx_to_string
[reg_idx
], "result.color.primary")
4083 || !strcmp(decl_idx_to_string
[reg_idx
], "result.color.secondary"))
4085 compiled
->need_color_unclamp
= TRUE
;
4089 /* Map declared to declared */
4090 for (i
= 0; i
< (sizeof(shader
->output_signature
) / sizeof(*shader
->output_signature
)); ++i
)
4092 /* Write unread output to TA to throw them away */
4093 priv_ctx
->vs_output
[i
] = "TA";
4094 semantic_name
= shader
->output_signature
[i
].semantic_name
;
4095 if (!semantic_name
) continue;
4097 if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_POSITION
)
4098 && !shader
->output_signature
[i
].semantic_idx
)
4100 priv_ctx
->vs_output
[i
] = "TMP_OUT";
4103 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_PSIZE
)
4104 && !shader
->output_signature
[i
].semantic_idx
)
4106 priv_ctx
->vs_output
[i
] = "result.pointsize";
4110 for(j
= 0; j
< MAX_REG_INPUT
; j
++)
4112 if (!ps_input_sig
[j
].semantic_name
) continue;
4114 if (!strcmp(ps_input_sig
[j
].semantic_name
, semantic_name
)
4115 && ps_input_sig
[j
].semantic_idx
== shader
->output_signature
[i
].semantic_idx
)
4117 priv_ctx
->vs_output
[i
] = decl_idx_to_string
[ps_input_sig
[j
].register_idx
];
4119 if (!strcmp(priv_ctx
->vs_output
[i
], "result.color.primary")
4120 || !strcmp(priv_ctx
->vs_output
[i
], "result.color.secondary"))
4122 compiled
->need_color_unclamp
= TRUE
;
4129 /* Context activation is done by the caller. */
4130 static GLuint
shader_arb_generate_vshader(const struct wined3d_shader
*shader
,
4131 const struct wined3d_gl_info
*gl_info
, struct wined3d_shader_buffer
*buffer
,
4132 const struct arb_vs_compile_args
*args
, struct arb_vs_compiled_shader
*compiled
,
4133 const struct wined3d_shader_signature_element
*ps_input_sig
)
4135 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
4136 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
4137 struct shader_arb_priv
*priv
= shader
->device
->shader_priv
;
4138 const DWORD
*function
= shader
->function
;
4140 DWORD next_local
= 0;
4141 struct shader_arb_ctx_priv priv_ctx
;
4145 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
4146 priv_ctx
.cur_vs_args
= args
;
4147 list_init(&priv_ctx
.control_frames
);
4148 init_output_registers(shader
, ps_input_sig
, &priv_ctx
, compiled
);
4150 /* Create the hw ARB shader */
4151 shader_addline(buffer
, "!!ARBvp1.0\n");
4153 /* Always enable the NV extension if available. Unlike fragment shaders, there is no
4154 * mesurable performance penalty, and we can always make use of it for clipplanes.
4156 if (gl_info
->supported
[NV_VERTEX_PROGRAM3
])
4158 shader_addline(buffer
, "OPTION NV_vertex_program3;\n");
4159 priv_ctx
.target_version
= NV3
;
4160 shader_addline(buffer
, "ADDRESS aL;\n");
4162 else if (gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
])
4164 shader_addline(buffer
, "OPTION NV_vertex_program2;\n");
4165 priv_ctx
.target_version
= NV2
;
4166 shader_addline(buffer
, "ADDRESS aL;\n");
4168 priv_ctx
.target_version
= ARB
;
4171 shader_addline(buffer
, "TEMP TMP_OUT;\n");
4173 shader_addline(buffer
, "TEMP TMP_FOGCOORD;\n");
4174 if (need_helper_const(shader_data
, reg_maps
, gl_info
))
4176 shader_addline(buffer
, "PARAM helper_const = { 0.0, 1.0, 2.0, %1.10f};\n", eps
);
4178 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
))
4180 shader_addline(buffer
, "PARAM rel_addr_const = { 0.5, %d.0, 0.0, 0.0 };\n", shader_data
->rel_offset
);
4181 shader_addline(buffer
, "TEMP A0_SHADOW;\n");
4184 shader_addline(buffer
, "TEMP TA;\n");
4185 shader_addline(buffer
, "TEMP TB;\n");
4187 /* Base Declarations */
4188 shader_generate_arb_declarations(shader
, reg_maps
, buffer
, gl_info
,
4189 &priv_ctx
.vs_clipplanes
, &priv_ctx
);
4191 for(i
= 0; i
< MAX_CONST_I
; i
++)
4193 compiled
->int_consts
[i
] = WINED3D_CONST_NUM_UNUSED
;
4194 if(reg_maps
->integer_constants
& (1 << i
) && priv_ctx
.target_version
>= NV2
)
4196 const DWORD
*control_values
= find_loop_control_values(shader
, i
);
4200 shader_addline(buffer
, "PARAM I%u = {%u, %u, %u, -1};\n", i
,
4201 control_values
[0], control_values
[1], control_values
[2]);
4205 compiled
->int_consts
[i
] = next_local
;
4206 compiled
->num_int_consts
++;
4207 shader_addline(buffer
, "PARAM I%u = program.local[%u];\n", i
, next_local
++);
4212 /* We need a constant to fixup the final position */
4213 shader_addline(buffer
, "PARAM posFixup = program.local[%u];\n", next_local
);
4214 compiled
->pos_fixup
= next_local
++;
4216 /* Initialize output parameters. GL_ARB_vertex_program does not require special initialization values
4217 * for output parameters. D3D in theory does not do that either, but some applications depend on a
4218 * proper initialization of the secondary color, and programs using the fixed function pipeline without
4219 * a replacement shader depend on the texcoord.w being set properly.
4221 * GL_NV_vertex_program defines that all output values are initialized to {0.0, 0.0, 0.0, 1.0}. This
4222 * assertion is in effect even when using GL_ARB_vertex_program without any NV specific additions. So
4223 * skip this if NV_vertex_program is supported. Otherwise, initialize the secondary color. For the tex-
4224 * coords, we have a flag in the opengl caps. Many cards do not require the texcoord being set, and
4225 * this can eat a number of instructions, so skip it unless this cap is set as well
4227 if (!gl_info
->supported
[NV_VERTEX_PROGRAM
])
4229 const char *color_init
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_0001
);
4230 shader_addline(buffer
, "MOV result.color.secondary, %s;\n", color_init
);
4232 if (gl_info
->quirks
& WINED3D_QUIRK_SET_TEXCOORD_W
&& !priv
->ffp_proj_control
)
4235 const char *one
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_ONE
);
4236 for(i
= 0; i
< min(8, MAX_REG_TEXCRD
); i
++)
4238 if (reg_maps
->texcoord_mask
[i
] && reg_maps
->texcoord_mask
[i
] != WINED3DSP_WRITEMASK_ALL
)
4239 shader_addline(buffer
, "MOV result.texcoord[%u].w, %s\n", i
, one
);
4244 /* The shader starts with the main function */
4245 priv_ctx
.in_main_func
= TRUE
;
4246 /* Base Shader Body */
4247 shader_generate_main(shader
, buffer
, reg_maps
, function
, &priv_ctx
);
4249 if (!priv_ctx
.footer_written
) vshader_add_footer(&priv_ctx
,
4250 shader_data
, args
, reg_maps
, gl_info
, buffer
);
4252 shader_addline(buffer
, "END\n");
4254 /* TODO: change to resource.glObjectHandle or something like that */
4255 GL_EXTCALL(glGenProgramsARB(1, &ret
));
4257 TRACE("Creating a hw vertex shader, prg=%d\n", ret
);
4258 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, ret
));
4260 TRACE("Created hw vertex shader, prg=%d\n", ret
);
4261 /* Create the program and check for errors */
4262 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
4263 buffer
->content_size
, buffer
->buffer
));
4264 checkGLcall("glProgramStringARB()");
4266 gl_info
->gl_ops
.gl
.p_glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &errPos
);
4269 FIXME("HW VertexShader Error at position %d: %s\n\n",
4270 errPos
, debugstr_a((const char *)gl_info
->gl_ops
.gl
.p_glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
4271 shader_arb_dump_program_source(buffer
->buffer
);
4278 GL_EXTCALL(glGetProgramivARB(GL_VERTEX_PROGRAM_ARB
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
4279 checkGLcall("glGetProgramivARB()");
4280 if (!native
) WARN("Program exceeds native resource limits.\n");
4286 /* Context activation is done by the caller. */
4287 static struct arb_ps_compiled_shader
*find_arb_pshader(struct wined3d_shader
*shader
,
4288 const struct arb_ps_compile_args
*args
)
4290 struct wined3d_device
*device
= shader
->device
;
4291 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
4292 const struct wined3d_d3d_info
*d3d_info
= &device
->adapter
->d3d_info
;
4295 struct arb_ps_compiled_shader
*new_array
;
4296 struct wined3d_shader_buffer buffer
;
4297 struct arb_pshader_private
*shader_data
;
4300 if (!shader
->backend_data
)
4302 struct shader_arb_priv
*priv
= device
->shader_priv
;
4304 shader
->backend_data
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
4305 shader_data
= shader
->backend_data
;
4306 shader_data
->clamp_consts
= shader
->reg_maps
.shader_version
.major
== 1;
4308 if (shader
->reg_maps
.shader_version
.major
< 3)
4309 shader_data
->input_signature_idx
= ~0U;
4311 shader_data
->input_signature_idx
= find_input_signature(priv
, shader
->input_signature
);
4313 TRACE("Shader got assigned input signature index %u\n", shader_data
->input_signature_idx
);
4315 if (!d3d_info
->vs_clipping
)
4316 shader_data
->clipplane_emulation
= shader_find_free_input_register(&shader
->reg_maps
,
4317 d3d_info
->limits
.ffp_blend_stages
- 1);
4319 shader_data
->clipplane_emulation
= ~0U;
4321 shader_data
= shader
->backend_data
;
4323 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4324 * so a linear search is more performant than a hashmap or a binary search
4325 * (cache coherency etc)
4327 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4329 if (!memcmp(&shader_data
->gl_shaders
[i
].args
, args
, sizeof(*args
)))
4330 return &shader_data
->gl_shaders
[i
];
4333 TRACE("No matching GL shader found, compiling a new shader\n");
4334 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4335 if (shader_data
->num_gl_shaders
)
4337 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4338 new_array
= HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, shader_data
->gl_shaders
,
4339 new_size
* sizeof(*shader_data
->gl_shaders
));
4341 new_array
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
->gl_shaders
));
4346 ERR("Out of memory\n");
4349 shader_data
->gl_shaders
= new_array
;
4350 shader_data
->shader_array_size
= new_size
;
4353 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4355 pixelshader_update_samplers(shader
, args
->super
.tex_types
);
4357 if (!shader_buffer_init(&buffer
))
4359 ERR("Failed to initialize shader buffer.\n");
4363 ret
= shader_arb_generate_pshader(shader
, gl_info
, &buffer
, args
,
4364 &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
]);
4365 shader_buffer_free(&buffer
);
4366 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].prgId
= ret
;
4368 return &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++];
4371 static inline BOOL
vs_args_equal(const struct arb_vs_compile_args
*stored
, const struct arb_vs_compile_args
*new,
4372 const DWORD use_map
, BOOL skip_int
) {
4373 if((stored
->super
.swizzle_map
& use_map
) != new->super
.swizzle_map
) return FALSE
;
4374 if(stored
->super
.clip_enabled
!= new->super
.clip_enabled
) return FALSE
;
4375 if(stored
->super
.fog_src
!= new->super
.fog_src
) return FALSE
;
4376 if(stored
->clip
.boolclip_compare
!= new->clip
.boolclip_compare
) return FALSE
;
4377 if(stored
->ps_signature
!= new->ps_signature
) return FALSE
;
4378 if(stored
->vertex
.samplers_compare
!= new->vertex
.samplers_compare
) return FALSE
;
4379 if(skip_int
) return TRUE
;
4381 return !memcmp(stored
->loop_ctrl
, new->loop_ctrl
, sizeof(stored
->loop_ctrl
));
4384 static struct arb_vs_compiled_shader
*find_arb_vshader(struct wined3d_shader
*shader
,
4385 const struct arb_vs_compile_args
*args
,
4386 const struct wined3d_shader_signature_element
*ps_input_sig
)
4388 struct wined3d_device
*device
= shader
->device
;
4389 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
4390 DWORD use_map
= device
->stream_info
.use_map
;
4393 struct arb_vs_compiled_shader
*new_array
;
4394 struct wined3d_shader_buffer buffer
;
4395 struct arb_vshader_private
*shader_data
;
4398 if (!shader
->backend_data
)
4400 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
4402 shader
->backend_data
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
4403 shader_data
= shader
->backend_data
;
4405 if ((gl_info
->quirks
& WINED3D_QUIRK_ARB_VS_OFFSET_LIMIT
)
4406 && reg_maps
->min_rel_offset
<= reg_maps
->max_rel_offset
)
4408 if (reg_maps
->max_rel_offset
- reg_maps
->min_rel_offset
> 127)
4410 FIXME("The difference between the minimum and maximum relative offset is > 127.\n");
4411 FIXME("Which this OpenGL implementation does not support. Try using GLSL.\n");
4412 FIXME("Min: %u, Max: %u.\n", reg_maps
->min_rel_offset
, reg_maps
->max_rel_offset
);
4414 else if (reg_maps
->max_rel_offset
- reg_maps
->min_rel_offset
> 63)
4415 shader_data
->rel_offset
= reg_maps
->min_rel_offset
+ 63;
4416 else if (reg_maps
->max_rel_offset
> 63)
4417 shader_data
->rel_offset
= reg_maps
->min_rel_offset
;
4420 shader_data
= shader
->backend_data
;
4422 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4423 * so a linear search is more performant than a hashmap or a binary search
4424 * (cache coherency etc)
4426 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
4427 if (vs_args_equal(&shader_data
->gl_shaders
[i
].args
, args
,
4428 use_map
, gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]))
4430 return &shader_data
->gl_shaders
[i
];
4434 TRACE("No matching GL shader found, compiling a new shader\n");
4436 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4437 if (shader_data
->num_gl_shaders
)
4439 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4440 new_array
= HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, shader_data
->gl_shaders
,
4441 new_size
* sizeof(*shader_data
->gl_shaders
));
4443 new_array
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
->gl_shaders
));
4448 ERR("Out of memory\n");
4451 shader_data
->gl_shaders
= new_array
;
4452 shader_data
->shader_array_size
= new_size
;
4455 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4457 if (!shader_buffer_init(&buffer
))
4459 ERR("Failed to initialize shader buffer.\n");
4463 ret
= shader_arb_generate_vshader(shader
, gl_info
, &buffer
, args
,
4464 &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
],
4466 shader_buffer_free(&buffer
);
4467 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].prgId
= ret
;
4469 return &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++];
4472 static void find_arb_ps_compile_args(const struct wined3d_state
*state
,
4473 const struct wined3d_context
*context
, const struct wined3d_shader
*shader
,
4474 struct arb_ps_compile_args
*args
)
4476 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4477 const struct wined3d_d3d_info
*d3d_info
= context
->d3d_info
;
4481 find_ps_compile_args(state
, shader
, &args
->super
);
4483 /* This forces all local boolean constants to 1 to make them stateblock independent */
4484 args
->bools
= shader
->reg_maps
.local_bool_consts
;
4486 for(i
= 0; i
< MAX_CONST_B
; i
++)
4488 if (state
->ps_consts_b
[i
])
4489 args
->bools
|= ( 1 << i
);
4492 /* Only enable the clip plane emulation KIL if at least one clipplane is enabled. The KIL instruction
4493 * is quite expensive because it forces the driver to disable early Z discards. It is cheaper to
4494 * duplicate the shader than have a no-op KIL instruction in every shader
4496 if (!d3d_info
->vs_clipping
&& use_vs(state
)
4497 && state
->render_states
[WINED3D_RS_CLIPPING
]
4498 && state
->render_states
[WINED3D_RS_CLIPPLANEENABLE
])
4503 /* Skip if unused or local, or supported natively */
4504 int_skip
= ~shader
->reg_maps
.integer_constants
| shader
->reg_maps
.local_int_consts
;
4505 if (int_skip
== 0xffff || gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
])
4507 memset(args
->loop_ctrl
, 0, sizeof(args
->loop_ctrl
));
4511 for(i
= 0; i
< MAX_CONST_I
; i
++)
4513 if(int_skip
& (1 << i
))
4515 args
->loop_ctrl
[i
][0] = 0;
4516 args
->loop_ctrl
[i
][1] = 0;
4517 args
->loop_ctrl
[i
][2] = 0;
4521 args
->loop_ctrl
[i
][0] = state
->ps_consts_i
[i
* 4];
4522 args
->loop_ctrl
[i
][1] = state
->ps_consts_i
[i
* 4 + 1];
4523 args
->loop_ctrl
[i
][2] = state
->ps_consts_i
[i
* 4 + 2];
4528 static void find_arb_vs_compile_args(const struct wined3d_state
*state
,
4529 const struct wined3d_context
*context
, const struct wined3d_shader
*shader
,
4530 struct arb_vs_compile_args
*args
)
4532 struct wined3d_device
*device
= shader
->device
;
4533 const struct wined3d_adapter
*adapter
= device
->adapter
;
4534 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4535 const struct wined3d_d3d_info
*d3d_info
= context
->d3d_info
;
4539 find_vs_compile_args(state
, shader
, &args
->super
);
4541 args
->clip
.boolclip_compare
= 0;
4544 const struct wined3d_shader
*ps
= state
->pixel_shader
;
4545 const struct arb_pshader_private
*shader_priv
= ps
->backend_data
;
4546 args
->ps_signature
= shader_priv
->input_signature_idx
;
4548 args
->clip
.boolclip
.clip_texcoord
= shader_priv
->clipplane_emulation
+ 1;
4552 args
->ps_signature
= ~0;
4553 if (!d3d_info
->vs_clipping
&& adapter
->fragment_pipe
== &arbfp_fragment_pipeline
)
4555 args
->clip
.boolclip
.clip_texcoord
= ffp_clip_emul(state
) ? d3d_info
->limits
.ffp_blend_stages
: 0;
4557 /* Otherwise: Setting boolclip_compare set clip_texcoord to 0 */
4560 if (args
->clip
.boolclip
.clip_texcoord
)
4562 if (state
->render_states
[WINED3D_RS_CLIPPING
])
4563 args
->clip
.boolclip
.clipplane_mask
= (unsigned char)state
->render_states
[WINED3D_RS_CLIPPLANEENABLE
];
4564 /* clipplane_mask was set to 0 by setting boolclip_compare to 0 */
4567 /* This forces all local boolean constants to 1 to make them stateblock independent */
4568 args
->clip
.boolclip
.bools
= shader
->reg_maps
.local_bool_consts
;
4569 /* TODO: Figure out if it would be better to store bool constants as bitmasks in the stateblock */
4570 for(i
= 0; i
< MAX_CONST_B
; i
++)
4572 if (state
->vs_consts_b
[i
])
4573 args
->clip
.boolclip
.bools
|= ( 1 << i
);
4576 args
->vertex
.samplers
[0] = device
->texUnitMap
[MAX_FRAGMENT_SAMPLERS
+ 0];
4577 args
->vertex
.samplers
[1] = device
->texUnitMap
[MAX_FRAGMENT_SAMPLERS
+ 1];
4578 args
->vertex
.samplers
[2] = device
->texUnitMap
[MAX_FRAGMENT_SAMPLERS
+ 2];
4579 args
->vertex
.samplers
[3] = 0;
4581 /* Skip if unused or local */
4582 int_skip
= ~shader
->reg_maps
.integer_constants
| shader
->reg_maps
.local_int_consts
;
4583 /* This is about flow control, not clipping. */
4584 if (int_skip
== 0xffff || gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
])
4586 memset(args
->loop_ctrl
, 0, sizeof(args
->loop_ctrl
));
4590 for(i
= 0; i
< MAX_CONST_I
; i
++)
4592 if(int_skip
& (1 << i
))
4594 args
->loop_ctrl
[i
][0] = 0;
4595 args
->loop_ctrl
[i
][1] = 0;
4596 args
->loop_ctrl
[i
][2] = 0;
4600 args
->loop_ctrl
[i
][0] = state
->vs_consts_i
[i
* 4];
4601 args
->loop_ctrl
[i
][1] = state
->vs_consts_i
[i
* 4 + 1];
4602 args
->loop_ctrl
[i
][2] = state
->vs_consts_i
[i
* 4 + 2];
4607 /* Context activation is done by the caller. */
4608 static void shader_arb_select(const struct wined3d_context
*context
, enum wined3d_shader_mode vertex_mode
,
4609 enum wined3d_shader_mode fragment_mode
)
4611 struct wined3d_device
*device
= context
->swapchain
->device
;
4612 struct shader_arb_priv
*priv
= device
->shader_priv
;
4613 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4614 const struct wined3d_state
*state
= &device
->stateBlock
->state
;
4617 /* Deal with pixel shaders first so the vertex shader arg function has the input signature ready */
4618 if (fragment_mode
== WINED3D_SHADER_MODE_SHADER
)
4620 struct wined3d_shader
*ps
= state
->pixel_shader
;
4621 struct arb_ps_compile_args compile_args
;
4622 struct arb_ps_compiled_shader
*compiled
;
4624 TRACE("Using pixel shader %p.\n", ps
);
4625 find_arb_ps_compile_args(state
, context
, ps
, &compile_args
);
4626 compiled
= find_arb_pshader(ps
, &compile_args
);
4627 priv
->current_fprogram_id
= compiled
->prgId
;
4628 priv
->compiled_fprog
= compiled
;
4630 /* Bind the fragment program */
4631 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, priv
->current_fprogram_id
));
4632 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->current_fprogram_id);");
4634 if (!priv
->use_arbfp_fixed_func
)
4635 priv
->fragment_pipe
->enable_extension(gl_info
, FALSE
);
4637 /* Enable OpenGL fragment programs. */
4638 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
4639 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
4641 TRACE("(%p) : Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n",
4642 device
, priv
->current_fprogram_id
);
4644 /* Pixel Shader 1.x constants are clamped to [-1;1], Pixel Shader 2.0 constants are not. If switching between
4645 * a 1.x and newer shader, reload the first 8 constants
4647 if (priv
->last_ps_const_clamped
!= ((struct arb_pshader_private
*)ps
->backend_data
)->clamp_consts
)
4649 priv
->last_ps_const_clamped
= ((struct arb_pshader_private
*)ps
->backend_data
)->clamp_consts
;
4650 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, 8);
4651 for(i
= 0; i
< 8; i
++)
4653 priv
->pshader_const_dirty
[i
] = 1;
4655 /* Also takes care of loading local constants */
4656 shader_arb_load_constants_internal(device
->shader_priv
, context
, state
, TRUE
, FALSE
, TRUE
);
4660 UINT rt_height
= state
->fb
->render_targets
[0]->resource
.height
;
4661 shader_arb_ps_local_constants(compiled
, context
, state
, rt_height
);
4664 /* Force constant reloading for the NP2 fixup (see comment in shader_glsl_select for more info) */
4665 if (compiled
->np2fixup_info
.super
.active
)
4666 shader_arb_load_np2fixup_constants(priv
, gl_info
, state
);
4670 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
] && !priv
->use_arbfp_fixed_func
)
4672 /* Disable only if we're not using arbfp fixed function fragment
4673 * processing. If this is used, keep GL_FRAGMENT_PROGRAM_ARB
4674 * enabled, and the fixed function pipeline will bind the fixed
4675 * function replacement shader. */
4676 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4677 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4678 priv
->current_fprogram_id
= 0;
4680 priv
->fragment_pipe
->enable_extension(gl_info
, TRUE
);
4683 if (vertex_mode
== WINED3D_SHADER_MODE_SHADER
)
4685 struct wined3d_shader
*vs
= state
->vertex_shader
;
4686 struct arb_vs_compile_args compile_args
;
4687 struct arb_vs_compiled_shader
*compiled
;
4688 const struct wined3d_shader_signature_element
*ps_input_sig
;
4690 TRACE("Using vertex shader %p\n", vs
);
4691 find_arb_vs_compile_args(state
, context
, vs
, &compile_args
);
4693 /* Instead of searching for the signature in the signature list, read the one from the
4694 * current pixel shader. It's maybe not the shader where the signature came from, but it
4695 * is the same signature and faster to find. */
4696 if (compile_args
.ps_signature
== ~0U)
4697 ps_input_sig
= NULL
;
4699 ps_input_sig
= state
->pixel_shader
->input_signature
;
4701 compiled
= find_arb_vshader(vs
, &compile_args
, ps_input_sig
);
4702 priv
->current_vprogram_id
= compiled
->prgId
;
4703 priv
->compiled_vprog
= compiled
;
4705 /* Bind the vertex program */
4706 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->current_vprogram_id
));
4707 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->current_vprogram_id);");
4709 /* Enable OpenGL vertex programs */
4710 gl_info
->gl_ops
.gl
.p_glEnable(GL_VERTEX_PROGRAM_ARB
);
4711 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
4712 TRACE("(%p) : Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", device
, priv
->current_vprogram_id
);
4713 shader_arb_vs_local_constants(compiled
, context
, state
);
4715 if(priv
->last_vs_color_unclamp
!= compiled
->need_color_unclamp
) {
4716 priv
->last_vs_color_unclamp
= compiled
->need_color_unclamp
;
4718 if (gl_info
->supported
[ARB_COLOR_BUFFER_FLOAT
])
4720 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, !compiled
->need_color_unclamp
));
4721 checkGLcall("glClampColorARB");
4723 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
4729 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
4731 priv
->current_vprogram_id
= 0;
4732 gl_info
->gl_ops
.gl
.p_glDisable(GL_VERTEX_PROGRAM_ARB
);
4733 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4735 priv
->vertex_pipe
->vp_enable(gl_info
, TRUE
);
4740 /* Context activation is done by the caller. */
4741 static void shader_arb_disable(void *shader_priv
, const struct wined3d_context
*context
)
4743 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4744 struct shader_arb_priv
*priv
= shader_priv
;
4746 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
4748 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4749 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4750 priv
->current_fprogram_id
= 0;
4752 priv
->fragment_pipe
->enable_extension(gl_info
, FALSE
);
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
, FALSE
);
4762 if (gl_info
->supported
[ARB_COLOR_BUFFER_FLOAT
] && priv
->last_vs_color_unclamp
)
4764 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, GL_FIXED_ONLY_ARB
));
4765 checkGLcall("glClampColorARB");
4766 priv
->last_vs_color_unclamp
= FALSE
;
4770 /* Context activation is done by the caller. */
4771 static void shader_arb_select_depth_blt(void *shader_priv
, const struct wined3d_gl_info
*gl_info
,
4772 enum tex_types tex_type
, const SIZE
*ds_mask_size
)
4774 const float mask
[] = {0.0f
, 0.0f
, (float)ds_mask_size
->cx
, (float)ds_mask_size
->cy
};
4775 BOOL masked
= ds_mask_size
->cx
&& ds_mask_size
->cy
;
4776 struct shader_arb_priv
*priv
= shader_priv
;
4777 GLuint
*blt_fprogram
;
4779 if (!priv
->depth_blt_vprogram_id
) priv
->depth_blt_vprogram_id
= create_arb_blt_vertex_program(gl_info
);
4780 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->depth_blt_vprogram_id
));
4781 gl_info
->gl_ops
.gl
.p_glEnable(GL_VERTEX_PROGRAM_ARB
);
4783 blt_fprogram
= masked
? &priv
->depth_blt_fprogram_id_masked
[tex_type
] : &priv
->depth_blt_fprogram_id_full
[tex_type
];
4784 if (!*blt_fprogram
) *blt_fprogram
= create_arb_blt_fragment_program(gl_info
, tex_type
, masked
);
4785 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, *blt_fprogram
));
4786 if (masked
) GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, 0, mask
));
4787 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
4790 /* Context activation is done by the caller. */
4791 static void shader_arb_deselect_depth_blt(void *shader_priv
, const struct wined3d_gl_info
*gl_info
)
4793 struct shader_arb_priv
*priv
= shader_priv
;
4795 if (priv
->current_vprogram_id
) {
4796 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->current_vprogram_id
));
4797 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
4799 TRACE("Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB.\n", priv
->current_vprogram_id
);
4803 gl_info
->gl_ops
.gl
.p_glDisable(GL_VERTEX_PROGRAM_ARB
);
4804 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4807 if (priv
->current_fprogram_id
) {
4808 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, priv
->current_fprogram_id
));
4809 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
4811 TRACE("Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB.\n", priv
->current_fprogram_id
);
4813 else if(!priv
->use_arbfp_fixed_func
)
4815 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4816 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4820 static void shader_arb_destroy(struct wined3d_shader
*shader
)
4822 struct wined3d_device
*device
= shader
->device
;
4823 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
4825 if (shader_is_pshader_version(shader
->reg_maps
.shader_version
.type
))
4827 struct arb_pshader_private
*shader_data
= shader
->backend_data
;
4830 if(!shader_data
) return; /* This can happen if a shader was never compiled */
4832 if (shader_data
->num_gl_shaders
)
4834 struct wined3d_context
*context
= context_acquire(device
, NULL
);
4836 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4838 GL_EXTCALL(glDeleteProgramsARB(1, &shader_data
->gl_shaders
[i
].prgId
));
4839 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader_data->gl_shaders[i].prgId))");
4842 context_release(context
);
4845 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
);
4846 HeapFree(GetProcessHeap(), 0, shader_data
);
4847 shader
->backend_data
= NULL
;
4851 struct arb_vshader_private
*shader_data
= shader
->backend_data
;
4854 if(!shader_data
) return; /* This can happen if a shader was never compiled */
4856 if (shader_data
->num_gl_shaders
)
4858 struct wined3d_context
*context
= context_acquire(device
, NULL
);
4860 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4862 GL_EXTCALL(glDeleteProgramsARB(1, &shader_data
->gl_shaders
[i
].prgId
));
4863 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader_data->gl_shaders[i].prgId))");
4866 context_release(context
);
4869 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
);
4870 HeapFree(GetProcessHeap(), 0, shader_data
);
4871 shader
->backend_data
= NULL
;
4875 static int sig_tree_compare(const void *key
, const struct wine_rb_entry
*entry
)
4877 struct ps_signature
*e
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
4878 return compare_sig(key
, e
->sig
);
4881 static const struct wine_rb_functions sig_tree_functions
=
4889 static HRESULT
shader_arb_alloc(struct wined3d_device
*device
, const struct wined3d_vertex_pipe_ops
*vertex_pipe
,
4890 const struct fragment_pipeline
*fragment_pipe
)
4892 struct shader_arb_priv
*priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*priv
));
4893 struct fragment_caps fragment_caps
;
4894 void *vertex_priv
, *fragment_priv
;
4895 const struct wined3d_d3d_info
*d3d_info
= &device
->adapter
->d3d_info
;
4897 if (!(vertex_priv
= vertex_pipe
->vp_alloc(&arb_program_shader_backend
, priv
)))
4899 ERR("Failed to initialize vertex pipe.\n");
4900 HeapFree(GetProcessHeap(), 0, priv
);
4904 if (!(fragment_priv
= fragment_pipe
->alloc_private(&arb_program_shader_backend
, priv
)))
4906 ERR("Failed to initialize fragment pipe.\n");
4907 vertex_pipe
->vp_free(device
);
4908 HeapFree(GetProcessHeap(), 0, priv
);
4912 priv
->vshader_const_dirty
= HeapAlloc(GetProcessHeap(), 0,
4913 sizeof(*priv
->vshader_const_dirty
) * d3d_info
->limits
.vs_uniform_count
);
4914 if (!priv
->vshader_const_dirty
)
4916 memset(priv
->vshader_const_dirty
, 1,
4917 sizeof(*priv
->vshader_const_dirty
) * d3d_info
->limits
.vs_uniform_count
);
4919 priv
->pshader_const_dirty
= HeapAlloc(GetProcessHeap(), 0,
4920 sizeof(*priv
->pshader_const_dirty
) * d3d_info
->limits
.ps_uniform_count
);
4921 if (!priv
->pshader_const_dirty
)
4923 memset(priv
->pshader_const_dirty
, 1,
4924 sizeof(*priv
->pshader_const_dirty
) * d3d_info
->limits
.ps_uniform_count
);
4926 if(wine_rb_init(&priv
->signature_tree
, &sig_tree_functions
) == -1)
4928 ERR("RB tree init failed\n");
4932 priv
->vertex_pipe
= vertex_pipe
;
4933 priv
->fragment_pipe
= fragment_pipe
;
4934 fragment_pipe
->get_caps(&device
->adapter
->gl_info
, &fragment_caps
);
4935 priv
->ffp_proj_control
= fragment_caps
.wined3d_caps
& WINED3D_FRAGMENT_CAP_PROJ_CONTROL
;
4937 device
->vertex_priv
= vertex_priv
;
4938 device
->fragment_priv
= fragment_priv
;
4939 device
->shader_priv
= priv
;
4944 HeapFree(GetProcessHeap(), 0, priv
->pshader_const_dirty
);
4945 HeapFree(GetProcessHeap(), 0, priv
->vshader_const_dirty
);
4946 fragment_pipe
->free_private(device
);
4947 vertex_pipe
->vp_free(device
);
4948 HeapFree(GetProcessHeap(), 0, priv
);
4949 return E_OUTOFMEMORY
;
4952 static void release_signature(struct wine_rb_entry
*entry
, void *context
)
4954 struct ps_signature
*sig
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
4956 for(i
= 0; i
< MAX_REG_INPUT
; i
++)
4958 HeapFree(GetProcessHeap(), 0, (char *) sig
->sig
[i
].semantic_name
);
4960 HeapFree(GetProcessHeap(), 0, sig
->sig
);
4961 HeapFree(GetProcessHeap(), 0, sig
);
4964 /* Context activation is done by the caller. */
4965 static void shader_arb_free(struct wined3d_device
*device
)
4967 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
4968 struct shader_arb_priv
*priv
= device
->shader_priv
;
4971 if (priv
->depth_blt_vprogram_id
)
4972 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_vprogram_id
));
4974 for (i
= 0; i
< tex_type_count
; ++i
)
4976 if (priv
->depth_blt_fprogram_id_full
[i
])
4978 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_fprogram_id_full
[i
]));
4980 if (priv
->depth_blt_fprogram_id_masked
[i
])
4982 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_fprogram_id_masked
[i
]));
4986 wine_rb_destroy(&priv
->signature_tree
, release_signature
, NULL
);
4987 HeapFree(GetProcessHeap(), 0, priv
->pshader_const_dirty
);
4988 HeapFree(GetProcessHeap(), 0, priv
->vshader_const_dirty
);
4989 priv
->fragment_pipe
->free_private(device
);
4990 priv
->vertex_pipe
->vp_free(device
);
4991 HeapFree(GetProcessHeap(), 0, device
->shader_priv
);
4994 static void shader_arb_context_destroyed(void *shader_priv
, const struct wined3d_context
*context
)
4996 struct shader_arb_priv
*priv
= shader_priv
;
4998 if (priv
->last_context
== context
)
4999 priv
->last_context
= NULL
;
5002 static void shader_arb_get_caps(const struct wined3d_gl_info
*gl_info
, struct shader_caps
*caps
)
5004 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
5009 /* 96 is the minimum allowed value of MAX_PROGRAM_ENV_PARAMETERS_ARB
5010 * for vertex programs. If the native limit is less than that it's
5011 * not very useful, and e.g. Mesa swrast returns 0, probably to
5012 * indicate it's a software implementation. */
5013 if (gl_info
->limits
.arb_vs_native_constants
< 96)
5014 vs_consts
= gl_info
->limits
.arb_vs_float_constants
;
5016 vs_consts
= min(gl_info
->limits
.arb_vs_float_constants
, gl_info
->limits
.arb_vs_native_constants
);
5018 if (gl_info
->supported
[NV_VERTEX_PROGRAM3
])
5021 TRACE("Hardware vertex shader version 3.0 enabled (NV_VERTEX_PROGRAM3)\n");
5023 else if (vs_consts
>= 256)
5025 /* Shader Model 2.0 requires at least 256 vertex shader constants */
5027 TRACE("Hardware vertex shader version 2.0 enabled (ARB_PROGRAM)\n");
5032 TRACE("Hardware vertex shader version 1.1 enabled (ARB_PROGRAM)\n");
5034 caps
->vs_version
= min(wined3d_settings
.max_sm_vs
, vs_version
);
5035 caps
->vs_uniform_count
= vs_consts
;
5039 caps
->vs_version
= 0;
5040 caps
->vs_uniform_count
= 0;
5043 caps
->gs_version
= 0;
5045 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
5050 /* Similar as above for vertex programs, but the minimum for fragment
5051 * programs is 24. */
5052 if (gl_info
->limits
.arb_ps_native_constants
< 24)
5053 ps_consts
= gl_info
->limits
.arb_ps_float_constants
;
5055 ps_consts
= min(gl_info
->limits
.arb_ps_float_constants
, gl_info
->limits
.arb_ps_native_constants
);
5057 if (gl_info
->supported
[NV_FRAGMENT_PROGRAM2
])
5060 TRACE("Hardware pixel shader version 3.0 enabled (NV_FRAGMENT_PROGRAM2)\n");
5062 else if (ps_consts
>= 32)
5064 /* Shader Model 2.0 requires at least 32 pixel shader constants */
5066 TRACE("Hardware pixel shader version 2.0 enabled (ARB_PROGRAM)\n");
5071 TRACE("Hardware pixel shader version 1.4 enabled (ARB_PROGRAM)\n");
5073 caps
->ps_version
= min(wined3d_settings
.max_sm_ps
, ps_version
);
5074 caps
->ps_uniform_count
= ps_consts
;
5075 caps
->ps_1x_max_value
= 8.0f
;
5079 caps
->ps_version
= 0;
5080 caps
->ps_uniform_count
= 0;
5081 caps
->ps_1x_max_value
= 0.0f
;
5084 caps
->wined3d_caps
= WINED3D_SHADER_CAP_SRGB_WRITE
;
5085 if (use_nv_clip(gl_info
))
5086 caps
->wined3d_caps
|= WINED3D_SHADER_CAP_VS_CLIPPING
;
5089 static BOOL
shader_arb_color_fixup_supported(struct color_fixup_desc fixup
)
5091 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
5093 TRACE("Checking support for color_fixup:\n");
5094 dump_color_fixup_desc(fixup
);
5097 /* We support everything except complex conversions. */
5098 if (!is_complex_fixup(fixup
))
5104 TRACE("[FAILED]\n");
5108 static void shader_arb_add_instruction_modifiers(const struct wined3d_shader_instruction
*ins
) {
5110 char write_mask
[20], regstr
[50];
5111 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
5112 BOOL is_color
= FALSE
;
5113 const struct wined3d_shader_dst_param
*dst
;
5115 if (!ins
->dst_count
) return;
5119 if (!shift
) return; /* Saturate alone is handled by the instructions */
5121 shader_arb_get_write_mask(ins
, dst
, write_mask
);
5122 shader_arb_get_register_name(ins
, &dst
->reg
, regstr
, &is_color
);
5124 /* Generate a line that does the output modifier computation
5125 * FIXME: _SAT vs shift? _SAT alone is already handled in the instructions, if this
5126 * maps problems in e.g. _d4_sat modify shader_arb_get_modifier
5128 shader_addline(buffer
, "MUL%s %s%s, %s, %s;\n", shader_arb_get_modifier(ins
),
5129 regstr
, write_mask
, regstr
, shift_tab
[shift
]);
5132 static const SHADER_HANDLER shader_arb_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
5134 /* WINED3DSIH_ABS */ shader_hw_map2gl
,
5135 /* WINED3DSIH_ADD */ shader_hw_map2gl
,
5136 /* WINED3DSIH_AND */ NULL
,
5137 /* WINED3DSIH_BEM */ pshader_hw_bem
,
5138 /* WINED3DSIH_BREAK */ shader_hw_break
,
5139 /* WINED3DSIH_BREAKC */ shader_hw_breakc
,
5140 /* WINED3DSIH_BREAKP */ NULL
,
5141 /* WINED3DSIH_CALL */ shader_hw_call
,
5142 /* WINED3DSIH_CALLNZ */ NULL
,
5143 /* WINED3DSIH_CMP */ pshader_hw_cmp
,
5144 /* WINED3DSIH_CND */ pshader_hw_cnd
,
5145 /* WINED3DSIH_CRS */ shader_hw_map2gl
,
5146 /* WINED3DSIH_CUT */ NULL
,
5147 /* WINED3DSIH_DCL */ shader_hw_nop
,
5148 /* WINED3DSIH_DCL_CONSTANT_BUFFER */ shader_hw_nop
,
5149 /* WINED3DSIH_DCL_INPUT_PRIMITIVE */ shader_hw_nop
,
5150 /* WINED3DSIH_DCL_OUTPUT_TOPOLOGY */ shader_hw_nop
,
5151 /* WINED3DSIH_DCL_VERTICES_OUT */ shader_hw_nop
,
5152 /* WINED3DSIH_DEF */ shader_hw_nop
,
5153 /* WINED3DSIH_DEFB */ shader_hw_nop
,
5154 /* WINED3DSIH_DEFI */ shader_hw_nop
,
5155 /* WINED3DSIH_DIV */ NULL
,
5156 /* WINED3DSIH_DP2ADD */ pshader_hw_dp2add
,
5157 /* WINED3DSIH_DP3 */ shader_hw_map2gl
,
5158 /* WINED3DSIH_DP4 */ shader_hw_map2gl
,
5159 /* WINED3DSIH_DST */ shader_hw_map2gl
,
5160 /* WINED3DSIH_DSX */ shader_hw_map2gl
,
5161 /* WINED3DSIH_DSY */ shader_hw_dsy
,
5162 /* WINED3DSIH_ELSE */ shader_hw_else
,
5163 /* WINED3DSIH_EMIT */ NULL
,
5164 /* WINED3DSIH_ENDIF */ shader_hw_endif
,
5165 /* WINED3DSIH_ENDLOOP */ shader_hw_endloop
,
5166 /* WINED3DSIH_ENDREP */ shader_hw_endrep
,
5167 /* WINED3DSIH_EQ */ NULL
,
5168 /* WINED3DSIH_EXP */ shader_hw_scalar_op
,
5169 /* WINED3DSIH_EXPP */ shader_hw_scalar_op
,
5170 /* WINED3DSIH_FRC */ shader_hw_map2gl
,
5171 /* WINED3DSIH_FTOI */ NULL
,
5172 /* WINED3DSIH_GE */ NULL
,
5173 /* WINED3DSIH_IADD */ NULL
,
5174 /* WINED3DSIH_IEQ */ NULL
,
5175 /* WINED3DSIH_IF */ NULL
/* Hardcoded into the shader */,
5176 /* WINED3DSIH_IFC */ shader_hw_ifc
,
5177 /* WINED3DSIH_IGE */ NULL
,
5178 /* WINED3DSIH_IMUL */ NULL
,
5179 /* WINED3DSIH_ITOF */ NULL
,
5180 /* WINED3DSIH_LABEL */ shader_hw_label
,
5181 /* WINED3DSIH_LD */ NULL
,
5182 /* WINED3DSIH_LIT */ shader_hw_map2gl
,
5183 /* WINED3DSIH_LOG */ shader_hw_log
,
5184 /* WINED3DSIH_LOGP */ shader_hw_log
,
5185 /* WINED3DSIH_LOOP */ shader_hw_loop
,
5186 /* WINED3DSIH_LRP */ shader_hw_lrp
,
5187 /* WINED3DSIH_LT */ NULL
,
5188 /* WINED3DSIH_M3x2 */ shader_hw_mnxn
,
5189 /* WINED3DSIH_M3x3 */ shader_hw_mnxn
,
5190 /* WINED3DSIH_M3x4 */ shader_hw_mnxn
,
5191 /* WINED3DSIH_M4x3 */ shader_hw_mnxn
,
5192 /* WINED3DSIH_M4x4 */ shader_hw_mnxn
,
5193 /* WINED3DSIH_MAD */ shader_hw_map2gl
,
5194 /* WINED3DSIH_MAX */ shader_hw_map2gl
,
5195 /* WINED3DSIH_MIN */ shader_hw_map2gl
,
5196 /* WINED3DSIH_MOV */ shader_hw_mov
,
5197 /* WINED3DSIH_MOVA */ shader_hw_mov
,
5198 /* WINED3DSIH_MOVC */ NULL
,
5199 /* WINED3DSIH_MUL */ shader_hw_map2gl
,
5200 /* WINED3DSIH_NOP */ shader_hw_nop
,
5201 /* WINED3DSIH_NRM */ shader_hw_nrm
,
5202 /* WINED3DSIH_PHASE */ shader_hw_nop
,
5203 /* WINED3DSIH_POW */ shader_hw_pow
,
5204 /* WINED3DSIH_RCP */ shader_hw_rcp
,
5205 /* WINED3DSIH_REP */ shader_hw_rep
,
5206 /* WINED3DSIH_RET */ shader_hw_ret
,
5207 /* WINED3DSIH_ROUND_NI */ NULL
,
5208 /* WINED3DSIH_RSQ */ shader_hw_scalar_op
,
5209 /* WINED3DSIH_SAMPLE */ NULL
,
5210 /* WINED3DSIH_SAMPLE_GRAD */ NULL
,
5211 /* WINED3DSIH_SAMPLE_LOD */ NULL
,
5212 /* WINED3DSIH_SETP */ NULL
,
5213 /* WINED3DSIH_SGE */ shader_hw_map2gl
,
5214 /* WINED3DSIH_SGN */ shader_hw_sgn
,
5215 /* WINED3DSIH_SINCOS */ shader_hw_sincos
,
5216 /* WINED3DSIH_SLT */ shader_hw_map2gl
,
5217 /* WINED3DSIH_SQRT */ NULL
,
5218 /* WINED3DSIH_SUB */ shader_hw_map2gl
,
5219 /* WINED3DSIH_TEX */ pshader_hw_tex
,
5220 /* WINED3DSIH_TEXBEM */ pshader_hw_texbem
,
5221 /* WINED3DSIH_TEXBEML */ pshader_hw_texbem
,
5222 /* WINED3DSIH_TEXCOORD */ pshader_hw_texcoord
,
5223 /* WINED3DSIH_TEXDEPTH */ pshader_hw_texdepth
,
5224 /* WINED3DSIH_TEXDP3 */ pshader_hw_texdp3
,
5225 /* WINED3DSIH_TEXDP3TEX */ pshader_hw_texdp3tex
,
5226 /* WINED3DSIH_TEXKILL */ pshader_hw_texkill
,
5227 /* WINED3DSIH_TEXLDD */ shader_hw_texldd
,
5228 /* WINED3DSIH_TEXLDL */ shader_hw_texldl
,
5229 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_hw_texm3x2depth
,
5230 /* WINED3DSIH_TEXM3x2PAD */ pshader_hw_texm3x2pad
,
5231 /* WINED3DSIH_TEXM3x2TEX */ pshader_hw_texm3x2tex
,
5232 /* WINED3DSIH_TEXM3x3 */ pshader_hw_texm3x3
,
5233 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
5234 /* WINED3DSIH_TEXM3x3PAD */ pshader_hw_texm3x3pad
,
5235 /* WINED3DSIH_TEXM3x3SPEC */ pshader_hw_texm3x3spec
,
5236 /* WINED3DSIH_TEXM3x3TEX */ pshader_hw_texm3x3tex
,
5237 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_hw_texm3x3vspec
,
5238 /* WINED3DSIH_TEXREG2AR */ pshader_hw_texreg2ar
,
5239 /* WINED3DSIH_TEXREG2GB */ pshader_hw_texreg2gb
,
5240 /* WINED3DSIH_TEXREG2RGB */ pshader_hw_texreg2rgb
,
5241 /* WINED3DSIH_UDIV */ NULL
,
5242 /* WINED3DSIH_USHR */ NULL
,
5243 /* WINED3DSIH_UTOF */ NULL
,
5244 /* WINED3DSIH_XOR */ NULL
,
5247 static BOOL
get_bool_const(const struct wined3d_shader_instruction
*ins
,
5248 const struct wined3d_shader
*shader
, DWORD idx
)
5250 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
5251 BOOL vshader
= shader_is_vshader_version(reg_maps
->shader_version
.type
);
5252 const struct wined3d_shader_lconst
*constant
;
5254 WORD flag
= (1 << idx
);
5255 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5257 if (reg_maps
->local_bool_consts
& flag
)
5259 /* What good is a if(bool) with a hardcoded local constant? I don't know, but handle it */
5260 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsB
, struct wined3d_shader_lconst
, entry
)
5262 if (constant
->idx
== idx
)
5264 return constant
->value
[0];
5267 ERR("Local constant not found\n");
5272 if(vshader
) bools
= priv
->cur_vs_args
->clip
.boolclip
.bools
;
5273 else bools
= priv
->cur_ps_args
->bools
;
5274 return bools
& flag
;
5278 static void get_loop_control_const(const struct wined3d_shader_instruction
*ins
,
5279 const struct wined3d_shader
*shader
, UINT idx
, struct wined3d_shader_loop_control
*loop_control
)
5281 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
5282 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5284 /* Integer constants can either be a local constant, or they can be stored in the shader
5285 * type specific compile args. */
5286 if (reg_maps
->local_int_consts
& (1 << idx
))
5288 const struct wined3d_shader_lconst
*constant
;
5290 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsI
, struct wined3d_shader_lconst
, entry
)
5292 if (constant
->idx
== idx
)
5294 loop_control
->count
= constant
->value
[0];
5295 loop_control
->start
= constant
->value
[1];
5296 /* Step is signed. */
5297 loop_control
->step
= (int)constant
->value
[2];
5301 /* If this happens the flag was set incorrectly */
5302 ERR("Local constant not found\n");
5303 loop_control
->count
= 0;
5304 loop_control
->start
= 0;
5305 loop_control
->step
= 0;
5309 switch (reg_maps
->shader_version
.type
)
5311 case WINED3D_SHADER_TYPE_VERTEX
:
5312 /* Count and aL start value are unsigned */
5313 loop_control
->count
= priv
->cur_vs_args
->loop_ctrl
[idx
][0];
5314 loop_control
->start
= priv
->cur_vs_args
->loop_ctrl
[idx
][1];
5315 /* Step is signed. */
5316 loop_control
->step
= ((char)priv
->cur_vs_args
->loop_ctrl
[idx
][2]);
5319 case WINED3D_SHADER_TYPE_PIXEL
:
5320 loop_control
->count
= priv
->cur_ps_args
->loop_ctrl
[idx
][0];
5321 loop_control
->start
= priv
->cur_ps_args
->loop_ctrl
[idx
][1];
5322 loop_control
->step
= ((char)priv
->cur_ps_args
->loop_ctrl
[idx
][2]);
5326 FIXME("Unhandled shader type %#x.\n", reg_maps
->shader_version
.type
);
5331 static void record_instruction(struct list
*list
, const struct wined3d_shader_instruction
*ins
)
5334 struct wined3d_shader_dst_param
*dst_param
= NULL
;
5335 struct wined3d_shader_src_param
*src_param
= NULL
, *rel_addr
= NULL
;
5336 struct recorded_instruction
*rec
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*rec
));
5339 ERR("Out of memory\n");
5344 dst_param
= HeapAlloc(GetProcessHeap(), 0, sizeof(*dst_param
));
5345 if(!dst_param
) goto free
;
5346 *dst_param
= *ins
->dst
;
5347 if (ins
->dst
->reg
.idx
[0].rel_addr
)
5349 rel_addr
= HeapAlloc(GetProcessHeap(), 0, sizeof(*rel_addr
));
5352 *rel_addr
= *ins
->dst
->reg
.idx
[0].rel_addr
;
5353 dst_param
->reg
.idx
[0].rel_addr
= rel_addr
;
5355 rec
->ins
.dst
= dst_param
;
5357 src_param
= HeapAlloc(GetProcessHeap(), 0, sizeof(*src_param
) * ins
->src_count
);
5360 for (i
= 0; i
< ins
->src_count
; ++i
)
5362 src_param
[i
] = ins
->src
[i
];
5363 if (ins
->src
[i
].reg
.idx
[0].rel_addr
)
5365 rel_addr
= HeapAlloc(GetProcessHeap(), 0, sizeof(*rel_addr
));
5368 *rel_addr
= *ins
->src
[i
].reg
.idx
[0].rel_addr
;
5369 src_param
[i
].reg
.idx
[0].rel_addr
= rel_addr
;
5372 rec
->ins
.src
= src_param
;
5373 list_add_tail(list
, &rec
->entry
);
5377 ERR("Out of memory\n");
5380 HeapFree(GetProcessHeap(), 0, (void *)dst_param
->reg
.idx
[0].rel_addr
);
5381 HeapFree(GetProcessHeap(), 0, dst_param
);
5385 for(i
= 0; i
< ins
->src_count
; i
++)
5387 HeapFree(GetProcessHeap(), 0, (void *)src_param
[i
].reg
.idx
[0].rel_addr
);
5389 HeapFree(GetProcessHeap(), 0, src_param
);
5391 HeapFree(GetProcessHeap(), 0, rec
);
5394 static void free_recorded_instruction(struct list
*list
)
5396 struct recorded_instruction
*rec_ins
, *entry2
;
5399 LIST_FOR_EACH_ENTRY_SAFE(rec_ins
, entry2
, list
, struct recorded_instruction
, entry
)
5401 list_remove(&rec_ins
->entry
);
5402 if (rec_ins
->ins
.dst
)
5404 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.dst
->reg
.idx
[0].rel_addr
);
5405 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.dst
);
5407 if (rec_ins
->ins
.src
)
5409 for (i
= 0; i
< rec_ins
->ins
.src_count
; ++i
)
5411 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.src
[i
].reg
.idx
[0].rel_addr
);
5413 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.src
);
5415 HeapFree(GetProcessHeap(), 0, rec_ins
);
5419 static void pop_control_frame(const struct wined3d_shader_instruction
*ins
)
5421 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5422 struct control_frame
*control_frame
;
5424 if (ins
->handler_idx
== WINED3DSIH_ENDLOOP
|| ins
->handler_idx
== WINED3DSIH_ENDREP
)
5426 struct list
*e
= list_head(&priv
->control_frames
);
5427 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5428 list_remove(&control_frame
->entry
);
5429 HeapFree(GetProcessHeap(), 0, control_frame
);
5432 else if (ins
->handler_idx
== WINED3DSIH_ENDIF
)
5434 /* Non-ifc ENDIFs were already handled previously. */
5435 struct list
*e
= list_head(&priv
->control_frames
);
5436 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5437 list_remove(&control_frame
->entry
);
5438 HeapFree(GetProcessHeap(), 0, control_frame
);
5442 static void shader_arb_handle_instruction(const struct wined3d_shader_instruction
*ins
) {
5443 SHADER_HANDLER hw_fct
;
5444 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5445 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
5446 struct control_frame
*control_frame
;
5447 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
5450 if(ins
->handler_idx
== WINED3DSIH_LOOP
|| ins
->handler_idx
== WINED3DSIH_REP
)
5452 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
5453 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5455 if(ins
->handler_idx
== WINED3DSIH_LOOP
) control_frame
->type
= LOOP
;
5456 if(ins
->handler_idx
== WINED3DSIH_REP
) control_frame
->type
= REP
;
5458 if(priv
->target_version
>= NV2
)
5460 control_frame
->no
.loop
= priv
->num_loops
++;
5465 /* Don't bother recording when we're in a not used if branch */
5471 if(!priv
->recording
)
5473 list_init(&priv
->record
);
5474 priv
->recording
= TRUE
;
5475 control_frame
->outer_loop
= TRUE
;
5476 get_loop_control_const(ins
, shader
, ins
->src
[0].reg
.idx
[0].offset
, &control_frame
->loop_control
);
5477 return; /* Instruction is handled */
5479 /* Record this loop in the outer loop's recording */
5482 else if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
|| ins
->handler_idx
== WINED3DSIH_ENDREP
)
5484 if(priv
->target_version
>= NV2
)
5486 /* Nothing to do. The control frame is popped after the HW instr handler */
5490 struct list
*e
= list_head(&priv
->control_frames
);
5491 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5492 list_remove(&control_frame
->entry
);
5494 if(control_frame
->outer_loop
)
5496 unsigned int iteration
;
5500 /* Turn off recording before playback */
5501 priv
->recording
= FALSE
;
5503 /* Move the recorded instructions to a separate list and get them out of the private data
5504 * structure. If there are nested loops, the shader_arb_handle_instruction below will
5505 * be recorded again, thus priv->record might be overwritten
5508 list_move_tail(©
, &priv
->record
);
5509 list_init(&priv
->record
);
5511 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5513 shader_addline(buffer
, "#unrolling loop: %u iterations, aL=%u, inc %d\n",
5514 control_frame
->loop_control
.count
, control_frame
->loop_control
.start
,
5515 control_frame
->loop_control
.step
);
5516 aL
= control_frame
->loop_control
.start
;
5520 shader_addline(buffer
, "#unrolling rep: %u iterations\n", control_frame
->loop_control
.count
);
5523 for (iteration
= 0; iteration
< control_frame
->loop_control
.count
; ++iteration
)
5525 struct recorded_instruction
*rec_ins
;
5526 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5529 shader_addline(buffer
, "#Iteration %u, aL=%d\n", iteration
, aL
);
5533 shader_addline(buffer
, "#Iteration %u\n", iteration
);
5536 LIST_FOR_EACH_ENTRY(rec_ins
, ©
, struct recorded_instruction
, entry
)
5538 shader_arb_handle_instruction(&rec_ins
->ins
);
5541 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5543 aL
+= control_frame
->loop_control
.step
;
5546 shader_addline(buffer
, "#end loop/rep\n");
5548 free_recorded_instruction(©
);
5549 HeapFree(GetProcessHeap(), 0, control_frame
);
5550 return; /* Instruction is handled */
5554 /* This is a nested loop. Proceed to the normal recording function */
5555 HeapFree(GetProcessHeap(), 0, control_frame
);
5562 record_instruction(&priv
->record
, ins
);
5567 if(ins
->handler_idx
== WINED3DSIH_IF
)
5569 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
5570 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5571 control_frame
->type
= IF
;
5573 bool_const
= get_bool_const(ins
, shader
, ins
->src
[0].reg
.idx
[0].offset
);
5574 if (ins
->src
[0].modifiers
== WINED3DSPSM_NOT
)
5575 bool_const
= !bool_const
;
5576 if (!priv
->muted
&& !bool_const
)
5578 shader_addline(buffer
, "#if(FALSE){\n");
5580 control_frame
->muting
= TRUE
;
5582 else shader_addline(buffer
, "#if(TRUE) {\n");
5584 return; /* Instruction is handled */
5586 else if(ins
->handler_idx
== WINED3DSIH_IFC
)
5588 /* IF(bool) and if_cond(a, b) use the same ELSE and ENDIF tokens */
5589 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
5590 control_frame
->type
= IFC
;
5591 control_frame
->no
.ifc
= priv
->num_ifcs
++;
5592 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5594 else if(ins
->handler_idx
== WINED3DSIH_ELSE
)
5596 struct list
*e
= list_head(&priv
->control_frames
);
5597 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5599 if(control_frame
->type
== IF
)
5601 shader_addline(buffer
, "#} else {\n");
5602 if(!priv
->muted
&& !control_frame
->muting
)
5605 control_frame
->muting
= TRUE
;
5607 else if(control_frame
->muting
) priv
->muted
= FALSE
;
5608 return; /* Instruction is handled. */
5610 /* In case of an ifc, generate a HW shader instruction */
5611 if (control_frame
->type
!= IFC
)
5612 ERR("Control frame does not match.\n");
5614 else if(ins
->handler_idx
== WINED3DSIH_ENDIF
)
5616 struct list
*e
= list_head(&priv
->control_frames
);
5617 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5619 if(control_frame
->type
== IF
)
5621 shader_addline(buffer
, "#} endif\n");
5622 if(control_frame
->muting
) priv
->muted
= FALSE
;
5623 list_remove(&control_frame
->entry
);
5624 HeapFree(GetProcessHeap(), 0, control_frame
);
5625 return; /* Instruction is handled */
5627 /* In case of an ifc, generate a HW shader instruction */
5628 if (control_frame
->type
!= IFC
)
5629 ERR("Control frame does not match.\n");
5634 pop_control_frame(ins
);
5638 /* Select handler */
5639 hw_fct
= shader_arb_instruction_handler_table
[ins
->handler_idx
];
5641 /* Unhandled opcode */
5644 FIXME("Backend can't handle opcode %#x\n", ins
->handler_idx
);
5649 pop_control_frame(ins
);
5651 shader_arb_add_instruction_modifiers(ins
);
5654 static BOOL
shader_arb_has_ffp_proj_control(void *shader_priv
)
5656 struct shader_arb_priv
*priv
= shader_priv
;
5658 return priv
->ffp_proj_control
;
5661 const struct wined3d_shader_backend_ops arb_program_shader_backend
=
5663 shader_arb_handle_instruction
,
5666 shader_arb_select_depth_blt
,
5667 shader_arb_deselect_depth_blt
,
5668 shader_arb_update_float_vertex_constants
,
5669 shader_arb_update_float_pixel_constants
,
5670 shader_arb_load_constants
,
5671 shader_arb_load_np2fixup_constants
,
5675 shader_arb_context_destroyed
,
5676 shader_arb_get_caps
,
5677 shader_arb_color_fixup_supported
,
5678 shader_arb_has_ffp_proj_control
,
5681 /* ARB_fragment_program fixed function pipeline replacement definitions */
5682 #define ARB_FFP_CONST_TFACTOR 0
5683 #define ARB_FFP_CONST_SPECULAR_ENABLE ((ARB_FFP_CONST_TFACTOR) + 1)
5684 #define ARB_FFP_CONST_CONSTANT(i) ((ARB_FFP_CONST_SPECULAR_ENABLE) + 1 + i)
5685 #define ARB_FFP_CONST_BUMPMAT(i) ((ARB_FFP_CONST_CONSTANT(7)) + 1 + i)
5686 #define ARB_FFP_CONST_LUMINANCE(i) ((ARB_FFP_CONST_BUMPMAT(7)) + 1 + i)
5688 struct arbfp_ffp_desc
5690 struct ffp_frag_desc parent
;
5694 /* Context activation is done by the caller. */
5695 static void arbfp_enable(const struct wined3d_gl_info
*gl_info
, BOOL enable
)
5699 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
5700 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
5704 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
5705 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
5709 static void *arbfp_alloc(const struct wined3d_shader_backend_ops
*shader_backend
, void *shader_priv
)
5711 struct shader_arb_priv
*priv
;
5713 /* Share private data between the shader backend and the pipeline
5714 * replacement, if both are the arb implementation. This is needed to
5715 * figure out whether ARBfp should be disabled if no pixel shader is bound
5717 if (shader_backend
== &arb_program_shader_backend
)
5719 else if (!(priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*priv
))))
5722 if (wine_rb_init(&priv
->fragment_shaders
, &wined3d_ffp_frag_program_rb_functions
) == -1)
5724 ERR("Failed to initialize rbtree.\n");
5725 if (priv
!= shader_priv
)
5726 HeapFree(GetProcessHeap(), 0, priv
);
5729 priv
->use_arbfp_fixed_func
= TRUE
;
5734 /* Context activation is done by the caller. */
5735 static void arbfp_free_ffpshader(struct wine_rb_entry
*entry
, void *context
)
5737 const struct wined3d_gl_info
*gl_info
= context
;
5738 struct arbfp_ffp_desc
*entry_arb
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_ffp_desc
, parent
.entry
);
5740 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb
->shader
));
5741 checkGLcall("glDeleteProgramsARB(1, &entry_arb->shader)");
5742 HeapFree(GetProcessHeap(), 0, entry_arb
);
5745 /* Context activation is done by the caller. */
5746 static void arbfp_free(struct wined3d_device
*device
)
5748 struct shader_arb_priv
*priv
= device
->fragment_priv
;
5750 wine_rb_destroy(&priv
->fragment_shaders
, arbfp_free_ffpshader
, &device
->adapter
->gl_info
);
5751 priv
->use_arbfp_fixed_func
= FALSE
;
5753 if (device
->shader_backend
!= &arb_program_shader_backend
)
5755 HeapFree(GetProcessHeap(), 0, device
->fragment_priv
);
5759 static void arbfp_get_caps(const struct wined3d_gl_info
*gl_info
, struct fragment_caps
*caps
)
5761 caps
->wined3d_caps
= WINED3D_FRAGMENT_CAP_PROJ_CONTROL
5762 | WINED3D_FRAGMENT_CAP_SRGB_WRITE
;
5763 caps
->PrimitiveMiscCaps
= WINED3DPMISCCAPS_TSSARGTEMP
;
5764 caps
->TextureOpCaps
= WINED3DTEXOPCAPS_DISABLE
|
5765 WINED3DTEXOPCAPS_SELECTARG1
|
5766 WINED3DTEXOPCAPS_SELECTARG2
|
5767 WINED3DTEXOPCAPS_MODULATE4X
|
5768 WINED3DTEXOPCAPS_MODULATE2X
|
5769 WINED3DTEXOPCAPS_MODULATE
|
5770 WINED3DTEXOPCAPS_ADDSIGNED2X
|
5771 WINED3DTEXOPCAPS_ADDSIGNED
|
5772 WINED3DTEXOPCAPS_ADD
|
5773 WINED3DTEXOPCAPS_SUBTRACT
|
5774 WINED3DTEXOPCAPS_ADDSMOOTH
|
5775 WINED3DTEXOPCAPS_BLENDCURRENTALPHA
|
5776 WINED3DTEXOPCAPS_BLENDFACTORALPHA
|
5777 WINED3DTEXOPCAPS_BLENDTEXTUREALPHA
|
5778 WINED3DTEXOPCAPS_BLENDDIFFUSEALPHA
|
5779 WINED3DTEXOPCAPS_BLENDTEXTUREALPHAPM
|
5780 WINED3DTEXOPCAPS_MODULATEALPHA_ADDCOLOR
|
5781 WINED3DTEXOPCAPS_MODULATECOLOR_ADDALPHA
|
5782 WINED3DTEXOPCAPS_MODULATEINVCOLOR_ADDALPHA
|
5783 WINED3DTEXOPCAPS_MODULATEINVALPHA_ADDCOLOR
|
5784 WINED3DTEXOPCAPS_DOTPRODUCT3
|
5785 WINED3DTEXOPCAPS_MULTIPLYADD
|
5786 WINED3DTEXOPCAPS_LERP
|
5787 WINED3DTEXOPCAPS_BUMPENVMAP
|
5788 WINED3DTEXOPCAPS_BUMPENVMAPLUMINANCE
;
5790 /* TODO: Implement WINED3DTEXOPCAPS_PREMODULATE */
5792 caps
->MaxTextureBlendStages
= 8;
5793 caps
->MaxSimultaneousTextures
= min(gl_info
->limits
.fragment_samplers
, 8);
5796 static void state_texfactor_arbfp(struct wined3d_context
*context
,
5797 const struct wined3d_state
*state
, DWORD state_id
)
5799 struct wined3d_device
*device
= context
->swapchain
->device
;
5800 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5803 /* Don't load the parameter if we're using an arbfp pixel shader,
5804 * otherwise we'll overwrite application provided constants. */
5805 if (device
->shader_backend
== &arb_program_shader_backend
)
5807 struct shader_arb_priv
*priv
;
5809 if (use_ps(state
)) return;
5811 priv
= device
->shader_priv
;
5812 priv
->pshader_const_dirty
[ARB_FFP_CONST_TFACTOR
] = 1;
5813 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_TFACTOR
+ 1);
5816 D3DCOLORTOGLFLOAT4(state
->render_states
[WINED3D_RS_TEXTUREFACTOR
], col
);
5817 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_TFACTOR
, col
));
5818 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_TFACTOR, col)");
5821 static void state_arb_specularenable(struct wined3d_context
*context
,
5822 const struct wined3d_state
*state
, DWORD state_id
)
5824 struct wined3d_device
*device
= context
->swapchain
->device
;
5825 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5828 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
5829 * application provided constants
5831 if (device
->shader_backend
== &arb_program_shader_backend
)
5833 struct shader_arb_priv
*priv
;
5835 if (use_ps(state
)) return;
5837 priv
= device
->shader_priv
;
5838 priv
->pshader_const_dirty
[ARB_FFP_CONST_SPECULAR_ENABLE
] = 1;
5839 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_SPECULAR_ENABLE
+ 1);
5842 if (state
->render_states
[WINED3D_RS_SPECULARENABLE
])
5844 /* The specular color has no alpha */
5845 col
[0] = 1.0f
; col
[1] = 1.0f
;
5846 col
[2] = 1.0f
; col
[3] = 0.0f
;
5848 col
[0] = 0.0f
; col
[1] = 0.0f
;
5849 col
[2] = 0.0f
; col
[3] = 0.0f
;
5851 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_SPECULAR_ENABLE
, col
));
5852 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_SPECULAR_ENABLE, col)");
5855 static void set_bumpmat_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
5857 DWORD stage
= (state_id
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
5858 struct wined3d_device
*device
= context
->swapchain
->device
;
5859 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5864 /* The pixel shader has to know the bump env matrix. Do a constants
5866 if (stage
&& (state
->pixel_shader
->reg_maps
.bumpmat
& (1 << stage
)))
5867 context
->load_constants
= 1;
5869 if(device
->shader_backend
== &arb_program_shader_backend
) {
5870 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants */
5874 else if (device
->shader_backend
== &arb_program_shader_backend
)
5876 struct shader_arb_priv
*priv
= device
->shader_priv
;
5877 priv
->pshader_const_dirty
[ARB_FFP_CONST_BUMPMAT(stage
)] = 1;
5878 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_BUMPMAT(stage
) + 1);
5881 mat
[0][0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT00
]);
5882 mat
[0][1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT01
]);
5883 mat
[1][0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT10
]);
5884 mat
[1][1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT11
]);
5886 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_BUMPMAT(stage
), &mat
[0][0]));
5887 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_BUMPMAT(stage), &mat[0][0])");
5890 static void tex_bumpenvlum_arbfp(struct wined3d_context
*context
,
5891 const struct wined3d_state
*state
, DWORD state_id
)
5893 DWORD stage
= (state_id
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
5894 struct wined3d_device
*device
= context
->swapchain
->device
;
5895 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5900 /* The pixel shader has to know the luminance offset. Do a constants
5902 if (stage
&& (state
->pixel_shader
->reg_maps
.luminanceparams
& (1 << stage
)))
5903 context
->load_constants
= 1;
5905 if(device
->shader_backend
== &arb_program_shader_backend
) {
5906 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants */
5910 else if (device
->shader_backend
== &arb_program_shader_backend
)
5912 struct shader_arb_priv
*priv
= device
->shader_priv
;
5913 priv
->pshader_const_dirty
[ARB_FFP_CONST_LUMINANCE(stage
)] = 1;
5914 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_LUMINANCE(stage
) + 1);
5917 param
[0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_LSCALE
]);
5918 param
[1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_LOFFSET
]);
5922 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_LUMINANCE(stage
), param
));
5923 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_LUMINANCE(stage), param)");
5926 static const char *get_argreg(struct wined3d_shader_buffer
*buffer
, DWORD argnum
, unsigned int stage
, DWORD arg
)
5930 if(arg
== ARG_UNUSED
) return "unused"; /* This is the marker for unused registers */
5932 switch(arg
& WINED3DTA_SELECTMASK
) {
5933 case WINED3DTA_DIFFUSE
:
5934 ret
= "fragment.color.primary"; break;
5936 case WINED3DTA_CURRENT
:
5937 if (!stage
) ret
= "fragment.color.primary";
5941 case WINED3DTA_TEXTURE
:
5943 case 0: ret
= "tex0"; break;
5944 case 1: ret
= "tex1"; break;
5945 case 2: ret
= "tex2"; break;
5946 case 3: ret
= "tex3"; break;
5947 case 4: ret
= "tex4"; break;
5948 case 5: ret
= "tex5"; break;
5949 case 6: ret
= "tex6"; break;
5950 case 7: ret
= "tex7"; break;
5951 default: ret
= "unknown texture";
5955 case WINED3DTA_TFACTOR
:
5956 ret
= "tfactor"; break;
5958 case WINED3DTA_SPECULAR
:
5959 ret
= "fragment.color.secondary"; break;
5961 case WINED3DTA_TEMP
:
5962 ret
= "tempreg"; break;
5964 case WINED3DTA_CONSTANT
:
5965 FIXME("Implement perstage constants\n");
5967 case 0: ret
= "const0"; break;
5968 case 1: ret
= "const1"; break;
5969 case 2: ret
= "const2"; break;
5970 case 3: ret
= "const3"; break;
5971 case 4: ret
= "const4"; break;
5972 case 5: ret
= "const5"; break;
5973 case 6: ret
= "const6"; break;
5974 case 7: ret
= "const7"; break;
5975 default: ret
= "unknown constant";
5983 if(arg
& WINED3DTA_COMPLEMENT
) {
5984 shader_addline(buffer
, "SUB arg%u, const.x, %s;\n", argnum
, ret
);
5985 if(argnum
== 0) ret
= "arg0";
5986 if(argnum
== 1) ret
= "arg1";
5987 if(argnum
== 2) ret
= "arg2";
5989 if(arg
& WINED3DTA_ALPHAREPLICATE
) {
5990 shader_addline(buffer
, "MOV arg%u, %s.w;\n", argnum
, ret
);
5991 if(argnum
== 0) ret
= "arg0";
5992 if(argnum
== 1) ret
= "arg1";
5993 if(argnum
== 2) ret
= "arg2";
5998 static void gen_ffp_instr(struct wined3d_shader_buffer
*buffer
, unsigned int stage
, BOOL color
,
5999 BOOL alpha
, DWORD dst
, DWORD op
, DWORD dw_arg0
, DWORD dw_arg1
, DWORD dw_arg2
)
6001 const char *dstmask
, *dstreg
, *arg0
, *arg1
, *arg2
;
6002 unsigned int mul
= 1;
6004 if(color
&& alpha
) dstmask
= "";
6005 else if(color
) dstmask
= ".xyz";
6006 else dstmask
= ".w";
6008 if(dst
== tempreg
) dstreg
= "tempreg";
6009 else dstreg
= "ret";
6011 arg0
= get_argreg(buffer
, 0, stage
, dw_arg0
);
6012 arg1
= get_argreg(buffer
, 1, stage
, dw_arg1
);
6013 arg2
= get_argreg(buffer
, 2, stage
, dw_arg2
);
6017 case WINED3D_TOP_DISABLE
:
6019 shader_addline(buffer
, "MOV %s%s, fragment.color.primary;\n", dstreg
, dstmask
);
6022 case WINED3D_TOP_SELECT_ARG2
:
6025 case WINED3D_TOP_SELECT_ARG1
:
6026 shader_addline(buffer
, "MOV %s%s, %s;\n", dstreg
, dstmask
, arg1
);
6029 case WINED3D_TOP_MODULATE_4X
:
6032 case WINED3D_TOP_MODULATE_2X
:
6035 case WINED3D_TOP_MODULATE
:
6036 shader_addline(buffer
, "MUL %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6039 case WINED3D_TOP_ADD_SIGNED_2X
:
6042 case WINED3D_TOP_ADD_SIGNED
:
6043 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
6046 case WINED3D_TOP_ADD
:
6047 shader_addline(buffer
, "ADD_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6050 case WINED3D_TOP_SUBTRACT
:
6051 shader_addline(buffer
, "SUB_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6054 case WINED3D_TOP_ADD_SMOOTH
:
6055 shader_addline(buffer
, "SUB arg1, const.x, %s;\n", arg1
);
6056 shader_addline(buffer
, "MAD_SAT %s%s, arg1, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6059 case WINED3D_TOP_BLEND_CURRENT_ALPHA
:
6060 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_CURRENT
);
6061 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6063 case WINED3D_TOP_BLEND_FACTOR_ALPHA
:
6064 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TFACTOR
);
6065 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6067 case WINED3D_TOP_BLEND_TEXTURE_ALPHA
:
6068 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
6069 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6071 case WINED3D_TOP_BLEND_DIFFUSE_ALPHA
:
6072 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_DIFFUSE
);
6073 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6076 case WINED3D_TOP_BLEND_TEXTURE_ALPHA_PM
:
6077 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
6078 shader_addline(buffer
, "SUB arg0.w, const.x, %s.w;\n", arg0
);
6079 shader_addline(buffer
, "MAD_SAT %s%s, %s, arg0.w, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6082 /* D3DTOP_PREMODULATE ???? */
6084 case WINED3D_TOP_MODULATE_INVALPHA_ADD_COLOR
:
6085 shader_addline(buffer
, "SUB arg0.w, const.x, %s;\n", arg1
);
6086 shader_addline(buffer
, "MAD_SAT %s%s, arg0.w, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6088 case WINED3D_TOP_MODULATE_ALPHA_ADD_COLOR
:
6089 shader_addline(buffer
, "MAD_SAT %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
6091 case WINED3D_TOP_MODULATE_INVCOLOR_ADD_ALPHA
:
6092 shader_addline(buffer
, "SUB arg0, const.x, %s;\n", arg1
);
6093 shader_addline(buffer
, "MAD_SAT %s%s, arg0, %s, %s.w;\n", dstreg
, dstmask
, arg2
, arg1
);
6095 case WINED3D_TOP_MODULATE_COLOR_ADD_ALPHA
:
6096 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s.w;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
6099 case WINED3D_TOP_DOTPRODUCT3
:
6101 shader_addline(buffer
, "SUB arg1, %s, const.w;\n", arg1
);
6102 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
6103 shader_addline(buffer
, "DP3_SAT %s%s, arg1, arg2;\n", dstreg
, dstmask
);
6106 case WINED3D_TOP_MULTIPLY_ADD
:
6107 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg0
);
6110 case WINED3D_TOP_LERP
:
6111 /* The msdn is not quite right here */
6112 shader_addline(buffer
, "LRP %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6115 case WINED3D_TOP_BUMPENVMAP
:
6116 case WINED3D_TOP_BUMPENVMAP_LUMINANCE
:
6117 /* Those are handled in the first pass of the shader(generation pass 1 and 2) already */
6121 FIXME("Unhandled texture op %08x\n", op
);
6125 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.y;\n", dstreg
, dstmask
, dstreg
);
6127 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.z;\n", dstreg
, dstmask
, dstreg
);
6130 static GLuint
gen_arbfp_ffp_shader(const struct ffp_frag_settings
*settings
, const struct wined3d_gl_info
*gl_info
)
6133 struct wined3d_shader_buffer buffer
;
6134 BOOL tex_read
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
6135 BOOL bump_used
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
6136 BOOL luminance_used
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
6137 UINT lowest_disabled_stage
;
6138 const char *textype
;
6139 const char *instr
, *sat
;
6140 char colorcor_dst
[8];
6142 DWORD arg0
, arg1
, arg2
;
6143 BOOL tempreg_used
= FALSE
, tfactor_used
= FALSE
;
6145 const char *final_combiner_src
= "ret";
6148 /* Find out which textures are read */
6149 for (stage
= 0; stage
< MAX_TEXTURES
; ++stage
)
6151 if (settings
->op
[stage
].cop
== WINED3D_TOP_DISABLE
)
6153 arg0
= settings
->op
[stage
].carg0
& WINED3DTA_SELECTMASK
;
6154 arg1
= settings
->op
[stage
].carg1
& WINED3DTA_SELECTMASK
;
6155 arg2
= settings
->op
[stage
].carg2
& WINED3DTA_SELECTMASK
;
6156 if(arg0
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6157 if(arg1
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6158 if(arg2
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6160 if (settings
->op
[stage
].cop
== WINED3D_TOP_BLEND_TEXTURE_ALPHA
)
6161 tex_read
[stage
] = TRUE
;
6162 if (settings
->op
[stage
].cop
== WINED3D_TOP_BLEND_TEXTURE_ALPHA_PM
)
6163 tex_read
[stage
] = TRUE
;
6164 if (settings
->op
[stage
].cop
== WINED3D_TOP_BUMPENVMAP
)
6166 bump_used
[stage
] = TRUE
;
6167 tex_read
[stage
] = TRUE
;
6169 if (settings
->op
[stage
].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
)
6171 bump_used
[stage
] = TRUE
;
6172 tex_read
[stage
] = TRUE
;
6173 luminance_used
[stage
] = TRUE
;
6175 else if (settings
->op
[stage
].cop
== WINED3D_TOP_BLEND_FACTOR_ALPHA
)
6177 tfactor_used
= TRUE
;
6180 if(arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
) {
6181 tfactor_used
= TRUE
;
6184 if(settings
->op
[stage
].dst
== tempreg
) tempreg_used
= TRUE
;
6185 if(arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
) {
6186 tempreg_used
= TRUE
;
6189 if (settings
->op
[stage
].aop
== WINED3D_TOP_DISABLE
)
6191 arg0
= settings
->op
[stage
].aarg0
& WINED3DTA_SELECTMASK
;
6192 arg1
= settings
->op
[stage
].aarg1
& WINED3DTA_SELECTMASK
;
6193 arg2
= settings
->op
[stage
].aarg2
& WINED3DTA_SELECTMASK
;
6194 if(arg0
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6195 if(arg1
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6196 if(arg2
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6198 if(arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
) {
6199 tempreg_used
= TRUE
;
6201 if(arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
) {
6202 tfactor_used
= TRUE
;
6205 lowest_disabled_stage
= stage
;
6208 if (!shader_buffer_init(&buffer
))
6210 ERR("Failed to initialize shader buffer.\n");
6214 shader_addline(&buffer
, "!!ARBfp1.0\n");
6216 switch(settings
->fog
) {
6217 case FOG_OFF
: break;
6218 case FOG_LINEAR
: shader_addline(&buffer
, "OPTION ARB_fog_linear;\n"); break;
6219 case FOG_EXP
: shader_addline(&buffer
, "OPTION ARB_fog_exp;\n"); break;
6220 case FOG_EXP2
: shader_addline(&buffer
, "OPTION ARB_fog_exp2;\n"); break;
6221 default: FIXME("Unexpected fog setting %d\n", settings
->fog
);
6224 shader_addline(&buffer
, "PARAM const = {1, 2, 4, 0.5};\n");
6225 shader_addline(&buffer
, "TEMP TMP;\n");
6226 shader_addline(&buffer
, "TEMP ret;\n");
6227 if(tempreg_used
|| settings
->sRGB_write
) shader_addline(&buffer
, "TEMP tempreg;\n");
6228 shader_addline(&buffer
, "TEMP arg0;\n");
6229 shader_addline(&buffer
, "TEMP arg1;\n");
6230 shader_addline(&buffer
, "TEMP arg2;\n");
6231 for(stage
= 0; stage
< MAX_TEXTURES
; stage
++) {
6232 if(!tex_read
[stage
]) continue;
6233 shader_addline(&buffer
, "TEMP tex%u;\n", stage
);
6234 if(!bump_used
[stage
]) continue;
6235 shader_addline(&buffer
, "PARAM bumpmat%u = program.env[%u];\n", stage
, ARB_FFP_CONST_BUMPMAT(stage
));
6236 if(!luminance_used
[stage
]) continue;
6237 shader_addline(&buffer
, "PARAM luminance%u = program.env[%u];\n", stage
, ARB_FFP_CONST_LUMINANCE(stage
));
6240 shader_addline(&buffer
, "PARAM tfactor = program.env[%u];\n", ARB_FFP_CONST_TFACTOR
);
6242 shader_addline(&buffer
, "PARAM specular_enable = program.env[%u];\n", ARB_FFP_CONST_SPECULAR_ENABLE
);
6244 if(settings
->sRGB_write
) {
6245 shader_addline(&buffer
, "PARAM srgb_consts1 = {%f, %f, %f, %f};\n",
6246 srgb_mul_low
, srgb_cmp
, srgb_pow
, srgb_mul_high
);
6247 shader_addline(&buffer
, "PARAM srgb_consts2 = {%f, %f, %f, %f};\n",
6248 srgb_sub_high
, 0.0, 0.0, 0.0);
6251 if (lowest_disabled_stage
< 7 && settings
->emul_clipplanes
)
6252 shader_addline(&buffer
, "KIL fragment.texcoord[7];\n");
6254 /* Generate texture sampling instructions) */
6255 for (stage
= 0; stage
< MAX_TEXTURES
&& settings
->op
[stage
].cop
!= WINED3D_TOP_DISABLE
; ++stage
)
6257 if (!tex_read
[stage
])
6260 switch(settings
->op
[stage
].tex_type
) {
6261 case tex_1d
: textype
= "1D"; break;
6262 case tex_2d
: textype
= "2D"; break;
6263 case tex_3d
: textype
= "3D"; break;
6264 case tex_cube
: textype
= "CUBE"; break;
6265 case tex_rect
: textype
= "RECT"; break;
6266 default: textype
= "unexpected_textype"; break;
6269 if (settings
->op
[stage
].cop
== WINED3D_TOP_BUMPENVMAP
6270 || settings
->op
[stage
].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
)
6275 if(settings
->op
[stage
].projected
== proj_none
) {
6277 } else if(settings
->op
[stage
].projected
== proj_count4
||
6278 settings
->op
[stage
].projected
== proj_count3
) {
6281 FIXME("Unexpected projection mode %d\n", settings
->op
[stage
].projected
);
6286 && (settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP
6287 || settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
))
6289 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, x, z, 0, 0;\n", stage
- 1);
6290 shader_addline(&buffer
, "DP3 ret.x, arg1, tex%u;\n", stage
- 1);
6291 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, y, w, 0, 0;\n", stage
- 1);
6292 shader_addline(&buffer
, "DP3 ret.y, arg1, tex%u;\n", stage
- 1);
6294 /* with projective textures, texbem only divides the static texture coord, not the displacement,
6295 * so multiply the displacement with the dividing parameter before passing it to TXP
6297 if (settings
->op
[stage
].projected
!= proj_none
) {
6298 if(settings
->op
[stage
].projected
== proj_count4
) {
6299 shader_addline(&buffer
, "MOV ret.w, fragment.texcoord[%u].w;\n", stage
);
6300 shader_addline(&buffer
, "MUL ret.xyz, ret, fragment.texcoord[%u].w, fragment.texcoord[%u];\n", stage
, stage
);
6302 shader_addline(&buffer
, "MOV ret.w, fragment.texcoord[%u].z;\n", stage
);
6303 shader_addline(&buffer
, "MAD ret.xyz, ret, fragment.texcoord[%u].z, fragment.texcoord[%u];\n", stage
, stage
);
6306 shader_addline(&buffer
, "ADD ret, ret, fragment.texcoord[%u];\n", stage
);
6309 shader_addline(&buffer
, "%s%s tex%u, ret, texture[%u], %s;\n",
6310 instr
, sat
, stage
, stage
, textype
);
6311 if (settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
)
6313 shader_addline(&buffer
, "MAD_SAT ret.x, tex%u.z, luminance%u.x, luminance%u.y;\n",
6314 stage
- 1, stage
- 1, stage
- 1);
6315 shader_addline(&buffer
, "MUL tex%u, tex%u, ret.x;\n", stage
, stage
);
6317 } else if(settings
->op
[stage
].projected
== proj_count3
) {
6318 shader_addline(&buffer
, "MOV ret, fragment.texcoord[%u];\n", stage
);
6319 shader_addline(&buffer
, "MOV ret.w, ret.z;\n");
6320 shader_addline(&buffer
, "%s%s tex%u, ret, texture[%u], %s;\n",
6321 instr
, sat
, stage
, stage
, textype
);
6323 shader_addline(&buffer
, "%s%s tex%u, fragment.texcoord[%u], texture[%u], %s;\n",
6324 instr
, sat
, stage
, stage
, stage
, textype
);
6327 sprintf(colorcor_dst
, "tex%u", stage
);
6328 gen_color_correction(&buffer
, colorcor_dst
, WINED3DSP_WRITEMASK_ALL
, "const.x", "const.y",
6329 settings
->op
[stage
].color_fixup
);
6332 /* Generate the main shader */
6333 for (stage
= 0; stage
< MAX_TEXTURES
; ++stage
)
6335 if (settings
->op
[stage
].cop
== WINED3D_TOP_DISABLE
)
6338 final_combiner_src
= "fragment.color.primary";
6342 if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG1
6343 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG1
)
6344 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
;
6345 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG1
6346 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG2
)
6347 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg2
;
6348 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG2
6349 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG1
)
6350 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg1
;
6351 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG2
6352 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG2
)
6353 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
6355 op_equal
= settings
->op
[stage
].aop
== settings
->op
[stage
].cop
6356 && settings
->op
[stage
].carg0
== settings
->op
[stage
].aarg0
6357 && settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
6358 && settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
6360 if (settings
->op
[stage
].aop
== WINED3D_TOP_DISABLE
)
6362 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].dst
,
6363 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6364 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6366 shader_addline(&buffer
, "MOV ret.w, fragment.color.primary.w;\n");
6370 gen_ffp_instr(&buffer
, stage
, TRUE
, TRUE
, settings
->op
[stage
].dst
,
6371 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6372 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6374 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].dst
,
6375 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6376 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6377 gen_ffp_instr(&buffer
, stage
, FALSE
, TRUE
, settings
->op
[stage
].dst
,
6378 settings
->op
[stage
].aop
, settings
->op
[stage
].aarg0
,
6379 settings
->op
[stage
].aarg1
, settings
->op
[stage
].aarg2
);
6383 if(settings
->sRGB_write
) {
6384 shader_addline(&buffer
, "MAD ret, fragment.color.secondary, specular_enable, %s;\n", final_combiner_src
);
6385 arbfp_add_sRGB_correction(&buffer
, "ret", "arg0", "arg1", "arg2", "tempreg", FALSE
);
6386 shader_addline(&buffer
, "MOV result.color, ret;\n");
6388 shader_addline(&buffer
, "MAD result.color, fragment.color.secondary, specular_enable, %s;\n", final_combiner_src
);
6392 shader_addline(&buffer
, "END\n");
6394 /* Generate the shader */
6395 GL_EXTCALL(glGenProgramsARB(1, &ret
));
6396 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, ret
));
6397 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
6398 strlen(buffer
.buffer
), buffer
.buffer
));
6399 checkGLcall("glProgramStringARB()");
6401 gl_info
->gl_ops
.gl
.p_glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
6404 FIXME("Fragment program error at position %d: %s\n\n", pos
,
6405 debugstr_a((const char *)gl_info
->gl_ops
.gl
.p_glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
6406 shader_arb_dump_program_source(buffer
.buffer
);
6412 GL_EXTCALL(glGetProgramivARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
6413 checkGLcall("glGetProgramivARB()");
6414 if (!native
) WARN("Program exceeds native resource limits.\n");
6417 shader_buffer_free(&buffer
);
6421 static void fragment_prog_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6423 const struct wined3d_device
*device
= context
->swapchain
->device
;
6424 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
6425 struct shader_arb_priv
*priv
= device
->fragment_priv
;
6426 BOOL use_pshader
= use_ps(state
);
6427 struct ffp_frag_settings settings
;
6428 const struct arbfp_ffp_desc
*desc
;
6431 TRACE("context %p, state %p, state_id %#x.\n", context
, state
, state_id
);
6433 if (isStateDirty(context
, STATE_RENDER(WINED3D_RS_FOGENABLE
)))
6435 if (!use_pshader
&& device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
)
6437 /* Reload fixed function constants since they collide with the
6438 * pixel shader constants. */
6439 for (i
= 0; i
< MAX_TEXTURES
; ++i
)
6441 set_bumpmat_arbfp(context
, state
, STATE_TEXTURESTAGE(i
, WINED3D_TSS_BUMPENV_MAT00
));
6443 state_texfactor_arbfp(context
, state
, STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
));
6444 state_arb_specularenable(context
, state
, STATE_RENDER(WINED3D_RS_SPECULARENABLE
));
6446 else if (use_pshader
)
6448 context
->select_shader
= 1;
6455 /* Find or create a shader implementing the fixed function pipeline
6456 * settings, then activate it. */
6457 gen_ffp_frag_op(context
, state
, &settings
, FALSE
);
6458 desc
= (const struct arbfp_ffp_desc
*)find_ffp_frag_shader(&priv
->fragment_shaders
, &settings
);
6460 struct arbfp_ffp_desc
*new_desc
= HeapAlloc(GetProcessHeap(), 0, sizeof(*new_desc
));
6463 ERR("Out of memory\n");
6467 new_desc
->parent
.settings
= settings
;
6468 new_desc
->shader
= gen_arbfp_ffp_shader(&settings
, gl_info
);
6469 add_ffp_frag_shader(&priv
->fragment_shaders
, &new_desc
->parent
);
6470 TRACE("Allocated fixed function replacement shader descriptor %p\n", new_desc
);
6474 /* Now activate the replacement program. GL_FRAGMENT_PROGRAM_ARB is already active (however, note the
6475 * comment above the shader_select call below). If e.g. GLSL is active, the shader_select call will
6478 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, desc
->shader
));
6479 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, desc->shader)");
6480 priv
->current_fprogram_id
= desc
->shader
;
6482 if (device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
)
6484 /* Reload fixed function constants since they collide with the
6485 * pixel shader constants. */
6486 for (i
= 0; i
< MAX_TEXTURES
; ++i
)
6488 set_bumpmat_arbfp(context
, state
, STATE_TEXTURESTAGE(i
, WINED3D_TSS_BUMPENV_MAT00
));
6490 state_texfactor_arbfp(context
, state
, STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
));
6491 state_arb_specularenable(context
, state
, STATE_RENDER(WINED3D_RS_SPECULARENABLE
));
6493 context
->last_was_pshader
= FALSE
;
6495 context
->last_was_pshader
= TRUE
;
6498 context
->select_shader
= 1;
6499 context
->load_constants
= 1;
6502 /* We can't link the fog states to the fragment state directly since the
6503 * vertex pipeline links them to FOGENABLE. A different linking in different
6504 * pipeline parts can't be expressed in the combined state table, so we need
6505 * to handle that with a forwarding function. The other invisible side effect
6506 * is that changing the fog start and fog end (which links to FOGENABLE in
6507 * vertex) results in the fragment_prog_arbfp function being called because
6508 * FOGENABLE is dirty, which calls this function here. */
6509 static void state_arbfp_fog(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6511 enum fogsource new_source
;
6513 TRACE("context %p, state %p, state_id %#x.\n", context
, state
, state_id
);
6515 if (!isStateDirty(context
, STATE_PIXELSHADER
))
6516 fragment_prog_arbfp(context
, state
, state_id
);
6518 if (!state
->render_states
[WINED3D_RS_FOGENABLE
])
6521 if (state
->render_states
[WINED3D_RS_FOGTABLEMODE
] == WINED3D_FOG_NONE
)
6525 new_source
= FOGSOURCE_VS
;
6529 if (state
->render_states
[WINED3D_RS_FOGVERTEXMODE
] == WINED3D_FOG_NONE
|| context
->last_was_rhw
)
6530 new_source
= FOGSOURCE_COORD
;
6532 new_source
= FOGSOURCE_FFP
;
6537 new_source
= FOGSOURCE_FFP
;
6540 if (new_source
!= context
->fog_source
)
6542 context
->fog_source
= new_source
;
6543 state_fogstartend(context
, state
, STATE_RENDER(WINED3D_RS_FOGSTART
));
6547 static void textransform(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6549 if (!isStateDirty(context
, STATE_PIXELSHADER
))
6550 fragment_prog_arbfp(context
, state
, state_id
);
6553 static const struct StateEntryTemplate arbfp_fragmentstate_template
[] =
6555 {STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
), { STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
), state_texfactor_arbfp
}, WINED3D_GL_EXT_NONE
},
6556 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_OP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6557 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6558 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6559 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6560 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_OP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6561 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6562 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6563 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6564 {STATE_TEXTURESTAGE(0, WINED3D_TSS_RESULT_ARG
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6565 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6566 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6567 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6568 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6569 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6570 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6571 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_OP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6572 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6573 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6574 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6575 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_OP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6576 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6577 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6578 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6579 {STATE_TEXTURESTAGE(1, WINED3D_TSS_RESULT_ARG
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6580 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6581 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6582 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6583 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6584 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6585 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6586 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_OP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6587 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6588 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6589 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6590 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_OP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6591 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6592 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6593 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6594 {STATE_TEXTURESTAGE(2, WINED3D_TSS_RESULT_ARG
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6595 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6596 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6597 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6598 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6599 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6600 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6601 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_OP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6602 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6603 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6604 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6605 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_OP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6606 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6607 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6608 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6609 {STATE_TEXTURESTAGE(3, WINED3D_TSS_RESULT_ARG
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6610 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6611 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6612 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6613 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6614 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6615 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6616 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_OP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6617 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6618 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6619 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6620 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_OP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6621 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6622 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6623 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6624 {STATE_TEXTURESTAGE(4, WINED3D_TSS_RESULT_ARG
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6625 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6626 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6627 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6628 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6629 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6630 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6631 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_OP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6632 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6633 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6634 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6635 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_OP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6636 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6637 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6638 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6639 {STATE_TEXTURESTAGE(5, WINED3D_TSS_RESULT_ARG
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6640 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6641 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6642 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6643 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6644 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6645 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6646 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_OP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6647 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6648 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6649 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6650 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_OP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6651 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6652 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6653 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6654 {STATE_TEXTURESTAGE(6, WINED3D_TSS_RESULT_ARG
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6655 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6656 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6657 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6658 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6659 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6660 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6661 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_OP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6662 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6663 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6664 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6665 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_OP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6666 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6667 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6668 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6669 {STATE_TEXTURESTAGE(7, WINED3D_TSS_RESULT_ARG
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6670 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6671 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6672 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6673 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6674 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6675 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6676 {STATE_PIXELSHADER
, { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
6677 {STATE_RENDER(WINED3D_RS_FOGENABLE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), state_arbfp_fog
}, WINED3D_GL_EXT_NONE
},
6678 {STATE_RENDER(WINED3D_RS_FOGTABLEMODE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6679 {STATE_RENDER(WINED3D_RS_FOGVERTEXMODE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6680 {STATE_RENDER(WINED3D_RS_FOGSTART
), { STATE_RENDER(WINED3D_RS_FOGSTART
), state_fogstartend
}, WINED3D_GL_EXT_NONE
},
6681 {STATE_RENDER(WINED3D_RS_FOGEND
), { STATE_RENDER(WINED3D_RS_FOGSTART
), NULL
}, WINED3D_GL_EXT_NONE
},
6682 {STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), { STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), state_srgbwrite
}, ARB_FRAMEBUFFER_SRGB
},
6683 {STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6684 {STATE_RENDER(WINED3D_RS_FOGCOLOR
), { STATE_RENDER(WINED3D_RS_FOGCOLOR
), state_fogcolor
}, WINED3D_GL_EXT_NONE
},
6685 {STATE_RENDER(WINED3D_RS_FOGDENSITY
), { STATE_RENDER(WINED3D_RS_FOGDENSITY
), state_fogdensity
}, WINED3D_GL_EXT_NONE
},
6686 {STATE_TEXTURESTAGE(0,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(0, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6687 {STATE_TEXTURESTAGE(1,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(1, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6688 {STATE_TEXTURESTAGE(2,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(2, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6689 {STATE_TEXTURESTAGE(3,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(3, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6690 {STATE_TEXTURESTAGE(4,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(4, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6691 {STATE_TEXTURESTAGE(5,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(5, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6692 {STATE_TEXTURESTAGE(6,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(6, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6693 {STATE_TEXTURESTAGE(7,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(7, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6694 {STATE_RENDER(WINED3D_RS_SPECULARENABLE
), { STATE_RENDER(WINED3D_RS_SPECULARENABLE
), state_arb_specularenable
}, WINED3D_GL_EXT_NONE
},
6695 {0 /* Terminate */, { 0, 0 }, WINED3D_GL_EXT_NONE
},
6698 const struct fragment_pipeline arbfp_fragment_pipeline
= {
6703 shader_arb_color_fixup_supported
,
6704 arbfp_fragmentstate_template
,
6707 struct arbfp_blit_priv
{
6708 GLenum yuy2_rect_shader
, yuy2_2d_shader
;
6709 GLenum uyvy_rect_shader
, uyvy_2d_shader
;
6710 GLenum yv12_rect_shader
, yv12_2d_shader
;
6711 GLenum p8_rect_shader
, p8_2d_shader
;
6712 GLuint palette_texture
;
6715 static HRESULT
arbfp_blit_alloc(struct wined3d_device
*device
)
6717 device
->blit_priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct arbfp_blit_priv
));
6718 if(!device
->blit_priv
) {
6719 ERR("Out of memory\n");
6720 return E_OUTOFMEMORY
;
6725 /* Context activation is done by the caller. */
6726 static void arbfp_blit_free(struct wined3d_device
*device
)
6728 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
6729 struct arbfp_blit_priv
*priv
= device
->blit_priv
;
6731 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->yuy2_rect_shader
));
6732 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->yuy2_2d_shader
));
6733 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->uyvy_rect_shader
));
6734 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->uyvy_2d_shader
));
6735 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->yv12_rect_shader
));
6736 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->yv12_2d_shader
));
6737 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->p8_rect_shader
));
6738 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->p8_2d_shader
));
6739 checkGLcall("Delete yuv and p8 programs");
6741 if (priv
->palette_texture
)
6742 gl_info
->gl_ops
.gl
.p_glDeleteTextures(1, &priv
->palette_texture
);
6744 HeapFree(GetProcessHeap(), 0, device
->blit_priv
);
6745 device
->blit_priv
= NULL
;
6748 static BOOL
gen_planar_yuv_read(struct wined3d_shader_buffer
*buffer
, enum complex_fixup fixup
,
6749 GLenum textype
, char *luminance
)
6752 const char *tex
, *texinstr
;
6754 if (fixup
== COMPLEX_FIXUP_UYVY
) {
6762 case GL_TEXTURE_2D
: tex
= "2D"; texinstr
= "TXP"; break;
6763 case GL_TEXTURE_RECTANGLE_ARB
: tex
= "RECT"; texinstr
= "TEX"; break;
6765 /* This is more tricky than just replacing the texture type - we have to navigate
6766 * properly in the texture to find the correct chroma values
6768 FIXME("Implement yuv correction for non-2d, non-rect textures\n");
6772 /* First we have to read the chroma values. This means we need at least two pixels(no filtering),
6773 * or 4 pixels(with filtering). To get the unmodified chromas, we have to rid ourselves of the
6774 * filtering when we sample the texture.
6776 * These are the rules for reading the chroma:
6782 * So we have to get the sampling x position in non-normalized coordinates in integers
6784 if(textype
!= GL_TEXTURE_RECTANGLE_ARB
) {
6785 shader_addline(buffer
, "MUL texcrd.xy, fragment.texcoord[0], size.x;\n");
6786 shader_addline(buffer
, "MOV texcrd.w, size.x;\n");
6788 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
6790 /* We must not allow filtering between pixel x and x+1, this would mix U and V
6791 * Vertical filtering is ok. However, bear in mind that the pixel center is at
6794 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
6795 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, coef.y;\n");
6797 /* Divide the x coordinate by 0.5 and get the fraction. This gives 0.25 and 0.75 for the
6798 * even and odd pixels respectively
6800 shader_addline(buffer
, "MUL texcrd2, texcrd, coef.y;\n");
6801 shader_addline(buffer
, "FRC texcrd2, texcrd2;\n");
6803 /* Sample Pixel 1 */
6804 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
6806 /* Put the value into either of the chroma values */
6807 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
6808 shader_addline(buffer
, "MUL chroma.x, luminance.%c, temp.x;\n", chroma
);
6809 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
6810 shader_addline(buffer
, "MUL chroma.y, luminance.%c, temp.x;\n", chroma
);
6812 /* Sample pixel 2. If we read an even pixel(SLT above returned 1), sample
6813 * the pixel right to the current one. Otherwise, sample the left pixel.
6814 * Bias and scale the SLT result to -1;1 and add it to the texcrd.x.
6816 shader_addline(buffer
, "MAD temp.x, temp.x, coef.z, -coef.x;\n");
6817 shader_addline(buffer
, "ADD texcrd.x, texcrd, temp.x;\n");
6818 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
6820 /* Put the value into the other chroma */
6821 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
6822 shader_addline(buffer
, "MAD chroma.y, luminance.%c, temp.x, chroma.y;\n", chroma
);
6823 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
6824 shader_addline(buffer
, "MAD chroma.x, luminance.%c, temp.x, chroma.x;\n", chroma
);
6826 /* TODO: If filtering is enabled, sample a 2nd pair of pixels left or right of
6827 * the current one and lerp the two U and V values
6830 /* This gives the correctly filtered luminance value */
6831 shader_addline(buffer
, "TEX luminance, fragment.texcoord[0], texture[0], %s;\n", tex
);
6836 static BOOL
gen_yv12_read(struct wined3d_shader_buffer
*buffer
, GLenum textype
, char *luminance
)
6841 case GL_TEXTURE_2D
: tex
= "2D"; break;
6842 case GL_TEXTURE_RECTANGLE_ARB
: tex
= "RECT"; break;
6844 FIXME("Implement yv12 correction for non-2d, non-rect textures\n");
6848 /* YV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
6849 * V and a (W/2)x(H/2) U plane, each with 8 bit per pixel. So the effective
6850 * bitdepth is 12 bits per pixel. Since the U and V planes have only half the
6851 * pitch of the luminance plane, the packing into the gl texture is a bit
6852 * unfortunate. If the whole texture is interpreted as luminance data it looks
6853 * approximately like this:
6855 * +----------------------------------+----
6867 * +----------------+-----------------+----
6869 * | U even rows | U odd rows |
6871 * +----------------+------------------ -
6873 * | V even rows | V odd rows |
6875 * +----------------+-----------------+----
6879 * So it appears as if there are 4 chroma images, but in fact the odd rows
6880 * in the chroma images are in the same row as the even ones. So its is
6881 * kinda tricky to read
6883 * When reading from rectangle textures, keep in mind that the input y coordinates
6884 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height
6886 shader_addline(buffer
, "PARAM yv12_coef = {%f, %f, %f, %f};\n",
6887 2.0f
/ 3.0f
, 1.0f
/ 6.0f
, (2.0f
/ 3.0f
) + (1.0f
/ 6.0f
), 1.0f
/ 3.0f
);
6889 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
6890 /* the chroma planes have only half the width */
6891 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, coef.y;\n");
6893 /* The first value is between 2/3 and 5/6th of the texture's height, so scale+bias
6894 * the coordinate. Also read the right side of the image when reading odd lines
6896 * Don't forget to clamp the y values in into the range, otherwise we'll get filtering
6899 if(textype
== GL_TEXTURE_2D
) {
6901 shader_addline(buffer
, "RCP chroma.w, size.y;\n");
6903 shader_addline(buffer
, "MUL texcrd2.y, texcrd.y, size.y;\n");
6905 shader_addline(buffer
, "FLR texcrd2.y, texcrd2.y;\n");
6906 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, yv12_coef.y, yv12_coef.x;\n");
6908 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
6909 shader_addline(buffer
, "ADD texcrd2.x, texcrd2.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
6910 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
6911 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
6912 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
6914 /* clamp, keep the half pixel origin in mind */
6915 shader_addline(buffer
, "MAD temp.y, coef.y, chroma.w, yv12_coef.x;\n");
6916 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
6917 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.z;\n");
6918 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
6920 /* Read from [size - size+size/4] */
6921 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
6922 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, coef.w, size.y;\n");
6924 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
6925 shader_addline(buffer
, "ADD texcrd2.x, texcrd.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
6926 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
6927 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
6928 shader_addline(buffer
, "MUL texcrd2.x, texcrd2.x, size.x;\n");
6929 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
6931 /* Make sure to read exactly from the pixel center */
6932 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
6933 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, coef.y;\n");
6936 shader_addline(buffer
, "MAD temp.y, size.y, coef.w, size.y;\n");
6937 shader_addline(buffer
, "ADD temp.y, temp.y, -coef.y;\n");
6938 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
6939 shader_addline(buffer
, "ADD temp.y, size.y, -coef.y;\n");
6940 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
6942 /* Read the texture, put the result into the output register */
6943 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
6944 shader_addline(buffer
, "MOV chroma.x, temp.w;\n");
6946 /* The other chroma value is 1/6th of the texture lower, from 5/6th to 6/6th
6947 * No need to clamp because we're just reusing the already clamped value from above
6949 if(textype
== GL_TEXTURE_2D
) {
6950 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, yv12_coef.y;\n");
6952 shader_addline(buffer
, "MAD texcrd.y, size.y, coef.w, texcrd.y;\n");
6954 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
6955 shader_addline(buffer
, "MOV chroma.y, temp.w;\n");
6957 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
6958 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
6959 * values due to filtering
6961 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
6962 if(textype
== GL_TEXTURE_2D
) {
6963 /* Multiply the y coordinate by 2/3 and clamp it */
6964 shader_addline(buffer
, "MUL texcrd.y, texcrd.y, yv12_coef.x;\n");
6965 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.x;\n");
6966 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
6967 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
6969 /* Reading from texture_rectangles is pretty straightforward, just use the unmodified
6970 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
6973 shader_addline(buffer
, "ADD temp.x, size.y, -coef.y;\n");
6974 shader_addline(buffer
, "MIN texcrd.y, texcrd.y, size.x;\n");
6975 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
6982 static GLuint
gen_p8_shader(struct arbfp_blit_priv
*priv
,
6983 const struct wined3d_gl_info
*gl_info
, GLenum textype
)
6986 struct wined3d_shader_buffer buffer
;
6990 if (!shader_buffer_init(&buffer
))
6992 ERR("Failed to initialize shader buffer.\n");
6996 GL_EXTCALL(glGenProgramsARB(1, &shader
));
6997 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7000 shader_buffer_free(&buffer
);
7004 shader_addline(&buffer
, "!!ARBfp1.0\n");
7005 shader_addline(&buffer
, "TEMP index;\n");
7007 /* { 255/256, 0.5/255*255/256, 0, 0 } */
7008 shader_addline(&buffer
, "PARAM constants = { 0.996, 0.00195, 0, 0 };\n");
7010 /* The alpha-component contains the palette index */
7011 if(textype
== GL_TEXTURE_RECTANGLE_ARB
)
7012 shader_addline(&buffer
, "TXP index, fragment.texcoord[0], texture[0], RECT;\n");
7014 shader_addline(&buffer
, "TEX index, fragment.texcoord[0], texture[0], 2D;\n");
7016 /* Scale the index by 255/256 and add a bias of '0.5' in order to sample in the middle */
7017 shader_addline(&buffer
, "MAD index.a, index.a, constants.x, constants.y;\n");
7019 /* Use the alpha-component as an index in the palette to get the final color */
7020 shader_addline(&buffer
, "TEX result.color, index.a, texture[1], 1D;\n");
7021 shader_addline(&buffer
, "END\n");
7023 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
7024 strlen(buffer
.buffer
), buffer
.buffer
));
7025 checkGLcall("glProgramStringARB()");
7027 gl_info
->gl_ops
.gl
.p_glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
7030 FIXME("Fragment program error at position %d: %s\n\n", pos
,
7031 debugstr_a((const char *)gl_info
->gl_ops
.gl
.p_glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
7032 shader_arb_dump_program_source(buffer
.buffer
);
7035 if (textype
== GL_TEXTURE_RECTANGLE_ARB
)
7036 priv
->p8_rect_shader
= shader
;
7038 priv
->p8_2d_shader
= shader
;
7040 shader_buffer_free(&buffer
);
7045 /* Context activation is done by the caller. */
7046 static void upload_palette(const struct wined3d_surface
*surface
, struct wined3d_context
*context
)
7049 struct wined3d_device
*device
= surface
->resource
.device
;
7050 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
7051 struct arbfp_blit_priv
*priv
= device
->blit_priv
;
7052 BOOL colorkey
= (surface
->CKeyFlags
& WINEDDSD_CKSRCBLT
) != 0;
7054 d3dfmt_p8_init_palette(surface
, table
, colorkey
);
7056 if (gl_info
->supported
[APPLE_CLIENT_STORAGE
])
7058 gl_info
->gl_ops
.gl
.p_glPixelStorei(GL_UNPACK_CLIENT_STORAGE_APPLE
, GL_FALSE
);
7059 checkGLcall("glPixelStorei(GL_UNPACK_CLIENT_STORAGE_APPLE, GL_FALSE)");
7062 if (!priv
->palette_texture
)
7063 gl_info
->gl_ops
.gl
.p_glGenTextures(1, &priv
->palette_texture
);
7065 GL_EXTCALL(glActiveTextureARB(GL_TEXTURE1
));
7066 gl_info
->gl_ops
.gl
.p_glBindTexture(GL_TEXTURE_1D
, priv
->palette_texture
);
7068 gl_info
->gl_ops
.gl
.p_glTexEnvi(GL_TEXTURE_ENV
, GL_TEXTURE_ENV_MODE
, GL_REPLACE
);
7070 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_MIN_FILTER
, GL_NEAREST
);
7071 /* Make sure we have discrete color levels. */
7072 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_MAG_FILTER
, GL_NEAREST
);
7073 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_WRAP_S
, GL_CLAMP_TO_EDGE
);
7074 /* Upload the palette */
7075 /* TODO: avoid unneeded uploads in the future by adding some SFLAG_PALETTE_DIRTY mechanism */
7076 gl_info
->gl_ops
.gl
.p_glTexImage1D(GL_TEXTURE_1D
, 0, GL_RGBA
, 256, 0, GL_RGBA
, GL_UNSIGNED_BYTE
, table
);
7078 if (gl_info
->supported
[APPLE_CLIENT_STORAGE
])
7080 gl_info
->gl_ops
.gl
.p_glPixelStorei(GL_UNPACK_CLIENT_STORAGE_APPLE
, GL_TRUE
);
7081 checkGLcall("glPixelStorei(GL_UNPACK_CLIENT_STORAGE_APPLE, GL_TRUE)");
7084 /* Switch back to unit 0 in which the 2D texture will be stored. */
7085 context_active_texture(context
, gl_info
, 0);
7088 /* Context activation is done by the caller. */
7089 static GLuint
gen_yuv_shader(struct arbfp_blit_priv
*priv
, const struct wined3d_gl_info
*gl_info
,
7090 enum complex_fixup yuv_fixup
, GLenum textype
)
7093 struct wined3d_shader_buffer buffer
;
7094 char luminance_component
;
7098 if (!shader_buffer_init(&buffer
))
7100 ERR("Failed to initialize shader buffer.\n");
7104 GL_EXTCALL(glGenProgramsARB(1, &shader
));
7105 checkGLcall("GL_EXTCALL(glGenProgramsARB(1, &shader))");
7106 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7107 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
7110 shader_buffer_free(&buffer
);
7114 /* The YUY2 and UYVY formats contain two pixels packed into a 32 bit macropixel,
7115 * giving effectively 16 bit per pixel. The color consists of a luminance(Y) and
7116 * two chroma(U and V) values. Each macropixel has two luminance values, one for
7117 * each single pixel it contains, and one U and one V value shared between both
7120 * The data is loaded into an A8L8 texture. With YUY2, the luminance component
7121 * contains the luminance and alpha the chroma. With UYVY it is vice versa. Thus
7122 * take the format into account when generating the read swizzles
7124 * Reading the Y value is straightforward - just sample the texture. The hardware
7125 * takes care of filtering in the horizontal and vertical direction.
7127 * Reading the U and V values is harder. We have to avoid filtering horizontally,
7128 * because that would mix the U and V values of one pixel or two adjacent pixels.
7129 * Thus floor the texture coordinate and add 0.5 to get an unfiltered read,
7130 * regardless of the filtering setting. Vertical filtering works automatically
7131 * though - the U and V values of two rows are mixed nicely.
7133 * Apart of avoiding filtering issues, the code has to know which value it just
7134 * read, and where it can find the other one. To determine this, it checks if
7135 * it sampled an even or odd pixel, and shifts the 2nd read accordingly.
7137 * Handling horizontal filtering of U and V values requires reading a 2nd pair
7138 * of pixels, extracting U and V and mixing them. This is not implemented yet.
7140 * An alternative implementation idea is to load the texture as A8R8G8B8 texture,
7141 * with width / 2. This way one read gives all 3 values, finding U and V is easy
7142 * in an unfiltered situation. Finding the luminance on the other hand requires
7143 * finding out if it is an odd or even pixel. The real drawback of this approach
7144 * is filtering. This would have to be emulated completely in the shader, reading
7145 * up two 2 packed pixels in up to 2 rows and interpolating both horizontally and
7146 * vertically. Beyond that it would require adjustments to the texture handling
7147 * code to deal with the width scaling
7149 shader_addline(&buffer
, "!!ARBfp1.0\n");
7150 shader_addline(&buffer
, "TEMP luminance;\n");
7151 shader_addline(&buffer
, "TEMP temp;\n");
7152 shader_addline(&buffer
, "TEMP chroma;\n");
7153 shader_addline(&buffer
, "TEMP texcrd;\n");
7154 shader_addline(&buffer
, "TEMP texcrd2;\n");
7155 shader_addline(&buffer
, "PARAM coef = {1.0, 0.5, 2.0, 0.25};\n");
7156 shader_addline(&buffer
, "PARAM yuv_coef = {1.403, 0.344, 0.714, 1.770};\n");
7157 shader_addline(&buffer
, "PARAM size = program.local[0];\n");
7161 case COMPLEX_FIXUP_UYVY
:
7162 case COMPLEX_FIXUP_YUY2
:
7163 if (!gen_planar_yuv_read(&buffer
, yuv_fixup
, textype
, &luminance_component
))
7165 shader_buffer_free(&buffer
);
7170 case COMPLEX_FIXUP_YV12
:
7171 if (!gen_yv12_read(&buffer
, textype
, &luminance_component
))
7173 shader_buffer_free(&buffer
);
7179 FIXME("Unsupported YUV fixup %#x\n", yuv_fixup
);
7180 shader_buffer_free(&buffer
);
7184 /* Calculate the final result. Formula is taken from
7185 * http://www.fourcc.org/fccyvrgb.php. Note that the chroma
7186 * ranges from -0.5 to 0.5
7188 shader_addline(&buffer
, "SUB chroma.xy, chroma, coef.y;\n");
7190 shader_addline(&buffer
, "MAD result.color.x, chroma.x, yuv_coef.x, luminance.%c;\n", luminance_component
);
7191 shader_addline(&buffer
, "MAD temp.x, -chroma.y, yuv_coef.y, luminance.%c;\n", luminance_component
);
7192 shader_addline(&buffer
, "MAD result.color.y, -chroma.x, yuv_coef.z, temp.x;\n");
7193 shader_addline(&buffer
, "MAD result.color.z, chroma.y, yuv_coef.w, luminance.%c;\n", luminance_component
);
7194 shader_addline(&buffer
, "END\n");
7196 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
7197 strlen(buffer
.buffer
), buffer
.buffer
));
7198 checkGLcall("glProgramStringARB()");
7200 gl_info
->gl_ops
.gl
.p_glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
7203 FIXME("Fragment program error at position %d: %s\n\n", pos
,
7204 debugstr_a((const char *)gl_info
->gl_ops
.gl
.p_glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
7205 shader_arb_dump_program_source(buffer
.buffer
);
7211 GL_EXTCALL(glGetProgramivARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
7212 checkGLcall("glGetProgramivARB()");
7213 if (!native
) WARN("Program exceeds native resource limits.\n");
7216 shader_buffer_free(&buffer
);
7220 case COMPLEX_FIXUP_YUY2
:
7221 if (textype
== GL_TEXTURE_RECTANGLE_ARB
) priv
->yuy2_rect_shader
= shader
;
7222 else priv
->yuy2_2d_shader
= shader
;
7225 case COMPLEX_FIXUP_UYVY
:
7226 if (textype
== GL_TEXTURE_RECTANGLE_ARB
) priv
->uyvy_rect_shader
= shader
;
7227 else priv
->uyvy_2d_shader
= shader
;
7230 case COMPLEX_FIXUP_YV12
:
7231 if (textype
== GL_TEXTURE_RECTANGLE_ARB
) priv
->yv12_rect_shader
= shader
;
7232 else priv
->yv12_2d_shader
= shader
;
7235 ERR("Unsupported complex fixup: %d\n", yuv_fixup
);
7241 /* Context activation is done by the caller. */
7242 static HRESULT
arbfp_blit_set(void *blit_priv
, struct wined3d_context
*context
, const struct wined3d_surface
*surface
)
7245 float size
[4] = {(float) surface
->pow2Width
, (float) surface
->pow2Height
, 1.0f
, 1.0f
};
7246 struct arbfp_blit_priv
*priv
= blit_priv
;
7247 enum complex_fixup fixup
;
7248 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
7251 if (surface
->container
.type
== WINED3D_CONTAINER_TEXTURE
)
7252 textype
= surface
->container
.u
.texture
->target
;
7254 textype
= surface
->texture_target
;
7256 if (surface
->flags
& SFLAG_CONVERTED
)
7258 gl_info
->gl_ops
.gl
.p_glEnable(textype
);
7259 checkGLcall("glEnable(textype)");
7263 if (!is_complex_fixup(surface
->resource
.format
->color_fixup
))
7266 dump_color_fixup_desc(surface
->resource
.format
->color_fixup
);
7267 /* Don't bother setting up a shader for unconverted formats */
7268 gl_info
->gl_ops
.gl
.p_glEnable(textype
);
7269 checkGLcall("glEnable(textype)");
7273 fixup
= get_complex_fixup(surface
->resource
.format
->color_fixup
);
7277 case COMPLEX_FIXUP_YUY2
:
7278 shader
= textype
== GL_TEXTURE_RECTANGLE_ARB
? priv
->yuy2_rect_shader
: priv
->yuy2_2d_shader
;
7281 case COMPLEX_FIXUP_UYVY
:
7282 shader
= textype
== GL_TEXTURE_RECTANGLE_ARB
? priv
->uyvy_rect_shader
: priv
->uyvy_2d_shader
;
7285 case COMPLEX_FIXUP_YV12
:
7286 shader
= textype
== GL_TEXTURE_RECTANGLE_ARB
? priv
->yv12_rect_shader
: priv
->yv12_2d_shader
;
7289 case COMPLEX_FIXUP_P8
:
7290 shader
= textype
== GL_TEXTURE_RECTANGLE_ARB
? priv
->p8_rect_shader
: priv
->p8_2d_shader
;
7291 if (!shader
) shader
= gen_p8_shader(priv
, gl_info
, textype
);
7293 upload_palette(surface
, context
);
7297 FIXME("Unsupported complex fixup %#x, not setting a shader\n", fixup
);
7298 gl_info
->gl_ops
.gl
.p_glEnable(textype
);
7299 checkGLcall("glEnable(textype)");
7303 if (!shader
) shader
= gen_yuv_shader(priv
, gl_info
, fixup
, textype
);
7305 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
7306 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
7307 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7308 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
7309 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, 0, size
));
7310 checkGLcall("glProgramLocalParameter4fvARB");
7315 /* Context activation is done by the caller. */
7316 static void arbfp_blit_unset(const struct wined3d_gl_info
*gl_info
)
7318 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
7319 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
7320 gl_info
->gl_ops
.gl
.p_glDisable(GL_TEXTURE_2D
);
7321 checkGLcall("glDisable(GL_TEXTURE_2D)");
7322 if (gl_info
->supported
[ARB_TEXTURE_CUBE_MAP
])
7324 gl_info
->gl_ops
.gl
.p_glDisable(GL_TEXTURE_CUBE_MAP_ARB
);
7325 checkGLcall("glDisable(GL_TEXTURE_CUBE_MAP_ARB)");
7327 if (gl_info
->supported
[ARB_TEXTURE_RECTANGLE
])
7329 gl_info
->gl_ops
.gl
.p_glDisable(GL_TEXTURE_RECTANGLE_ARB
);
7330 checkGLcall("glDisable(GL_TEXTURE_RECTANGLE_ARB)");
7334 static BOOL
arbfp_blit_supported(const struct wined3d_gl_info
*gl_info
, enum wined3d_blit_op blit_op
,
7335 const RECT
*src_rect
, DWORD src_usage
, enum wined3d_pool src_pool
, const struct wined3d_format
*src_format
,
7336 const RECT
*dst_rect
, DWORD dst_usage
, enum wined3d_pool dst_pool
, const struct wined3d_format
*dst_format
)
7338 enum complex_fixup src_fixup
;
7340 if (!gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
7343 if (blit_op
!= WINED3D_BLIT_OP_COLOR_BLIT
)
7345 TRACE("Unsupported blit_op=%d\n", blit_op
);
7349 if (src_pool
== WINED3D_POOL_SYSTEM_MEM
|| dst_pool
== WINED3D_POOL_SYSTEM_MEM
)
7352 src_fixup
= get_complex_fixup(src_format
->color_fixup
);
7353 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
7355 TRACE("Checking support for fixup:\n");
7356 dump_color_fixup_desc(src_format
->color_fixup
);
7359 if (!is_identity_fixup(dst_format
->color_fixup
))
7361 TRACE("Destination fixups are not supported\n");
7365 if (is_identity_fixup(src_format
->color_fixup
))
7371 /* We only support YUV conversions. */
7372 if (!is_complex_fixup(src_format
->color_fixup
))
7374 TRACE("[FAILED]\n");
7380 case COMPLEX_FIXUP_YUY2
:
7381 case COMPLEX_FIXUP_UYVY
:
7382 case COMPLEX_FIXUP_YV12
:
7383 case COMPLEX_FIXUP_P8
:
7388 FIXME("Unsupported YUV fixup %#x\n", src_fixup
);
7389 TRACE("[FAILED]\n");
7394 HRESULT
arbfp_blit_surface(struct wined3d_device
*device
, DWORD filter
,
7395 struct wined3d_surface
*src_surface
, const RECT
*src_rect_in
,
7396 struct wined3d_surface
*dst_surface
, const RECT
*dst_rect_in
)
7398 struct wined3d_context
*context
;
7399 RECT src_rect
= *src_rect_in
;
7400 RECT dst_rect
= *dst_rect_in
;
7402 /* Now load the surface */
7403 if (wined3d_settings
.offscreen_rendering_mode
!= ORM_FBO
7404 && (src_surface
->flags
& (SFLAG_INTEXTURE
| SFLAG_INDRAWABLE
)) == SFLAG_INDRAWABLE
)
7406 /* Without FBO blits transferring from the drawable to the texture is
7407 * expensive, because we have to flip the data in sysmem. Since we can
7408 * flip in the blitter, we don't actually need that flip anyway. So we
7409 * use the surface's texture as scratch texture, and flip the source
7410 * rectangle instead. */
7411 surface_load_fb_texture(src_surface
, FALSE
);
7413 src_rect
.top
= src_surface
->resource
.height
- src_rect
.top
;
7414 src_rect
.bottom
= src_surface
->resource
.height
- src_rect
.bottom
;
7417 surface_internal_preload(src_surface
, SRGB_RGB
);
7419 /* Activate the destination context, set it up for blitting */
7420 context
= context_acquire(device
, dst_surface
);
7421 context_apply_blit_state(context
, device
);
7423 if (!surface_is_offscreen(dst_surface
))
7424 surface_translate_drawable_coords(dst_surface
, context
->win_handle
, &dst_rect
);
7426 arbfp_blit_set(device
->blit_priv
, context
, src_surface
);
7428 /* Draw a textured quad */
7429 draw_textured_quad(src_surface
, context
, &src_rect
, &dst_rect
, filter
);
7431 /* Leave the opengl state valid for blitting */
7432 arbfp_blit_unset(context
->gl_info
);
7434 if (wined3d_settings
.strict_draw_ordering
7435 || (dst_surface
->container
.type
== WINED3D_CONTAINER_SWAPCHAIN
7436 && (dst_surface
->container
.u
.swapchain
->front_buffer
== dst_surface
)))
7437 context
->gl_info
->gl_ops
.gl
.p_glFlush(); /* Flush to ensure ordering across contexts. */
7439 context_release(context
);
7441 surface_modify_location(dst_surface
, dst_surface
->draw_binding
, TRUE
);
7445 /* Do not call while under the GL lock. */
7446 static HRESULT
arbfp_blit_color_fill(struct wined3d_device
*device
, struct wined3d_surface
*dst_surface
,
7447 const RECT
*dst_rect
, const struct wined3d_color
*color
)
7449 FIXME("Color filling not implemented by arbfp_blit\n");
7450 return WINED3DERR_INVALIDCALL
;
7453 /* Do not call while under the GL lock. */
7454 static HRESULT
arbfp_blit_depth_fill(struct wined3d_device
*device
,
7455 struct wined3d_surface
*surface
, const RECT
*rect
, float depth
)
7457 FIXME("Depth filling not implemented by arbfp_blit.\n");
7458 return WINED3DERR_INVALIDCALL
;
7461 const struct blit_shader arbfp_blit
= {
7466 arbfp_blit_supported
,
7467 arbfp_blit_color_fill
,
7468 arbfp_blit_depth_fill
,