2 * Pixel and vertex shaders implementation using ARB_vertex_program
3 * and ARB_fragment_program GL extensions.
5 * Copyright 2002-2003 Jason Edmeades
6 * Copyright 2002-2003 Raphael Junqueira
7 * Copyright 2004 Christian Costa
8 * Copyright 2005 Oliver Stieber
9 * Copyright 2006 Ivan Gyurdiev
10 * Copyright 2006 Jason Green
11 * Copyright 2006 Henri Verbeet
12 * Copyright 2007-2011, 2013-2014 Stefan Dösinger for CodeWeavers
13 * Copyright 2009 Henri Verbeet for CodeWeavers
15 * This library is free software; you can redistribute it and/or
16 * modify it under the terms of the GNU Lesser General Public
17 * License as published by the Free Software Foundation; either
18 * version 2.1 of the License, or (at your option) any later version.
20 * This library is distributed in the hope that it will be useful,
21 * but WITHOUT ANY WARRANTY; without even the implied warranty of
22 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
23 * Lesser General Public License for more details.
25 * You should have received a copy of the GNU Lesser General Public
26 * License along with this library; if not, write to the Free Software
27 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
32 #include "wined3d_private.h"
34 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader
);
35 WINE_DECLARE_DEBUG_CHANNEL(d3d_constants
);
36 WINE_DECLARE_DEBUG_CHANNEL(d3d
);
37 WINE_DECLARE_DEBUG_CHANNEL(d3d_perf
);
39 static BOOL
shader_is_pshader_version(enum wined3d_shader_type type
)
41 return type
== WINED3D_SHADER_TYPE_PIXEL
;
44 static BOOL
shader_is_vshader_version(enum wined3d_shader_type type
)
46 return type
== WINED3D_SHADER_TYPE_VERTEX
;
60 static const char *arb_get_helper_value(enum wined3d_shader_type shader
, enum arb_helper_value value
)
62 if (shader
!= WINED3D_SHADER_TYPE_VERTEX
&& shader
!= WINED3D_SHADER_TYPE_PIXEL
)
64 ERR("Unsupported shader type '%s'.\n", debug_shader_type(shader
));
68 if (shader
== WINED3D_SHADER_TYPE_PIXEL
)
72 case ARB_ZERO
: return "ps_helper_const.x";
73 case ARB_ONE
: return "ps_helper_const.y";
74 case ARB_TWO
: return "coefmul.x";
75 case ARB_0001
: return "ps_helper_const.xxxy";
76 case ARB_EPS
: return "ps_helper_const.z";
84 case ARB_ZERO
: return "helper_const.x";
85 case ARB_ONE
: return "helper_const.y";
86 case ARB_TWO
: return "helper_const.z";
87 case ARB_EPS
: return "helper_const.w";
88 case ARB_0001
: return "helper_const.xxxy";
89 case ARB_VS_REL_OFFSET
: return "rel_addr_const.y";
92 FIXME("Unmanaged %s shader helper constant requested: %u.\n",
93 shader
== WINED3D_SHADER_TYPE_PIXEL
? "pixel" : "vertex", value
);
96 case ARB_ZERO
: return "0.0";
97 case ARB_ONE
: return "1.0";
98 case ARB_TWO
: return "2.0";
99 case ARB_0001
: return "{0.0, 0.0, 0.0, 1.0}";
100 case ARB_EPS
: return "1e-8";
101 default: return "bad";
105 static inline BOOL
ffp_clip_emul(const struct wined3d_context
*context
)
107 return context
->lowest_disabled_stage
< 7;
110 /* ARB_program_shader private data */
129 struct wined3d_shader_loop_control loop_control
;
133 struct arb_ps_np2fixup_info
135 struct ps_np2fixup_info super
;
136 /* For ARB we need an offset value:
137 * With both GLSL and ARB mode the NP2 fixup information (the texture dimensions) are stored in a
138 * consecutive way (GLSL uses a uniform array). Since ARB doesn't know the notion of a "standalone"
139 * array we need an offset to the index inside the program local parameter array. */
143 struct arb_ps_compile_args
145 struct ps_compile_args super
;
147 WORD clip
; /* only a boolean, use a WORD for alignment */
148 unsigned char loop_ctrl
[WINED3D_MAX_CONSTS_I
][3];
151 struct stb_const_desc
153 unsigned char texunit
;
157 struct arb_ps_compiled_shader
159 struct arb_ps_compile_args args
;
160 struct arb_ps_np2fixup_info np2fixup_info
;
161 struct stb_const_desc bumpenvmatconst
[WINED3D_MAX_TEXTURES
];
162 struct stb_const_desc luminanceconst
[WINED3D_MAX_TEXTURES
];
163 UINT int_consts
[WINED3D_MAX_CONSTS_I
];
166 unsigned char numbumpenvmatconsts
;
170 struct arb_vs_compile_args
172 struct vs_compile_args super
;
178 unsigned char clip_texcoord
;
179 unsigned char clipplane_mask
;
181 DWORD boolclip_compare
;
183 unsigned int ps_signature
;
186 unsigned char samplers
[4];
187 DWORD samplers_compare
;
189 unsigned char loop_ctrl
[WINED3D_MAX_CONSTS_I
][3];
192 struct arb_vs_compiled_shader
194 struct arb_vs_compile_args args
;
196 UINT int_consts
[WINED3D_MAX_CONSTS_I
];
198 char need_color_unclamp
;
202 struct recorded_instruction
204 struct wined3d_shader_instruction ins
;
208 struct shader_arb_ctx_priv
213 /* plain GL_ARB_vertex_program or GL_ARB_fragment_program */
215 /* GL_NV_vertex_program2_option or GL_NV_fragment_program_option */
217 /* GL_NV_vertex_program3 or GL_NV_fragment_program2 */
221 const struct wined3d_gl_info
*gl_info
;
222 const struct arb_vs_compile_args
*cur_vs_args
;
223 const struct arb_ps_compile_args
*cur_ps_args
;
224 const struct arb_ps_compiled_shader
*compiled_fprog
;
225 const struct arb_vs_compiled_shader
*compiled_vprog
;
226 struct arb_ps_np2fixup_info
*cur_np2fixup_info
;
227 struct list control_frames
;
231 unsigned int num_loops
, loop_depth
, num_ifcs
;
233 BOOL ps_post_process
;
235 unsigned int vs_clipplanes
;
239 /* For 3.0 vertex shaders */
240 const char *vs_output
[MAX_REG_OUTPUT
];
241 /* For 2.x and earlier vertex shaders */
242 const char *texcrd_output
[8], *color_output
[2], *fog_output
;
244 /* 3.0 pshader input for compatibility with fixed function */
245 const char *ps_input
[MAX_REG_INPUT
];
250 struct wined3d_shader_signature sig
;
252 struct wine_rb_entry entry
;
255 struct arb_pshader_private
{
256 struct arb_ps_compiled_shader
*gl_shaders
;
257 UINT num_gl_shaders
, shader_array_size
;
258 unsigned int input_signature_idx
;
259 unsigned int clipplane_emulation
;
263 struct arb_vshader_private
{
264 struct arb_vs_compiled_shader
*gl_shaders
;
265 UINT num_gl_shaders
, shader_array_size
;
269 struct shader_arb_priv
271 GLuint current_vprogram_id
;
272 GLuint current_fprogram_id
;
273 const struct arb_ps_compiled_shader
*compiled_fprog
;
274 const struct arb_vs_compiled_shader
*compiled_vprog
;
275 BOOL use_arbfp_fixed_func
;
276 struct wine_rb_tree fragment_shaders
;
277 BOOL last_ps_const_clamped
;
278 BOOL last_vs_color_unclamp
;
280 struct wine_rb_tree signature_tree
;
283 unsigned int highest_dirty_ps_const
, highest_dirty_vs_const
;
284 char vshader_const_dirty
[WINED3D_MAX_VS_CONSTS_F
];
285 char pshader_const_dirty
[WINED3D_MAX_PS_CONSTS_F
];
286 const struct wined3d_context
*last_context
;
288 const struct wined3d_vertex_pipe_ops
*vertex_pipe
;
289 const struct wined3d_fragment_pipe_ops
*fragment_pipe
;
290 BOOL ffp_proj_control
;
293 /* Context activation for state handlers is done by the caller. */
295 static BOOL
need_rel_addr_const(const struct arb_vshader_private
*shader_data
,
296 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
)
298 if (shader_data
->rel_offset
) return TRUE
;
299 if (!reg_maps
->usesmova
) return FALSE
;
300 return !gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
];
303 /* Returns TRUE if result.clip from GL_NV_vertex_program2 should be used and FALSE otherwise */
304 static inline BOOL
use_nv_clip(const struct wined3d_gl_info
*gl_info
)
306 return gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]
307 && !(gl_info
->quirks
& WINED3D_QUIRK_NV_CLIP_BROKEN
);
310 static BOOL
need_helper_const(const struct arb_vshader_private
*shader_data
,
311 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
)
313 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
)) return TRUE
;
314 if (!gl_info
->supported
[NV_VERTEX_PROGRAM
]) return TRUE
; /* Need to init colors. */
315 if (gl_info
->quirks
& WINED3D_QUIRK_ARB_VS_OFFSET_LIMIT
) return TRUE
; /* Load the immval offset. */
316 if (!use_nv_clip(gl_info
)) return TRUE
; /* Init the clip texcoord */
317 if (reg_maps
->usesnrm
) return TRUE
; /* 0.0 */
318 if (reg_maps
->usespow
) return TRUE
; /* EPS, 0.0 and 1.0 */
319 if (reg_maps
->fog
) return TRUE
; /* Clamping fog coord, 0.0 and 1.0 */
323 static unsigned int reserved_vs_const(const struct arb_vshader_private
*shader_data
,
324 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
)
326 unsigned int ret
= 1;
327 /* We use one PARAM for the pos fixup, and in some cases one to load
328 * some immediate values into the shader. */
329 if (need_helper_const(shader_data
, reg_maps
, gl_info
)) ++ret
;
330 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
)) ++ret
;
334 /* Loads floating point constants into the currently set ARB_vertex/fragment_program.
335 * When constant_list == NULL, it will load all the constants.
337 * @target_type should be either GL_VERTEX_PROGRAM_ARB (for vertex shaders)
338 * or GL_FRAGMENT_PROGRAM_ARB (for pixel shaders)
340 /* Context activation is done by the caller. */
341 static unsigned int shader_arb_load_constants_f(const struct wined3d_shader
*shader
,
342 const struct wined3d_gl_info
*gl_info
, GLuint target_type
, unsigned int max_constants
,
343 const struct wined3d_vec4
*constants
, char *dirty_consts
)
345 struct wined3d_shader_lconst
*lconst
;
346 unsigned int ret
, i
, j
;
348 if (TRACE_ON(d3d_constants
))
350 for (i
= 0; i
< max_constants
; ++i
)
352 if (!dirty_consts
[i
])
354 TRACE_(d3d_constants
)("Loading constant %u: %s.\n", i
, debug_vec4(&constants
[i
]));
360 /* In 1.X pixel shaders constants are implicitly clamped in the range [-1;1] */
361 if (target_type
== GL_FRAGMENT_PROGRAM_ARB
&& shader
->reg_maps
.shader_version
.major
== 1)
364 /* ps 1.x supports only 8 constants, clamp only those. When switching between 1.x and higher
365 * shaders, the first 8 constants are marked dirty for reload
367 for (; i
< min(8, max_constants
); ++i
)
369 if (!dirty_consts
[i
])
373 if (constants
[i
].x
> 1.0f
)
375 else if (constants
[i
].x
< -1.0f
)
376 lcl_const
[0] = -1.0f
;
378 lcl_const
[0] = constants
[i
].x
;
380 if (constants
[i
].y
> 1.0f
)
382 else if (constants
[i
].y
< -1.0f
)
383 lcl_const
[1] = -1.0f
;
385 lcl_const
[1] = constants
[i
].y
;
387 if (constants
[i
].z
> 1.0f
)
389 else if (constants
[i
].z
< -1.0f
)
390 lcl_const
[2] = -1.0f
;
392 lcl_const
[2] = constants
[i
].z
;
394 if (constants
[i
].w
> 1.0f
)
396 else if (constants
[i
].w
< -1.0f
)
397 lcl_const
[3] = -1.0f
;
399 lcl_const
[3] = constants
[i
].w
;
401 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, i
, lcl_const
));
404 /* If further constants are dirty, reload them without clamping.
406 * The alternative is not to touch them, but then we cannot reset the dirty constant count
407 * to zero. That's bad for apps that only use PS 1.x shaders, because in that case the code
408 * above would always re-check the first 8 constants since max_constant remains at the init
413 if (gl_info
->supported
[EXT_GPU_PROGRAM_PARAMETERS
])
415 /* TODO: Benchmark if we're better of with finding the dirty constants ourselves,
416 * or just reloading *all* constants at once
418 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type, i, max_constants, constants + (i * 4)));
420 for (; i
< max_constants
; ++i
)
422 if (!dirty_consts
[i
])
425 /* Find the next block of dirty constants */
428 for (++i
; (i
< max_constants
) && dirty_consts
[i
]; ++i
)
433 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type
, j
, i
- j
, &constants
[j
].x
));
438 for (; i
< max_constants
; ++i
)
443 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, i
, &constants
[i
].x
));
447 checkGLcall("glProgramEnvParameter4fvARB()");
449 /* Load immediate constants */
450 if (shader
->load_local_constsF
)
452 if (TRACE_ON(d3d_shader
))
454 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
456 GLfloat
* values
= (GLfloat
*)lconst
->value
;
457 TRACE_(d3d_constants
)("Loading local constants %i: %f, %f, %f, %f\n", lconst
->idx
,
458 values
[0], values
[1], values
[2], values
[3]);
461 /* Immediate constants are clamped for 1.X shaders at loading times */
463 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
465 dirty_consts
[lconst
->idx
] = 1; /* Dirtify so the non-immediate constant overwrites it next time */
466 ret
= max(ret
, lconst
->idx
+ 1);
467 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, lconst
->idx
, (GLfloat
*)lconst
->value
));
469 checkGLcall("glProgramEnvParameter4fvARB()");
470 return ret
; /* The loaded immediate constants need reloading for the next shader */
472 return 0; /* No constants are dirty now */
476 /* Loads the texture dimensions for NP2 fixup into the currently set
477 * ARB_[vertex/fragment]_programs. */
478 static void shader_arb_load_np2fixup_constants(const struct arb_ps_np2fixup_info
*fixup
,
479 const struct wined3d_gl_info
*gl_info
, const struct wined3d_state
*state
)
481 GLfloat np2fixup_constants
[4 * WINED3D_MAX_FRAGMENT_SAMPLERS
];
482 uint32_t active
= fixup
->super
.active
;
483 const struct wined3d_texture
*tex
;
493 i
= wined3d_bit_scan(&active
);
494 if (!(tex
= state
->textures
[i
]))
496 ERR("Nonexistent texture is flagged for NP2 texcoord fixup.\n");
500 idx
= fixup
->super
.idx
[i
];
501 tex_dim
= &np2fixup_constants
[(idx
>> 1) * 4];
504 tex_dim
[2] = tex
->pow2_matrix
[0];
505 tex_dim
[3] = tex
->pow2_matrix
[5];
509 tex_dim
[0] = tex
->pow2_matrix
[0];
510 tex_dim
[1] = tex
->pow2_matrix
[5];
514 for (i
= 0; i
< fixup
->super
.num_consts
; ++i
)
516 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
517 fixup
->offset
+ i
, &np2fixup_constants
[i
* 4]));
521 /* Context activation is done by the caller. */
522 static void shader_arb_ps_local_constants(const struct arb_ps_compiled_shader
*gl_shader
,
523 const struct wined3d_context_gl
*context_gl
, const struct wined3d_state
*state
, unsigned int rt_height
)
525 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
528 for(i
= 0; i
< gl_shader
->numbumpenvmatconsts
; i
++)
530 int texunit
= gl_shader
->bumpenvmatconst
[i
].texunit
;
532 /* The state manager takes care that this function is always called if the bump env matrix changes */
533 const float *data
= (const float *)&state
->texture_states
[texunit
][WINED3D_TSS_BUMPENV_MAT00
];
534 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
535 gl_shader
->bumpenvmatconst
[i
].const_num
, data
));
537 if (gl_shader
->luminanceconst
[i
].const_num
!= WINED3D_CONST_NUM_UNUSED
)
539 /* WINED3D_TSS_BUMPENVLSCALE and WINED3D_TSS_BUMPENVLOFFSET are next to each other.
540 * point gl to the scale, and load 4 floats. x = scale, y = offset, z and w are junk, we
541 * don't care about them. The pointers are valid for sure because the stateblock is bigger.
542 * (they're WINED3D_TSS_TEXTURETRANSFORMFLAGS and WINED3D_TSS_ADDRESSW, so most likely 0 or NaN
544 const float *scale
= (const float *)&state
->texture_states
[texunit
][WINED3D_TSS_BUMPENV_LSCALE
];
545 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
546 gl_shader
->luminanceconst
[i
].const_num
, scale
));
549 checkGLcall("Load bumpmap consts");
551 if(gl_shader
->ycorrection
!= WINED3D_CONST_NUM_UNUSED
)
553 /* ycorrection.x: Backbuffer height(onscreen) or 0(offscreen).
554 * ycorrection.y: -1.0(onscreen), 1.0(offscreen)
559 val
[0] = context_gl
->c
.render_offscreen
? 0.0f
: (float)rt_height
;
560 val
[1] = context_gl
->c
.render_offscreen
? 1.0f
: -1.0f
;
563 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, gl_shader
->ycorrection
, val
));
564 checkGLcall("y correction loading");
567 if (!gl_shader
->num_int_consts
) return;
569 for (i
= 0; i
< WINED3D_MAX_CONSTS_I
; ++i
)
571 if(gl_shader
->int_consts
[i
] != WINED3D_CONST_NUM_UNUSED
)
574 val
[0] = (float)state
->ps_consts_i
[i
].x
;
575 val
[1] = (float)state
->ps_consts_i
[i
].y
;
576 val
[2] = (float)state
->ps_consts_i
[i
].z
;
579 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, gl_shader
->int_consts
[i
], val
));
582 checkGLcall("Load ps int consts");
585 /* Context activation is done by the caller. */
586 static void shader_arb_vs_local_constants(const struct arb_vs_compiled_shader
*gl_shader
,
587 const struct wined3d_context_gl
*context_gl
, const struct wined3d_state
*state
)
589 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
590 float position_fixup
[4];
593 /* Upload the position fixup */
594 shader_get_position_fixup(&context_gl
->c
, state
, 1, position_fixup
);
595 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, gl_shader
->pos_fixup
, position_fixup
));
597 if (!gl_shader
->num_int_consts
) return;
599 for (i
= 0; i
< WINED3D_MAX_CONSTS_I
; ++i
)
601 if(gl_shader
->int_consts
[i
] != WINED3D_CONST_NUM_UNUSED
)
604 val
[0] = (float)state
->vs_consts_i
[i
].x
;
605 val
[1] = (float)state
->vs_consts_i
[i
].y
;
606 val
[2] = (float)state
->vs_consts_i
[i
].z
;
609 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, gl_shader
->int_consts
[i
], val
));
612 checkGLcall("Load vs int consts");
615 static void shader_arb_select(void *shader_priv
, struct wined3d_context
*context
,
616 const struct wined3d_state
*state
);
619 * Loads the app-supplied constants into the currently set ARB_[vertex/fragment]_programs.
621 * We only support float constants in ARB at the moment, so don't
622 * worry about the Integers or Booleans
624 /* Context activation is done by the caller (state handler). */
625 static void shader_arb_load_constants_internal(struct shader_arb_priv
*priv
, struct wined3d_context_gl
*context_gl
,
626 const struct wined3d_state
*state
, BOOL use_ps
, BOOL use_vs
, BOOL from_shader_select
)
628 const struct wined3d_d3d_info
*d3d_info
= context_gl
->c
.d3d_info
;
629 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
631 if (!from_shader_select
)
633 const struct wined3d_shader
*vshader
= state
->shader
[WINED3D_SHADER_TYPE_VERTEX
];
634 const struct wined3d_shader
*pshader
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
637 && (vshader
->reg_maps
.boolean_constants
638 || (!gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]
639 && (vshader
->reg_maps
.integer_constants
& ~vshader
->reg_maps
.local_int_consts
))))
641 TRACE("bool/integer vertex shader constants potentially modified, forcing shader reselection.\n");
642 shader_arb_select(priv
, &context_gl
->c
, state
);
645 && (pshader
->reg_maps
.boolean_constants
646 || (!gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
]
647 && (pshader
->reg_maps
.integer_constants
& ~pshader
->reg_maps
.local_int_consts
))))
649 TRACE("bool/integer pixel shader constants potentially modified, forcing shader reselection.\n");
650 shader_arb_select(priv
, &context_gl
->c
, state
);
654 if (&context_gl
->c
!= priv
->last_context
)
656 memset(priv
->vshader_const_dirty
, 1,
657 sizeof(*priv
->vshader_const_dirty
) * d3d_info
->limits
.vs_uniform_count
);
658 priv
->highest_dirty_vs_const
= d3d_info
->limits
.vs_uniform_count
;
660 memset(priv
->pshader_const_dirty
, 1,
661 sizeof(*priv
->pshader_const_dirty
) * d3d_info
->limits
.ps_uniform_count
);
662 priv
->highest_dirty_ps_const
= d3d_info
->limits
.ps_uniform_count
;
664 priv
->last_context
= &context_gl
->c
;
669 const struct wined3d_shader
*vshader
= state
->shader
[WINED3D_SHADER_TYPE_VERTEX
];
670 const struct arb_vs_compiled_shader
*gl_shader
= priv
->compiled_vprog
;
672 /* Load DirectX 9 float constants for vertex shader */
673 priv
->highest_dirty_vs_const
= shader_arb_load_constants_f(vshader
, gl_info
, GL_VERTEX_PROGRAM_ARB
,
674 priv
->highest_dirty_vs_const
, state
->vs_consts_f
, priv
->vshader_const_dirty
);
675 shader_arb_vs_local_constants(gl_shader
, context_gl
, state
);
680 const struct wined3d_shader
*pshader
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
681 const struct arb_ps_compiled_shader
*gl_shader
= priv
->compiled_fprog
;
682 UINT rt_height
= state
->fb
.render_targets
[0]->height
;
684 /* Load DirectX 9 float constants for pixel shader */
685 priv
->highest_dirty_ps_const
= shader_arb_load_constants_f(pshader
, gl_info
, GL_FRAGMENT_PROGRAM_ARB
,
686 priv
->highest_dirty_ps_const
, state
->ps_consts_f
, priv
->pshader_const_dirty
);
687 shader_arb_ps_local_constants(gl_shader
, context_gl
, state
, rt_height
);
689 if (context_gl
->c
.constant_update_mask
& WINED3D_SHADER_CONST_PS_NP2_FIXUP
)
690 shader_arb_load_np2fixup_constants(&gl_shader
->np2fixup_info
, gl_info
, state
);
694 static void shader_arb_load_constants(void *shader_priv
, struct wined3d_context
*context
,
695 const struct wined3d_state
*state
)
697 shader_arb_load_constants_internal(shader_priv
, wined3d_context_gl(context
),
698 state
, use_ps(state
), use_vs(state
), FALSE
);
701 static void shader_arb_update_float_vertex_constants(struct wined3d_device
*device
, UINT start
, UINT count
)
703 struct wined3d_context_gl
*context_gl
= wined3d_context_gl_get_current();
704 struct shader_arb_priv
*priv
= device
->shader_priv
;
706 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
707 * context. On a context switch the old context will be fully dirtified */
708 if (!context_gl
|| context_gl
->c
.device
!= device
)
711 memset(priv
->vshader_const_dirty
+ start
, 1, sizeof(*priv
->vshader_const_dirty
) * count
);
712 priv
->highest_dirty_vs_const
= max(priv
->highest_dirty_vs_const
, start
+ count
);
715 static void shader_arb_update_float_pixel_constants(struct wined3d_device
*device
, UINT start
, UINT count
)
717 struct wined3d_context_gl
*context_gl
= wined3d_context_gl_get_current();
718 struct shader_arb_priv
*priv
= device
->shader_priv
;
720 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
721 * context. On a context switch the old context will be fully dirtified */
722 if (!context_gl
|| context_gl
->c
.device
!= device
)
725 memset(priv
->pshader_const_dirty
+ start
, 1, sizeof(*priv
->pshader_const_dirty
) * count
);
726 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, start
+ count
);
729 static void shader_arb_append_imm_vec4(struct wined3d_string_buffer
*buffer
, const float *values
)
733 wined3d_ftoa(values
[0], str
[0]);
734 wined3d_ftoa(values
[1], str
[1]);
735 wined3d_ftoa(values
[2], str
[2]);
736 wined3d_ftoa(values
[3], str
[3]);
737 shader_addline(buffer
, "{%s, %s, %s, %s}", str
[0], str
[1], str
[2], str
[3]);
740 /* Generate the variable & register declarations for the ARB_vertex_program output target */
741 static void shader_generate_arb_declarations(const struct wined3d_shader
*shader
,
742 const struct wined3d_shader_reg_maps
*reg_maps
, struct wined3d_string_buffer
*buffer
,
743 const struct wined3d_gl_info
*gl_info
, unsigned int *num_clipplanes
,
744 const struct shader_arb_ctx_priv
*ctx
)
747 char pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
748 const struct wined3d_shader_lconst
*lconst
;
749 unsigned max_constantsF
;
752 /* In pixel shaders, all private constants are program local, we don't need anything
753 * from program.env. Thus we can advertise the full set of constants in pixel shaders.
754 * If we need a private constant the GL implementation will squeeze it in somewhere
756 * With vertex shaders we need the posFixup and on some GL implementations 4 helper
757 * immediate values. The posFixup is loaded using program.env for now, so always
758 * subtract one from the number of constants. If the shader uses indirect addressing,
759 * account for the helper const too because we have to declare all available d3d constants
760 * and don't know which are actually used.
764 max_constantsF
= gl_info
->limits
.arb_ps_native_constants
;
765 /* 24 is the minimum MAX_PROGRAM_ENV_PARAMETERS_ARB value. */
766 if (max_constantsF
< 24)
767 max_constantsF
= gl_info
->limits
.arb_ps_float_constants
;
771 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
772 max_constantsF
= gl_info
->limits
.arb_vs_native_constants
;
773 /* 96 is the minimum MAX_PROGRAM_ENV_PARAMETERS_ARB value.
774 * Also prevents max_constantsF from becoming less than 0 and
776 if (max_constantsF
< 96)
777 max_constantsF
= gl_info
->limits
.arb_vs_float_constants
;
779 if (reg_maps
->usesrelconstF
)
781 DWORD highest_constf
= 0, clip_limit
;
783 max_constantsF
-= reserved_vs_const(shader_data
, reg_maps
, gl_info
);
784 max_constantsF
-= wined3d_popcount(reg_maps
->integer_constants
);
785 max_constantsF
-= gl_info
->reserved_arb_constants
;
787 for (i
= 0; i
< shader
->limits
->constant_float
; ++i
)
789 if (wined3d_bitmap_is_set(reg_maps
->constf
, i
))
793 if(use_nv_clip(gl_info
) && ctx
->target_version
>= NV2
)
795 if(ctx
->cur_vs_args
->super
.clip_enabled
)
796 clip_limit
= gl_info
->limits
.user_clip_distances
;
802 unsigned int mask
= ctx
->cur_vs_args
->clip
.boolclip
.clipplane_mask
;
803 clip_limit
= min(wined3d_popcount(mask
), 4);
805 *num_clipplanes
= min(clip_limit
, max_constantsF
- highest_constf
- 1);
806 max_constantsF
-= *num_clipplanes
;
807 if(*num_clipplanes
< clip_limit
)
809 WARN("Only %u clip planes out of %u enabled.\n", *num_clipplanes
,
810 gl_info
->limits
.user_clip_distances
);
815 if (ctx
->target_version
>= NV2
)
816 *num_clipplanes
= gl_info
->limits
.user_clip_distances
;
818 *num_clipplanes
= min(gl_info
->limits
.user_clip_distances
, 4);
822 map
= reg_maps
->temporary
;
825 i
= wined3d_bit_scan(&map
);
826 shader_addline(buffer
, "TEMP R%u;\n", i
);
829 map
= reg_maps
->address
;
832 i
= wined3d_bit_scan(&map
);
833 shader_addline(buffer
, "ADDRESS A%u;\n", i
);
836 if (pshader
&& reg_maps
->shader_version
.major
== 1 && reg_maps
->shader_version
.minor
<= 3)
838 map
= reg_maps
->texcoord
;
841 i
= wined3d_bit_scan(&map
);
842 shader_addline(buffer
, "TEMP T%u;\n", i
);
846 if (!shader
->load_local_constsF
)
848 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
851 value
= (const float *)lconst
->value
;
852 shader_addline(buffer
, "PARAM C%u = ", lconst
->idx
);
853 shader_arb_append_imm_vec4(buffer
, value
);
854 shader_addline(buffer
, ";\n");
858 /* After subtracting privately used constants from the hardware limit(they are loaded as
859 * local constants), make sure the shader doesn't violate the env constant limit
863 max_constantsF
= min(max_constantsF
, gl_info
->limits
.arb_ps_float_constants
);
867 max_constantsF
= min(max_constantsF
, gl_info
->limits
.arb_vs_float_constants
);
870 /* Avoid declaring more constants than needed */
871 max_constantsF
= min(max_constantsF
, shader
->limits
->constant_float
);
873 /* we use the array-based constants array if the local constants are marked for loading,
874 * because then we use indirect addressing, or when the local constant list is empty,
875 * because then we don't know if we're using indirect addressing or not. If we're hardcoding
876 * local constants do not declare the loaded constants as an array because ARB compilers usually
877 * do not optimize unused constants away
879 if (reg_maps
->usesrelconstF
)
881 /* Need to PARAM the environment parameters (constants) so we can use relative addressing */
882 shader_addline(buffer
, "PARAM C[%d] = { program.env[0..%d] };\n",
883 max_constantsF
, max_constantsF
- 1);
887 for (i
= 0; i
< max_constantsF
; ++i
)
889 if (!shader_constant_is_local(shader
, i
) && wined3d_extract_bits(reg_maps
->constf
, i
, 1))
891 shader_addline(buffer
, "PARAM C%d = program.env[%d];\n",i
, i
);
897 static const char * const shift_tab
[] = {
898 "dummy", /* 0 (none) */
899 "coefmul.x", /* 1 (x2) */
900 "coefmul.y", /* 2 (x4) */
901 "coefmul.z", /* 3 (x8) */
902 "coefmul.w", /* 4 (x16) */
903 "dummy", /* 5 (x32) */
904 "dummy", /* 6 (x64) */
905 "dummy", /* 7 (x128) */
906 "dummy", /* 8 (d256) */
907 "dummy", /* 9 (d128) */
908 "dummy", /* 10 (d64) */
909 "dummy", /* 11 (d32) */
910 "coefdiv.w", /* 12 (d16) */
911 "coefdiv.z", /* 13 (d8) */
912 "coefdiv.y", /* 14 (d4) */
913 "coefdiv.x" /* 15 (d2) */
916 static void shader_arb_get_write_mask(const struct wined3d_shader_instruction
*ins
,
917 const struct wined3d_shader_dst_param
*dst
, char *write_mask
)
919 char *ptr
= write_mask
;
921 if (dst
->write_mask
!= WINED3DSP_WRITEMASK_ALL
)
924 if (dst
->write_mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
925 if (dst
->write_mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
926 if (dst
->write_mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
927 if (dst
->write_mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
933 static void shader_arb_get_swizzle(const struct wined3d_shader_src_param
*param
, BOOL fixup
, char *swizzle_str
)
935 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
936 * but addressed as "rgba". To fix this we need to swap the register's x
937 * and z components. */
938 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
939 char *ptr
= swizzle_str
;
941 /* swizzle bits fields: wwzzyyxx */
942 DWORD swizzle
= param
->swizzle
;
943 DWORD swizzle_x
= swizzle
& 0x03;
944 DWORD swizzle_y
= (swizzle
>> 2) & 0x03;
945 DWORD swizzle_z
= (swizzle
>> 4) & 0x03;
946 DWORD swizzle_w
= (swizzle
>> 6) & 0x03;
948 /* If the swizzle is the default swizzle (ie, "xyzw"), we don't need to
949 * generate a swizzle string. Unless we need to our own swizzling. */
950 if (swizzle
!= WINED3DSP_NOSWIZZLE
|| fixup
)
953 if (swizzle_x
== swizzle_y
&& swizzle_x
== swizzle_z
&& swizzle_x
== swizzle_w
) {
954 *ptr
++ = swizzle_chars
[swizzle_x
];
956 *ptr
++ = swizzle_chars
[swizzle_x
];
957 *ptr
++ = swizzle_chars
[swizzle_y
];
958 *ptr
++ = swizzle_chars
[swizzle_z
];
959 *ptr
++ = swizzle_chars
[swizzle_w
];
966 static void shader_arb_request_a0(const struct wined3d_shader_instruction
*ins
, const char *src
)
968 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
969 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
971 if (!strcmp(priv
->addr_reg
, src
))
974 snprintf(priv
->addr_reg
, sizeof(priv
->addr_reg
), "%s", src
);
975 shader_addline(buffer
, "ARL A0.x, %s;\n", src
);
978 static void shader_arb_get_src_param(const struct wined3d_shader_instruction
*ins
,
979 const struct wined3d_shader_src_param
*src
, unsigned int tmpreg
, char *outregstr
);
981 static void shader_arb_get_register_name(const struct wined3d_shader_instruction
*ins
,
982 const struct wined3d_shader_register
*reg
, char *register_name
, BOOL
*is_color
)
984 /* oPos, oFog and oPts in D3D */
985 static const char * const rastout_reg_names
[] = {"TMP_OUT", "TMP_FOGCOORD", "result.pointsize"};
986 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
987 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
988 BOOL pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
989 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
995 case WINED3DSPR_TEMP
:
996 sprintf(register_name
, "R%u", reg
->idx
[0].offset
);
999 case WINED3DSPR_INPUT
:
1002 if (reg_maps
->shader_version
.major
< 3)
1004 if (!reg
->idx
[0].offset
)
1005 strcpy(register_name
, "fragment.color.primary");
1007 strcpy(register_name
, "fragment.color.secondary");
1011 if (reg
->idx
[0].rel_addr
)
1014 shader_arb_get_src_param(ins
, reg
->idx
[0].rel_addr
, 0, rel_reg
);
1016 if (!strcmp(rel_reg
, "**aL_emul**"))
1018 unsigned int idx
= ctx
->aL
+ reg
->idx
[0].offset
;
1019 if(idx
< MAX_REG_INPUT
)
1021 strcpy(register_name
, ctx
->ps_input
[idx
]);
1025 ERR("Pixel shader input register out of bounds: %u\n", idx
);
1026 sprintf(register_name
, "out_of_bounds_%u", idx
);
1029 else if (reg_maps
->input_registers
& 0x0300)
1031 /* There are two ways basically:
1033 * 1) Use the unrolling code that is used for loop emulation and unroll the loop.
1034 * That means trouble if the loop also contains a breakc or if the control values
1035 * aren't local constants.
1036 * 2) Generate an if block that checks if aL.y < 8, == 8 or == 9 and selects the
1037 * source dynamically. The trouble is that we cannot simply read aL.y because it
1038 * is an ADDRESS register. We could however push it, load .zw with a value and use
1039 * ADAC to load the condition code register and pop it again afterwards
1041 FIXME("Relative input register addressing with more than 8 registers\n");
1043 /* This is better than nothing for now */
1044 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1046 else if(ctx
->cur_ps_args
->super
.vp_mode
!= WINED3D_VP_MODE_SHADER
)
1048 /* This is problematic because we'd have to consult the ctx->ps_input strings
1049 * for where to find the varying. Some may be "0.0", others can be texcoords or
1050 * colors. This needs either a pipeline replacement to make the vertex shader feed
1051 * proper varyings, or loop unrolling
1053 * For now use the texcoords and hope for the best
1055 FIXME("Non-vertex shader varying input with indirect addressing\n");
1056 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1060 /* D3D supports indirect addressing only with aL in loop registers. The loop instruction
1061 * pulls GL_NV_fragment_program2 in
1063 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1068 if (reg
->idx
[0].offset
< MAX_REG_INPUT
)
1070 strcpy(register_name
, ctx
->ps_input
[reg
->idx
[0].offset
]);
1074 ERR("Pixel shader input register out of bounds: %u\n", reg
->idx
[0].offset
);
1075 sprintf(register_name
, "out_of_bounds_%u", reg
->idx
[0].offset
);
1082 if (ctx
->cur_vs_args
->super
.swizzle_map
& (1u << reg
->idx
[0].offset
))
1084 sprintf(register_name
, "vertex.attrib[%u]", reg
->idx
[0].offset
);
1088 case WINED3DSPR_CONST
:
1089 if (!pshader
&& reg
->idx
[0].rel_addr
)
1091 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1092 UINT rel_offset
= ctx
->target_version
== ARB
? shader_data
->rel_offset
: 0;
1095 if (reg_maps
->shader_version
.major
< 2)
1097 sprintf(rel_reg
, "A0.x");
1101 shader_arb_get_src_param(ins
, reg
->idx
[0].rel_addr
, 0, rel_reg
);
1102 if (ctx
->target_version
== ARB
)
1104 if (!strcmp(rel_reg
, "**aL_emul**"))
1108 shader_arb_request_a0(ins
, rel_reg
);
1109 sprintf(rel_reg
, "A0.x");
1114 sprintf(register_name
, "C[%u]", ctx
->aL
+ reg
->idx
[0].offset
);
1115 else if (reg
->idx
[0].offset
>= rel_offset
)
1116 sprintf(register_name
, "C[%s + %u]", rel_reg
, reg
->idx
[0].offset
- rel_offset
);
1118 sprintf(register_name
, "C[%s - %u]", rel_reg
, rel_offset
- reg
->idx
[0].offset
);
1122 if (reg_maps
->usesrelconstF
)
1123 sprintf(register_name
, "C[%u]", reg
->idx
[0].offset
);
1125 sprintf(register_name
, "C%u", reg
->idx
[0].offset
);
1129 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
1132 if (reg_maps
->shader_version
.major
== 1
1133 && reg_maps
->shader_version
.minor
<= 3)
1134 /* In ps <= 1.3, Tx is a temporary register as destination
1135 * to all instructions, and as source to most instructions.
1136 * For some instructions it is the texcoord input. Those
1137 * instructions know about the special use. */
1138 sprintf(register_name
, "T%u", reg
->idx
[0].offset
);
1140 /* In ps 1.4 and 2.x Tx is always a (read-only) varying. */
1141 sprintf(register_name
, "fragment.texcoord[%u]", reg
->idx
[0].offset
);
1145 if (reg_maps
->shader_version
.major
== 1 || ctx
->target_version
>= NV2
)
1146 sprintf(register_name
, "A%u", reg
->idx
[0].offset
);
1148 sprintf(register_name
, "A%u_SHADOW", reg
->idx
[0].offset
);
1152 case WINED3DSPR_COLOROUT
:
1153 if (ctx
->ps_post_process
&& !reg
->idx
[0].offset
)
1155 strcpy(register_name
, "TMP_COLOR");
1159 if (ctx
->cur_ps_args
->super
.srgb_correction
)
1160 FIXME("sRGB correction on higher render targets.\n");
1161 if (reg_maps
->rt_mask
> 1)
1162 sprintf(register_name
, "result.color[%u]", reg
->idx
[0].offset
);
1164 strcpy(register_name
, "result.color");
1168 case WINED3DSPR_RASTOUT
:
1169 if (reg
->idx
[0].offset
== 1)
1170 sprintf(register_name
, "%s", ctx
->fog_output
);
1172 sprintf(register_name
, "%s", rastout_reg_names
[reg
->idx
[0].offset
]);
1175 case WINED3DSPR_DEPTHOUT
:
1176 strcpy(register_name
, "result.depth");
1179 case WINED3DSPR_ATTROUT
:
1180 /* case WINED3DSPR_OUTPUT: */
1182 sprintf(register_name
, "oD[%u]", reg
->idx
[0].offset
);
1184 strcpy(register_name
, ctx
->color_output
[reg
->idx
[0].offset
]);
1187 case WINED3DSPR_TEXCRDOUT
:
1189 sprintf(register_name
, "oT[%u]", reg
->idx
[0].offset
);
1190 else if (reg_maps
->shader_version
.major
< 3)
1191 strcpy(register_name
, ctx
->texcrd_output
[reg
->idx
[0].offset
]);
1193 strcpy(register_name
, ctx
->vs_output
[reg
->idx
[0].offset
]);
1196 case WINED3DSPR_LOOP
:
1197 if(ctx
->target_version
>= NV2
)
1199 /* Pshader has an implicitly declared loop index counter A0.x that cannot be renamed */
1200 if(pshader
) sprintf(register_name
, "A0.x");
1201 else sprintf(register_name
, "aL.y");
1205 /* Unfortunately this code cannot return the value of ctx->aL here. An immediate value
1206 * would be valid, but if aL is used for indexing(its only use), there's likely an offset,
1207 * thus the result would be something like C[15 + 30], which is not valid in the ARB program
1208 * grammar. So return a marker for the emulated aL and intercept it in constant and varying
1211 sprintf(register_name
, "**aL_emul**");
1216 case WINED3DSPR_CONSTINT
:
1217 sprintf(register_name
, "I%u", reg
->idx
[0].offset
);
1220 case WINED3DSPR_MISCTYPE
:
1221 if (!reg
->idx
[0].offset
)
1222 sprintf(register_name
, "vpos");
1223 else if (reg
->idx
[0].offset
== 1)
1224 sprintf(register_name
, "fragment.facing.x");
1226 FIXME("Unknown MISCTYPE register index %u.\n", reg
->idx
[0].offset
);
1230 FIXME("Unhandled register type %#x[%u].\n", reg
->type
, reg
->idx
[0].offset
);
1231 sprintf(register_name
, "unrecognized_register[%u]", reg
->idx
[0].offset
);
1236 static void shader_arb_get_dst_param(const struct wined3d_shader_instruction
*ins
,
1237 const struct wined3d_shader_dst_param
*wined3d_dst
, char *str
)
1239 char register_name
[255];
1243 shader_arb_get_register_name(ins
, &wined3d_dst
->reg
, register_name
, &is_color
);
1244 strcpy(str
, register_name
);
1246 shader_arb_get_write_mask(ins
, wined3d_dst
, write_mask
);
1247 strcat(str
, write_mask
);
1250 static const char *shader_arb_get_fixup_swizzle(enum fixup_channel_source channel_source
)
1252 switch(channel_source
)
1254 case CHANNEL_SOURCE_ZERO
: return "0";
1255 case CHANNEL_SOURCE_ONE
: return "1";
1256 case CHANNEL_SOURCE_X
: return "x";
1257 case CHANNEL_SOURCE_Y
: return "y";
1258 case CHANNEL_SOURCE_Z
: return "z";
1259 case CHANNEL_SOURCE_W
: return "w";
1261 FIXME("Unhandled channel source %#x\n", channel_source
);
1266 struct color_fixup_masks
1272 static struct color_fixup_masks
calc_color_correction(struct color_fixup_desc fixup
, DWORD dst_mask
)
1274 struct color_fixup_masks masks
= {0, 0};
1276 if (is_complex_fixup(fixup
))
1278 enum complex_fixup complex_fixup
= get_complex_fixup(fixup
);
1279 FIXME("Complex fixup (%#x) not supported\n", complex_fixup
);
1283 if (fixup
.x_source
!= CHANNEL_SOURCE_X
)
1284 masks
.source
|= WINED3DSP_WRITEMASK_0
;
1285 if (fixup
.y_source
!= CHANNEL_SOURCE_Y
)
1286 masks
.source
|= WINED3DSP_WRITEMASK_1
;
1287 if (fixup
.z_source
!= CHANNEL_SOURCE_Z
)
1288 masks
.source
|= WINED3DSP_WRITEMASK_2
;
1289 if (fixup
.w_source
!= CHANNEL_SOURCE_W
)
1290 masks
.source
|= WINED3DSP_WRITEMASK_3
;
1291 masks
.source
&= dst_mask
;
1293 if (fixup
.x_sign_fixup
)
1294 masks
.sign
|= WINED3DSP_WRITEMASK_0
;
1295 if (fixup
.y_sign_fixup
)
1296 masks
.sign
|= WINED3DSP_WRITEMASK_1
;
1297 if (fixup
.z_sign_fixup
)
1298 masks
.sign
|= WINED3DSP_WRITEMASK_2
;
1299 if (fixup
.w_sign_fixup
)
1300 masks
.sign
|= WINED3DSP_WRITEMASK_3
;
1301 masks
.sign
&= dst_mask
;
1306 static void gen_color_correction(struct wined3d_string_buffer
*buffer
, const char *dst
,
1307 const char *src
, const char *one
, const char *two
,
1308 struct color_fixup_desc fixup
, struct color_fixup_masks masks
)
1310 const char *sign_fixup_src
= dst
;
1315 sign_fixup_src
= "TA";
1317 shader_addline(buffer
, "SWZ %s, %s, %s, %s, %s, %s;\n", sign_fixup_src
, src
,
1318 shader_arb_get_fixup_swizzle(fixup
.x_source
), shader_arb_get_fixup_swizzle(fixup
.y_source
),
1319 shader_arb_get_fixup_swizzle(fixup
.z_source
), shader_arb_get_fixup_swizzle(fixup
.w_source
));
1321 else if (masks
.sign
)
1323 sign_fixup_src
= src
;
1329 char *ptr
= reg_mask
;
1331 if (masks
.sign
!= WINED3DSP_WRITEMASK_ALL
)
1334 if (masks
.sign
& WINED3DSP_WRITEMASK_0
)
1336 if (masks
.sign
& WINED3DSP_WRITEMASK_1
)
1338 if (masks
.sign
& WINED3DSP_WRITEMASK_2
)
1340 if (masks
.sign
& WINED3DSP_WRITEMASK_3
)
1345 shader_addline(buffer
, "MAD %s%s, %s, %s, -%s;\n", dst
, reg_mask
, sign_fixup_src
, two
, one
);
1349 static const char *shader_arb_get_modifier(const struct wined3d_shader_instruction
*ins
)
1352 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1353 if (!ins
->dst_count
) return "";
1355 mod
= ins
->dst
[0].modifiers
;
1357 /* Silently ignore PARTIALPRECISION if it's not supported */
1358 if(priv
->target_version
== ARB
) mod
&= ~WINED3DSPDM_PARTIALPRECISION
;
1360 if(mod
& WINED3DSPDM_MSAMPCENTROID
)
1362 FIXME("Unhandled modifier WINED3DSPDM_MSAMPCENTROID\n");
1363 mod
&= ~WINED3DSPDM_MSAMPCENTROID
;
1368 case WINED3DSPDM_SATURATE
| WINED3DSPDM_PARTIALPRECISION
:
1371 case WINED3DSPDM_SATURATE
:
1374 case WINED3DSPDM_PARTIALPRECISION
:
1381 FIXME("Unknown modifiers 0x%08x\n", mod
);
1386 #define TEX_PROJ 0x1
1387 #define TEX_BIAS 0x2
1389 #define TEX_DERIV 0x10
1391 static void shader_hw_sample(const struct wined3d_shader_instruction
*ins
, unsigned int sampler_idx
,
1392 const char *dst_str
, const char *coord_reg
, WORD flags
, const char *dsx
, const char *dsy
)
1394 BOOL pshader
= shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
1395 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1396 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1397 enum wined3d_shader_resource_type resource_type
;
1398 struct color_fixup_masks masks
;
1399 const char *tex_dst
= dst_str
;
1400 BOOL np2_fixup
= FALSE
;
1401 const char *tex_type
;
1406 resource_type
= pixelshader_get_resource_type(ins
->ctx
->reg_maps
, sampler_idx
,
1407 priv
->cur_ps_args
->super
.tex_types
);
1411 resource_type
= ins
->ctx
->reg_maps
->resource_info
[sampler_idx
].type
;
1413 /* D3D vertex shader sampler IDs are vertex samplers(0-3), not global d3d samplers */
1414 sampler_idx
+= WINED3D_MAX_FRAGMENT_SAMPLERS
;
1417 switch (resource_type
)
1419 case WINED3D_SHADER_RESOURCE_TEXTURE_1D
:
1423 case WINED3D_SHADER_RESOURCE_TEXTURE_2D
:
1424 if (pshader
&& priv
->cur_ps_args
->super
.np2_fixup
& (1u << sampler_idx
)
1425 && priv
->gl_info
->supported
[ARB_TEXTURE_RECTANGLE
])
1432 if (priv
->cur_np2fixup_info
->super
.active
& (1u << sampler_idx
))
1434 if (flags
) FIXME("Only ordinary sampling from NP2 textures is supported.\n");
1435 else np2_fixup
= TRUE
;
1440 case WINED3D_SHADER_RESOURCE_TEXTURE_3D
:
1444 case WINED3D_SHADER_RESOURCE_TEXTURE_CUBE
:
1449 ERR("Unexpected resource type %#x.\n", resource_type
);
1453 /* TEX, TXL, TXD and TXP do not support the "H" modifier,
1454 * so don't use shader_arb_get_modifier
1456 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
) mod
= "_SAT";
1459 /* Fragment samplers always have identity mapping */
1460 if(sampler_idx
>= WINED3D_MAX_FRAGMENT_SAMPLERS
)
1462 sampler_idx
= priv
->cur_vs_args
->vertex
.samplers
[sampler_idx
- WINED3D_MAX_FRAGMENT_SAMPLERS
];
1467 masks
= calc_color_correction(priv
->cur_ps_args
->super
.color_fixup
[sampler_idx
],
1468 ins
->dst
[0].write_mask
);
1470 if (masks
.source
|| masks
.sign
)
1474 if (flags
& TEX_DERIV
)
1476 if(flags
& TEX_PROJ
) FIXME("Projected texture sampling with custom derivatives\n");
1477 if(flags
& TEX_BIAS
) FIXME("Biased texture sampling with custom derivatives\n");
1478 shader_addline(buffer
, "TXD%s %s, %s, %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
,
1479 dsx
, dsy
, sampler_idx
, tex_type
);
1481 else if(flags
& TEX_LOD
)
1483 if(flags
& TEX_PROJ
) FIXME("Projected texture sampling with explicit lod\n");
1484 if(flags
& TEX_BIAS
) FIXME("Biased texture sampling with explicit lod\n");
1485 shader_addline(buffer
, "TXL%s %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
,
1486 sampler_idx
, tex_type
);
1488 else if (flags
& TEX_BIAS
)
1490 /* Shouldn't be possible, but let's check for it */
1491 if(flags
& TEX_PROJ
) FIXME("Biased and Projected texture sampling\n");
1492 /* TXB takes the 4th component of the source vector automatically, as d3d. Nothing more to do */
1493 shader_addline(buffer
, "TXB%s %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
, sampler_idx
, tex_type
);
1495 else if (flags
& TEX_PROJ
)
1497 shader_addline(buffer
, "TXP%s %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
, sampler_idx
, tex_type
);
1503 const unsigned char idx
= priv
->cur_np2fixup_info
->super
.idx
[sampler_idx
];
1504 shader_addline(buffer
, "MUL TA, np2fixup[%u].%s, %s;\n", idx
>> 1,
1505 (idx
% 2) ? "zwxy" : "xyzw", coord_reg
);
1507 shader_addline(buffer
, "TEX%s %s, TA, texture[%u], %s;\n", mod
, tex_dst
, sampler_idx
, tex_type
);
1510 shader_addline(buffer
, "TEX%s %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
, sampler_idx
, tex_type
);
1515 gen_color_correction(buffer
, dst_str
, tex_dst
,
1516 arb_get_helper_value(WINED3D_SHADER_TYPE_PIXEL
, ARB_ONE
),
1517 arb_get_helper_value(WINED3D_SHADER_TYPE_PIXEL
, ARB_TWO
),
1518 priv
->cur_ps_args
->super
.color_fixup
[sampler_idx
], masks
);
1522 static void shader_arb_get_src_param(const struct wined3d_shader_instruction
*ins
,
1523 const struct wined3d_shader_src_param
*src
, unsigned int tmpreg
, char *outregstr
)
1525 /* Generate a line that does the input modifier computation and return the input register to use */
1526 BOOL is_color
= FALSE
, insert_line
;
1529 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1530 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1531 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
1532 const char *two
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_TWO
);
1534 /* Assume a new line will be added */
1537 /* Get register name */
1538 shader_arb_get_register_name(ins
, &src
->reg
, regstr
, &is_color
);
1539 shader_arb_get_swizzle(src
, is_color
, swzstr
);
1541 switch (src
->modifiers
)
1543 case WINED3DSPSM_NONE
:
1544 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
1545 insert_line
= FALSE
;
1547 case WINED3DSPSM_NEG
:
1548 sprintf(outregstr
, "-%s%s", regstr
, swzstr
);
1549 insert_line
= FALSE
;
1551 case WINED3DSPSM_BIAS
:
1552 shader_addline(buffer
, "ADD T%c, %s, -coefdiv.x;\n", 'A' + tmpreg
, regstr
);
1554 case WINED3DSPSM_BIASNEG
:
1555 shader_addline(buffer
, "ADD T%c, -%s, coefdiv.x;\n", 'A' + tmpreg
, regstr
);
1557 case WINED3DSPSM_SIGN
:
1558 shader_addline(buffer
, "MAD T%c, %s, %s, -%s;\n", 'A' + tmpreg
, regstr
, two
, one
);
1560 case WINED3DSPSM_SIGNNEG
:
1561 shader_addline(buffer
, "MAD T%c, %s, -%s, %s;\n", 'A' + tmpreg
, regstr
, two
, one
);
1563 case WINED3DSPSM_COMP
:
1564 shader_addline(buffer
, "SUB T%c, %s, %s;\n", 'A' + tmpreg
, one
, regstr
);
1566 case WINED3DSPSM_X2
:
1567 shader_addline(buffer
, "ADD T%c, %s, %s;\n", 'A' + tmpreg
, regstr
, regstr
);
1569 case WINED3DSPSM_X2NEG
:
1570 shader_addline(buffer
, "ADD T%c, -%s, -%s;\n", 'A' + tmpreg
, regstr
, regstr
);
1572 case WINED3DSPSM_DZ
:
1573 shader_addline(buffer
, "RCP T%c, %s.z;\n", 'A' + tmpreg
, regstr
);
1574 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
1576 case WINED3DSPSM_DW
:
1577 shader_addline(buffer
, "RCP T%c, %s.w;\n", 'A' + tmpreg
, regstr
);
1578 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
1580 case WINED3DSPSM_ABS
:
1581 if(ctx
->target_version
>= NV2
) {
1582 sprintf(outregstr
, "|%s%s|", regstr
, swzstr
);
1583 insert_line
= FALSE
;
1585 shader_addline(buffer
, "ABS T%c, %s;\n", 'A' + tmpreg
, regstr
);
1588 case WINED3DSPSM_ABSNEG
:
1589 if(ctx
->target_version
>= NV2
) {
1590 sprintf(outregstr
, "-|%s%s|", regstr
, swzstr
);
1592 shader_addline(buffer
, "ABS T%c, %s;\n", 'A' + tmpreg
, regstr
);
1593 sprintf(outregstr
, "-T%c%s", 'A' + tmpreg
, swzstr
);
1595 insert_line
= FALSE
;
1598 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
1599 insert_line
= FALSE
;
1602 /* Return modified or original register, with swizzle */
1604 sprintf(outregstr
, "T%c%s", 'A' + tmpreg
, swzstr
);
1607 static void pshader_hw_bem(const struct wined3d_shader_instruction
*ins
)
1609 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1610 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1611 unsigned int sampler_code
= dst
->reg
.idx
[0].offset
;
1613 char src_name
[2][50];
1615 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1617 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed
1619 * Keep in mind that src_name[1] can be "TB" and src_name[0] can be "TA" because modifiers like _x2 are valid
1620 * with bem. So delay loading the first parameter until after the perturbation calculation which needs two
1623 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1624 shader_addline(buffer
, "SWZ TA, bumpenvmat%d, x, z, 0, 0;\n", sampler_code
);
1625 shader_addline(buffer
, "DP3 TC.r, TA, %s;\n", src_name
[1]);
1626 shader_addline(buffer
, "SWZ TA, bumpenvmat%d, y, w, 0, 0;\n", sampler_code
);
1627 shader_addline(buffer
, "DP3 TC.g, TA, %s;\n", src_name
[1]);
1629 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1630 shader_addline(buffer
, "ADD %s, %s, TC;\n", dst_name
, src_name
[0]);
1633 static enum wined3d_shader_src_modifier
negate_modifiers(enum wined3d_shader_src_modifier mod
, char *extra_char
)
1638 case WINED3DSPSM_NONE
: return WINED3DSPSM_NEG
;
1639 case WINED3DSPSM_NEG
: return WINED3DSPSM_NONE
;
1640 case WINED3DSPSM_BIAS
: return WINED3DSPSM_BIASNEG
;
1641 case WINED3DSPSM_BIASNEG
: return WINED3DSPSM_BIAS
;
1642 case WINED3DSPSM_SIGN
: return WINED3DSPSM_SIGNNEG
;
1643 case WINED3DSPSM_SIGNNEG
: return WINED3DSPSM_SIGN
;
1644 case WINED3DSPSM_COMP
: *extra_char
= '-'; return WINED3DSPSM_COMP
;
1645 case WINED3DSPSM_X2
: return WINED3DSPSM_X2NEG
;
1646 case WINED3DSPSM_X2NEG
: return WINED3DSPSM_X2
;
1647 case WINED3DSPSM_DZ
: *extra_char
= '-'; return WINED3DSPSM_DZ
;
1648 case WINED3DSPSM_DW
: *extra_char
= '-'; return WINED3DSPSM_DW
;
1649 case WINED3DSPSM_ABS
: return WINED3DSPSM_ABSNEG
;
1650 case WINED3DSPSM_ABSNEG
: return WINED3DSPSM_ABS
;
1651 case WINED3DSPSM_NOT
: FIXME("Unknown modifier %u\n", mod
);
1656 static void pshader_hw_cnd(const struct wined3d_shader_instruction
*ins
)
1658 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1659 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1661 char src_name
[3][50];
1662 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1663 ins
->ctx
->reg_maps
->shader_version
.minor
);
1665 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1666 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1668 if (shader_version
<= WINED3D_SHADER_VERSION(1, 3) && ins
->coissue
1669 && ins
->dst
->write_mask
!= WINED3DSP_WRITEMASK_3
)
1671 shader_addline(buffer
, "MOV%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
[1]);
1675 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
1678 /* src0 may have a negate srcmod set, so we can't blindly add "-" to the name */
1679 src0_copy
.modifiers
= negate_modifiers(src0_copy
.modifiers
, &extra_neg
);
1681 shader_arb_get_src_param(ins
, &src0_copy
, 0, src_name
[0]);
1682 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1683 shader_addline(buffer
, "ADD TA, %c%s, coefdiv.x;\n", extra_neg
, src_name
[0]);
1684 shader_addline(buffer
, "CMP%s %s, TA, %s, %s;\n", shader_arb_get_modifier(ins
),
1685 dst_name
, src_name
[1], src_name
[2]);
1689 static void pshader_hw_cmp(const struct wined3d_shader_instruction
*ins
)
1691 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1692 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1694 char src_name
[3][50];
1696 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1698 /* Generate input register names (with modifiers) */
1699 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1700 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1701 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1703 shader_addline(buffer
, "CMP%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1704 dst_name
, src_name
[0], src_name
[2], src_name
[1]);
1707 /** Process the WINED3DSIO_DP2ADD instruction in ARB.
1708 * dst = dot2(src0, src1) + src2 */
1709 static void pshader_hw_dp2add(const struct wined3d_shader_instruction
*ins
)
1711 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1712 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1714 char src_name
[3][50];
1715 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1717 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1718 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1719 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1721 if(ctx
->target_version
>= NV3
)
1723 /* GL_NV_fragment_program2 has a 1:1 matching instruction */
1724 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1725 shader_addline(buffer
, "DP2A%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1726 dst_name
, src_name
[0], src_name
[1], src_name
[2]);
1728 else if(ctx
->target_version
>= NV2
)
1730 /* dst.x = src2.?, src0.x, src1.x + src0.y * src1.y
1731 * dst.y = src2.?, src0.x, src1.z + src0.y * src1.w
1732 * dst.z = src2.?, src0.x, src1.x + src0.y * src1.y
1733 * dst.z = src2.?, src0.x, src1.z + src0.y * src1.w
1735 * Make sure that src1.zw = src1.xy, then we get a classic dp2add
1737 * .xyxy and other swizzles that we could get with this are not valid in
1738 * plain ARBfp, but luckily the NV extension grammar lifts this limitation.
1740 struct wined3d_shader_src_param tmp_param
= ins
->src
[1];
1741 DWORD swizzle
= tmp_param
.swizzle
& 0xf; /* Selects .xy */
1742 tmp_param
.swizzle
= swizzle
| (swizzle
<< 4); /* Creates .xyxy */
1744 shader_arb_get_src_param(ins
, &tmp_param
, 1, src_name
[1]);
1746 shader_addline(buffer
, "X2D%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1747 dst_name
, src_name
[2], src_name
[0], src_name
[1]);
1751 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1752 /* 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]
1753 * src_name[0] can be TA, but TA is a private temp for modifiers, so it is save to overwrite
1755 shader_addline(buffer
, "MOV TA, %s;\n", src_name
[0]);
1756 shader_addline(buffer
, "MOV TA.z, 0.0;\n");
1757 shader_addline(buffer
, "DP3 TA, TA, %s;\n", src_name
[1]);
1758 shader_addline(buffer
, "ADD%s %s, TA, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
[2]);
1762 /* Map the opcode 1-to-1 to the GL code */
1763 static void shader_hw_map2gl(const struct wined3d_shader_instruction
*ins
)
1765 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1766 const char *instruction
;
1767 char arguments
[256], dst_str
[50];
1769 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1771 switch (ins
->handler_idx
)
1773 case WINED3DSIH_ABS
: instruction
= "ABS"; break;
1774 case WINED3DSIH_ADD
: instruction
= "ADD"; break;
1775 case WINED3DSIH_CRS
: instruction
= "XPD"; break;
1776 case WINED3DSIH_DP3
: instruction
= "DP3"; break;
1777 case WINED3DSIH_DP4
: instruction
= "DP4"; break;
1778 case WINED3DSIH_DST
: instruction
= "DST"; break;
1779 case WINED3DSIH_FRC
: instruction
= "FRC"; break;
1780 case WINED3DSIH_LIT
: instruction
= "LIT"; break;
1781 case WINED3DSIH_LRP
: instruction
= "LRP"; break;
1782 case WINED3DSIH_MAD
: instruction
= "MAD"; break;
1783 case WINED3DSIH_MAX
: instruction
= "MAX"; break;
1784 case WINED3DSIH_MIN
: instruction
= "MIN"; break;
1785 case WINED3DSIH_MOV
: instruction
= "MOV"; break;
1786 case WINED3DSIH_MUL
: instruction
= "MUL"; break;
1787 case WINED3DSIH_SGE
: instruction
= "SGE"; break;
1788 case WINED3DSIH_SLT
: instruction
= "SLT"; break;
1789 case WINED3DSIH_SUB
: instruction
= "SUB"; break;
1790 case WINED3DSIH_MOVA
:instruction
= "ARR"; break;
1791 case WINED3DSIH_DSX
: instruction
= "DDX"; break;
1792 default: instruction
= "";
1793 FIXME("Unhandled opcode %s.\n", debug_d3dshaderinstructionhandler(ins
->handler_idx
));
1797 /* Note that shader_arb_add_dst_param() adds spaces. */
1798 arguments
[0] = '\0';
1799 shader_arb_get_dst_param(ins
, dst
, dst_str
);
1800 for (i
= 0; i
< ins
->src_count
; ++i
)
1803 strcat(arguments
, ", ");
1804 shader_arb_get_src_param(ins
, &ins
->src
[i
], i
, operand
);
1805 strcat(arguments
, operand
);
1807 shader_addline(buffer
, "%s%s %s%s;\n", instruction
, shader_arb_get_modifier(ins
), dst_str
, arguments
);
1810 static void shader_hw_nop(const struct wined3d_shader_instruction
*ins
) {}
1812 static DWORD
shader_arb_select_component(DWORD swizzle
, DWORD component
)
1814 return ((swizzle
>> 2 * component
) & 0x3) * 0x55;
1817 static void shader_hw_mov(const struct wined3d_shader_instruction
*ins
)
1819 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
1820 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
1821 BOOL pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
1822 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1823 const char *zero
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ZERO
);
1824 const char *one
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ONE
);
1825 const char *two
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_TWO
);
1827 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1828 char src0_param
[256];
1830 if (ins
->handler_idx
== WINED3DSIH_MOVA
)
1832 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1834 const char *offset
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_VS_REL_OFFSET
);
1836 if(ctx
->target_version
>= NV2
) {
1837 shader_hw_map2gl(ins
);
1840 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_param
);
1841 shader_arb_get_write_mask(ins
, &ins
->dst
[0], write_mask
);
1843 /* This implements the mova formula used in GLSL. The first two instructions
1844 * prepare the sign() part. Note that it is fine to have my_sign(0.0) = 1.0
1848 * A0.x = arl(floor(abs(0.0) + 0.5) * 1.0) = floor(0.5) = 0.0 since arl does a floor
1850 * The ARL is performed when A0 is used - the requested component is read from A0_SHADOW into
1851 * A0.x. We can use the overwritten component of A0_shadow as temporary storage for the sign.
1853 shader_addline(buffer
, "SGE A0_SHADOW%s, %s, %s;\n", write_mask
, src0_param
, zero
);
1854 shader_addline(buffer
, "MAD A0_SHADOW%s, A0_SHADOW, %s, -%s;\n", write_mask
, two
, one
);
1856 shader_addline(buffer
, "ABS TA%s, %s;\n", write_mask
, src0_param
);
1857 shader_addline(buffer
, "ADD TA%s, TA, rel_addr_const.x;\n", write_mask
);
1858 shader_addline(buffer
, "FLR TA%s, TA;\n", write_mask
);
1859 if (shader_data
->rel_offset
)
1861 shader_addline(buffer
, "ADD TA%s, TA, %s;\n", write_mask
, offset
);
1863 shader_addline(buffer
, "MUL A0_SHADOW%s, TA, A0_SHADOW;\n", write_mask
);
1865 ((struct shader_arb_ctx_priv
*)ins
->ctx
->backend_data
)->addr_reg
[0] = '\0';
1867 else if (reg_maps
->shader_version
.major
== 1
1868 && !shader_is_pshader_version(reg_maps
->shader_version
.type
)
1869 && ins
->dst
[0].reg
.type
== WINED3DSPR_ADDR
)
1871 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1872 src0_param
[0] = '\0';
1874 if (shader_data
->rel_offset
&& ctx
->target_version
== ARB
)
1876 const char *offset
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_VS_REL_OFFSET
);
1877 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_param
);
1878 shader_addline(buffer
, "ADD TA.x, %s, %s;\n", src0_param
, offset
);
1879 shader_addline(buffer
, "ARL A0.x, TA.x;\n");
1883 /* Apple's ARB_vertex_program implementation does not accept an ARL source argument
1884 * with more than one component. Thus replicate the first source argument over all
1885 * 4 components. For example, .xyzw -> .x (or better: .xxxx), .zwxy -> .z, etc) */
1886 struct wined3d_shader_src_param tmp_src
= ins
->src
[0];
1887 tmp_src
.swizzle
= shader_arb_select_component(tmp_src
.swizzle
, 0);
1888 shader_arb_get_src_param(ins
, &tmp_src
, 0, src0_param
);
1889 shader_addline(buffer
, "ARL A0.x, %s;\n", src0_param
);
1892 else if (ins
->dst
[0].reg
.type
== WINED3DSPR_COLOROUT
&& !ins
->dst
[0].reg
.idx
[0].offset
&& pshader
)
1894 if (ctx
->ps_post_process
&& shader
->u
.ps
.color0_mov
)
1896 shader_addline(buffer
, "#mov handled in srgb write or fog code\n");
1899 shader_hw_map2gl(ins
);
1903 shader_hw_map2gl(ins
);
1907 static void pshader_hw_texkill(const struct wined3d_shader_instruction
*ins
)
1909 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1910 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1913 /* No swizzles are allowed in d3d's texkill. PS 1.x ignores the 4th component as documented,
1914 * but >= 2.0 honors it (undocumented, but tested by the d3d9 testsuite)
1916 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1918 if (ins
->ctx
->reg_maps
->shader_version
.major
>= 2)
1920 const char *kilsrc
= "TA";
1923 shader_arb_get_register_name(ins
, &dst
->reg
, reg_dest
, &is_color
);
1924 if(dst
->write_mask
== WINED3DSP_WRITEMASK_ALL
)
1930 /* Sigh. KIL doesn't support swizzles/writemasks. KIL passes a writemask, but ".xy" for example
1931 * is not valid as a swizzle in ARB (needs ".xyyy"). Use SWZ to load the register properly, and set
1932 * masked out components to 0(won't kill)
1934 char x
= '0', y
= '0', z
= '0', w
= '0';
1935 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
) x
= 'x';
1936 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
) y
= 'y';
1937 if(dst
->write_mask
& WINED3DSP_WRITEMASK_2
) z
= 'z';
1938 if(dst
->write_mask
& WINED3DSP_WRITEMASK_3
) w
= 'w';
1939 shader_addline(buffer
, "SWZ TA, %s, %c, %c, %c, %c;\n", reg_dest
, x
, y
, z
, w
);
1941 shader_addline(buffer
, "KIL %s;\n", kilsrc
);
1945 /* ARB fp doesn't like swizzles on the parameter of the KIL instruction. To mask the 4th component,
1946 * copy the register into our general purpose TMP variable, overwrite .w and pass TMP to KIL
1948 * ps_1_3 shaders use the texcoord incarnation of the Tx register. ps_1_4 shaders can use the same,
1949 * or pass in any temporary register(in shader phase 2)
1951 if (ins
->ctx
->reg_maps
->shader_version
.minor
<= 3)
1952 sprintf(reg_dest
, "fragment.texcoord[%u]", dst
->reg
.idx
[0].offset
);
1954 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1955 shader_addline(buffer
, "SWZ TA, %s, x, y, z, 1;\n", reg_dest
);
1956 shader_addline(buffer
, "KIL TA;\n");
1960 static void pshader_hw_tex(const struct wined3d_shader_instruction
*ins
)
1962 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1963 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1964 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1965 ins
->ctx
->reg_maps
->shader_version
.minor
);
1966 struct wined3d_shader_src_param src
;
1970 unsigned int reg_sampler_code
;
1972 BOOL swizzle_coord
= FALSE
;
1974 /* All versions have a destination register */
1975 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1977 /* 1.0-1.4: Use destination register number as texture code.
1978 2.0+: Use provided sampler number as texture code. */
1979 if (shader_version
< WINED3D_SHADER_VERSION(2,0))
1980 reg_sampler_code
= dst
->reg
.idx
[0].offset
;
1982 reg_sampler_code
= ins
->src
[1].reg
.idx
[0].offset
;
1984 /* 1.0-1.3: Use the texcoord varying.
1985 1.4+: Use provided coordinate source register. */
1986 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
1987 sprintf(reg_coord
, "fragment.texcoord[%u]", reg_sampler_code
);
1989 /* TEX is the only instruction that can handle DW and DZ natively */
1991 if(src
.modifiers
== WINED3DSPSM_DW
) src
.modifiers
= WINED3DSPSM_NONE
;
1992 if(src
.modifiers
== WINED3DSPSM_DZ
) src
.modifiers
= WINED3DSPSM_NONE
;
1993 shader_arb_get_src_param(ins
, &src
, 0, reg_coord
);
1997 * 1.1, 1.2, 1.3: Use WINED3D_TSS_TEXTURETRANSFORMFLAGS
1998 * 1.4: Use WINED3DSPSM_DZ or WINED3DSPSM_DW on src[0]
1999 * 2.0+: Use WINED3DSI_TEXLD_PROJECT on the opcode
2001 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
2004 if (reg_sampler_code
< WINED3D_MAX_TEXTURES
)
2005 flags
= priv
->cur_ps_args
->super
.tex_transform
>> reg_sampler_code
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
;
2006 if (flags
& WINED3D_PSARGS_PROJECTED
)
2008 myflags
|= TEX_PROJ
;
2009 if ((flags
& ~WINED3D_PSARGS_PROJECTED
) == WINED3D_TTFF_COUNT3
)
2010 swizzle_coord
= TRUE
;
2013 else if (shader_version
< WINED3D_SHADER_VERSION(2,0))
2015 enum wined3d_shader_src_modifier src_mod
= ins
->src
[0].modifiers
;
2016 if (src_mod
== WINED3DSPSM_DZ
)
2018 swizzle_coord
= TRUE
;
2019 myflags
|= TEX_PROJ
;
2020 } else if(src_mod
== WINED3DSPSM_DW
) {
2021 myflags
|= TEX_PROJ
;
2024 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) myflags
|= TEX_PROJ
;
2025 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) myflags
|= TEX_BIAS
;
2030 /* TXP cannot handle DZ natively, so move the z coordinate to .w.
2031 * reg_coord is a read-only varying register, so we need a temp reg */
2032 shader_addline(ins
->ctx
->buffer
, "SWZ TA, %s, x, y, z, z;\n", reg_coord
);
2033 strcpy(reg_coord
, "TA");
2036 shader_hw_sample(ins
, reg_sampler_code
, reg_dest
, reg_coord
, myflags
, NULL
, NULL
);
2039 static void pshader_hw_texcoord(const struct wined3d_shader_instruction
*ins
)
2041 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2042 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2043 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
2044 ins
->ctx
->reg_maps
->shader_version
.minor
);
2047 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
2049 unsigned int reg
= dst
->reg
.idx
[0].offset
;
2051 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2052 shader_addline(buffer
, "MOV_SAT %s, fragment.texcoord[%u];\n", dst_str
, reg
);
2056 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_src
);
2057 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2058 shader_addline(buffer
, "MOV %s, %s;\n", dst_str
, reg_src
);
2062 static void pshader_hw_texreg2ar(const struct wined3d_shader_instruction
*ins
)
2064 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2067 unsigned int reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2071 /* Note that texreg2ar treats Tx as a temporary register, not as a varying */
2072 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2073 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2074 /* Move .x first in case src_str is "TA" */
2075 shader_addline(buffer
, "MOV TA.y, %s.x;\n", src_str
);
2076 shader_addline(buffer
, "MOV TA.x, %s.w;\n", src_str
);
2077 if (reg1
< WINED3D_MAX_TEXTURES
)
2079 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2080 flags
= priv
->cur_ps_args
->super
.tex_transform
>> reg1
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
;
2082 shader_hw_sample(ins
, reg1
, dst_str
, "TA", flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2085 static void pshader_hw_texreg2gb(const struct wined3d_shader_instruction
*ins
)
2087 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2089 unsigned int reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2093 /* Note that texreg2gb treats Tx as a temporary register, not as a varying */
2094 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2095 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2096 shader_addline(buffer
, "MOV TA.x, %s.y;\n", src_str
);
2097 shader_addline(buffer
, "MOV TA.y, %s.z;\n", src_str
);
2098 shader_hw_sample(ins
, reg1
, dst_str
, "TA", 0, NULL
, NULL
);
2101 static void pshader_hw_texreg2rgb(const struct wined3d_shader_instruction
*ins
)
2103 unsigned int reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2107 /* Note that texreg2rg treats Tx as a temporary register, not as a varying */
2108 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2109 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2110 shader_hw_sample(ins
, reg1
, dst_str
, src_str
, 0, NULL
, NULL
);
2113 static void pshader_hw_texbem(const struct wined3d_shader_instruction
*ins
)
2115 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2116 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2117 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2118 char reg_coord
[40], dst_reg
[50], src_reg
[50];
2119 unsigned int reg_dest_code
;
2121 /* All versions have a destination register. The Tx where the texture coordinates come
2122 * from is the varying incarnation of the texture register
2124 reg_dest_code
= dst
->reg
.idx
[0].offset
;
2125 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_reg
);
2126 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_reg
);
2127 sprintf(reg_coord
, "fragment.texcoord[%u]", reg_dest_code
);
2129 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed
2130 * The Tx in which the perturbation map is stored is the tempreg incarnation of the texture register
2132 * GL_NV_fragment_program_option could handle this in one instruction via X2D:
2133 * X2D TA.xy, fragment.texcoord, T%u, bumpenvmat%u.xzyw
2135 * However, the NV extensions are never enabled for <= 2.0 shaders because of the performance penalty that
2136 * comes with it, and texbem is an 1.x only instruction. No 1.x instruction forces us to enable the NV
2139 shader_addline(buffer
, "SWZ TB, bumpenvmat%d, x, z, 0, 0;\n", reg_dest_code
);
2140 shader_addline(buffer
, "DP3 TA.x, TB, %s;\n", src_reg
);
2141 shader_addline(buffer
, "SWZ TB, bumpenvmat%d, y, w, 0, 0;\n", reg_dest_code
);
2142 shader_addline(buffer
, "DP3 TA.y, TB, %s;\n", src_reg
);
2144 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2145 * so we can't let the GL handle this.
2147 if ((priv
->cur_ps_args
->super
.tex_transform
>> reg_dest_code
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
)
2148 & WINED3D_PSARGS_PROJECTED
)
2150 shader_addline(buffer
, "RCP TB.w, %s.w;\n", reg_coord
);
2151 shader_addline(buffer
, "MUL TB.xy, %s, TB.w;\n", reg_coord
);
2152 shader_addline(buffer
, "ADD TA.xy, TA, TB;\n");
2154 shader_addline(buffer
, "ADD TA.xy, TA, %s;\n", reg_coord
);
2157 shader_hw_sample(ins
, reg_dest_code
, dst_reg
, "TA", 0, NULL
, NULL
);
2159 if (ins
->handler_idx
== WINED3DSIH_TEXBEML
)
2161 /* No src swizzles are allowed, so this is ok */
2162 shader_addline(buffer
, "MAD TA, %s.z, luminance%d.x, luminance%d.y;\n",
2163 src_reg
, reg_dest_code
, reg_dest_code
);
2164 shader_addline(buffer
, "MUL %s, %s, TA;\n", dst_reg
, dst_reg
);
2168 static void pshader_hw_texm3x2pad(const struct wined3d_shader_instruction
*ins
)
2170 unsigned int reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2171 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2172 char src0_name
[50], dst_name
[50];
2174 struct wined3d_shader_register tmp_reg
= ins
->dst
[0].reg
;
2176 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2177 /* The next instruction will be a texm3x2tex or texm3x2depth that writes to the uninitialized
2178 * T<reg+1> register. Use this register to store the calculated vector
2180 tmp_reg
.idx
[0].offset
= reg
+ 1;
2181 shader_arb_get_register_name(ins
, &tmp_reg
, dst_name
, &is_color
);
2182 shader_addline(buffer
, "DP3 %s.x, fragment.texcoord[%u], %s;\n", dst_name
, reg
, src0_name
);
2185 static void pshader_hw_texm3x2tex(const struct wined3d_shader_instruction
*ins
)
2187 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2189 unsigned int reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2190 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2196 /* We know that we're writing to the uninitialized T<reg> register, so use it for temporary storage */
2197 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2199 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2200 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2201 shader_addline(buffer
, "DP3 %s.y, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2202 flags
= reg
< WINED3D_MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2203 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2206 static void pshader_hw_texm3x3pad(const struct wined3d_shader_instruction
*ins
)
2208 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2209 unsigned int reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2210 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2211 char src0_name
[50], dst_name
[50];
2212 struct wined3d_shader_register tmp_reg
= ins
->dst
[0].reg
;
2215 /* There are always 2 texm3x3pad instructions followed by one texm3x3[tex,vspec, ...] instruction, with
2216 * incrementing ins->dst[0].register_idx numbers. So the pad instruction already knows the final destination
2217 * register, and this register is uninitialized(otherwise the assembler complains that it is 'redeclared')
2219 tmp_reg
.idx
[0].offset
= reg
+ 2 - tex_mx
->current_row
;
2220 shader_arb_get_register_name(ins
, &tmp_reg
, dst_name
, &is_color
);
2222 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2223 shader_addline(buffer
, "DP3 %s.%c, fragment.texcoord[%u], %s;\n",
2224 dst_name
, 'x' + tex_mx
->current_row
, reg
, src0_name
);
2225 tex_mx
->texcoord_w
[tex_mx
->current_row
++] = reg
;
2228 static void pshader_hw_texm3x3tex(const struct wined3d_shader_instruction
*ins
)
2230 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2231 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2233 unsigned int reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2234 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2236 char src0_name
[50], dst_name
[50];
2239 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2240 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2241 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_name
, reg
, src0_name
);
2243 /* Sample the texture using the calculated coordinates */
2244 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2245 flags
= reg
< WINED3D_MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2246 shader_hw_sample(ins
, reg
, dst_str
, dst_name
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2247 tex_mx
->current_row
= 0;
2250 static void pshader_hw_texm3x3vspec(const struct wined3d_shader_instruction
*ins
)
2252 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2253 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2255 unsigned int reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2256 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2262 /* Get the dst reg without writemask strings. We know this register is uninitialized, so we can use all
2263 * components for temporary data storage
2265 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2266 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2267 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2269 /* Construct the eye-ray vector from w coordinates */
2270 shader_addline(buffer
, "MOV TB.x, fragment.texcoord[%u].w;\n", tex_mx
->texcoord_w
[0]);
2271 shader_addline(buffer
, "MOV TB.y, fragment.texcoord[%u].w;\n", tex_mx
->texcoord_w
[1]);
2272 shader_addline(buffer
, "MOV TB.z, fragment.texcoord[%u].w;\n", reg
);
2274 /* Calculate reflection vector
2276 shader_addline(buffer
, "DP3 %s.w, %s, TB;\n", dst_reg
, dst_reg
);
2277 /* The .w is ignored when sampling, so I can use TB.w to calculate dot(N, N) */
2278 shader_addline(buffer
, "DP3 TB.w, %s, %s;\n", dst_reg
, dst_reg
);
2279 shader_addline(buffer
, "RCP TB.w, TB.w;\n");
2280 shader_addline(buffer
, "MUL %s.w, %s.w, TB.w;\n", dst_reg
, dst_reg
);
2281 shader_addline(buffer
, "MUL %s, %s.w, %s;\n", dst_reg
, dst_reg
, dst_reg
);
2282 shader_addline(buffer
, "MAD %s, coefmul.x, %s, -TB;\n", dst_reg
, dst_reg
);
2284 /* Sample the texture using the calculated coordinates */
2285 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2286 flags
= reg
< WINED3D_MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2287 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2288 tex_mx
->current_row
= 0;
2291 static void pshader_hw_texm3x3spec(const struct wined3d_shader_instruction
*ins
)
2293 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2294 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2296 unsigned int reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2297 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2304 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2305 shader_arb_get_src_param(ins
, &ins
->src
[0], 1, src1_name
);
2306 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2307 /* Note: dst_reg.xy is input here, generated by two texm3x3pad instructions */
2308 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2310 /* Calculate reflection vector.
2313 * dst_reg.xyz = 2 * --------- * N - E
2316 * Which normalizes the normal vector
2318 shader_addline(buffer
, "DP3 %s.w, %s, %s;\n", dst_reg
, dst_reg
, src1_name
);
2319 shader_addline(buffer
, "DP3 TC.w, %s, %s;\n", dst_reg
, dst_reg
);
2320 shader_addline(buffer
, "RCP TC.w, TC.w;\n");
2321 shader_addline(buffer
, "MUL %s.w, %s.w, TC.w;\n", dst_reg
, dst_reg
);
2322 shader_addline(buffer
, "MUL %s, %s.w, %s;\n", dst_reg
, dst_reg
, dst_reg
);
2323 shader_addline(buffer
, "MAD %s, coefmul.x, %s, -%s;\n", dst_reg
, dst_reg
, src1_name
);
2325 /* Sample the texture using the calculated coordinates */
2326 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2327 flags
= reg
< WINED3D_MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2328 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2329 tex_mx
->current_row
= 0;
2332 static void pshader_hw_texdepth(const struct wined3d_shader_instruction
*ins
)
2334 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2335 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2337 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2338 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2340 /* texdepth has an implicit destination, the fragment depth value. It's only parameter,
2341 * which is essentially an input, is the destination register because it is the first
2342 * parameter. According to the msdn, this must be register r5, but let's keep it more flexible
2343 * here(writemasks/swizzles are not valid on texdepth)
2345 shader_arb_get_dst_param(ins
, dst
, dst_name
);
2347 /* According to the msdn, the source register(must be r5) is unusable after
2348 * the texdepth instruction, so we're free to modify it
2350 shader_addline(buffer
, "MIN %s.y, %s.y, %s;\n", dst_name
, dst_name
, one
);
2352 /* How to deal with the special case dst_name.g == 0? if r != 0, then
2353 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
2354 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
2356 shader_addline(buffer
, "RCP %s.y, %s.y;\n", dst_name
, dst_name
);
2357 shader_addline(buffer
, "MUL TA.x, %s.x, %s.y;\n", dst_name
, dst_name
);
2358 shader_addline(buffer
, "MIN TA.x, TA.x, %s;\n", one
);
2359 shader_addline(buffer
, "MAX result.depth, TA.x, %s;\n", zero
);
2362 /** Process the WINED3DSIO_TEXDP3TEX instruction in ARB:
2363 * Take a 3-component dot product of the TexCoord[dstreg] and src,
2364 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
2365 static void pshader_hw_texdp3tex(const struct wined3d_shader_instruction
*ins
)
2367 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2368 unsigned int sampler_idx
= ins
->dst
[0].reg
.idx
[0].offset
;
2372 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2373 shader_addline(buffer
, "MOV TB, 0.0;\n");
2374 shader_addline(buffer
, "DP3 TB.x, fragment.texcoord[%u], %s;\n", sampler_idx
, src0
);
2376 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2377 shader_hw_sample(ins
, sampler_idx
, dst_str
, "TB", 0 /* Only one coord, can't be projected */, NULL
, NULL
);
2380 /** Process the WINED3DSIO_TEXDP3 instruction in ARB:
2381 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
2382 static void pshader_hw_texdp3(const struct wined3d_shader_instruction
*ins
)
2384 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2387 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2389 /* Handle output register */
2390 shader_arb_get_dst_param(ins
, dst
, dst_str
);
2391 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2392 shader_addline(buffer
, "DP3 %s, fragment.texcoord[%u], %s;\n", dst_str
, dst
->reg
.idx
[0].offset
, src0
);
2395 /** Process the WINED3DSIO_TEXM3X3 instruction in ARB
2396 * Perform the 3rd row of a 3x3 matrix multiply */
2397 static void pshader_hw_texm3x3(const struct wined3d_shader_instruction
*ins
)
2399 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2400 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2401 char dst_str
[50], dst_name
[50];
2405 shader_arb_get_dst_param(ins
, dst
, dst_str
);
2406 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2407 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2408 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_name
, dst
->reg
.idx
[0].offset
, src0
);
2409 shader_addline(buffer
, "MOV %s, %s;\n", dst_str
, dst_name
);
2412 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in ARB:
2413 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
2414 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
2415 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
2417 static void pshader_hw_texm3x2depth(const struct wined3d_shader_instruction
*ins
)
2419 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2420 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2421 char src0
[50], dst_name
[50];
2423 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2424 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2426 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2427 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2428 shader_addline(buffer
, "DP3 %s.y, fragment.texcoord[%u], %s;\n", dst_name
, dst
->reg
.idx
[0].offset
, src0
);
2430 /* How to deal with the special case dst_name.g == 0? if r != 0, then
2431 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
2432 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
2434 shader_addline(buffer
, "RCP %s.y, %s.y;\n", dst_name
, dst_name
);
2435 shader_addline(buffer
, "MUL %s.x, %s.x, %s.y;\n", dst_name
, dst_name
, dst_name
);
2436 shader_addline(buffer
, "MIN %s.x, %s.x, %s;\n", dst_name
, dst_name
, one
);
2437 shader_addline(buffer
, "MAX result.depth, %s.x, %s;\n", dst_name
, zero
);
2440 /** Handles transforming all WINED3DSIO_M?x? opcodes for
2441 Vertex/Pixel shaders to ARB_vertex_program codes */
2442 static void shader_hw_mnxn(const struct wined3d_shader_instruction
*ins
)
2445 int nComponents
= 0;
2446 struct wined3d_shader_dst_param tmp_dst
= {{0}};
2447 struct wined3d_shader_src_param tmp_src
[2] = {{{0}}};
2448 struct wined3d_shader_instruction tmp_ins
;
2450 memset(&tmp_ins
, 0, sizeof(tmp_ins
));
2452 /* Set constants for the temporary argument */
2453 tmp_ins
.ctx
= ins
->ctx
;
2454 tmp_ins
.dst_count
= 1;
2455 tmp_ins
.dst
= &tmp_dst
;
2456 tmp_ins
.src_count
= 2;
2457 tmp_ins
.src
= tmp_src
;
2459 switch(ins
->handler_idx
)
2461 case WINED3DSIH_M4x4
:
2463 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2465 case WINED3DSIH_M4x3
:
2467 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2469 case WINED3DSIH_M3x4
:
2471 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2473 case WINED3DSIH_M3x3
:
2475 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2477 case WINED3DSIH_M3x2
:
2479 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2482 FIXME("Unhandled opcode %s.\n", debug_d3dshaderinstructionhandler(ins
->handler_idx
));
2486 tmp_dst
= ins
->dst
[0];
2487 tmp_src
[0] = ins
->src
[0];
2488 tmp_src
[1] = ins
->src
[1];
2489 for (i
= 0; i
< nComponents
; ++i
)
2491 tmp_dst
.write_mask
= WINED3DSP_WRITEMASK_0
<< i
;
2492 shader_hw_map2gl(&tmp_ins
);
2493 ++tmp_src
[1].reg
.idx
[0].offset
;
2497 static enum wined3d_shader_src_modifier
abs_modifier(enum wined3d_shader_src_modifier mod
, BOOL
*need_abs
)
2503 case WINED3DSPSM_NONE
: return WINED3DSPSM_ABS
;
2504 case WINED3DSPSM_NEG
: return WINED3DSPSM_ABS
;
2505 case WINED3DSPSM_BIAS
: *need_abs
= TRUE
; return WINED3DSPSM_BIAS
;
2506 case WINED3DSPSM_BIASNEG
: *need_abs
= TRUE
; return WINED3DSPSM_BIASNEG
;
2507 case WINED3DSPSM_SIGN
: *need_abs
= TRUE
; return WINED3DSPSM_SIGN
;
2508 case WINED3DSPSM_SIGNNEG
: *need_abs
= TRUE
; return WINED3DSPSM_SIGNNEG
;
2509 case WINED3DSPSM_COMP
: *need_abs
= TRUE
; return WINED3DSPSM_COMP
;
2510 case WINED3DSPSM_X2
: *need_abs
= TRUE
; return WINED3DSPSM_X2
;
2511 case WINED3DSPSM_X2NEG
: *need_abs
= TRUE
; return WINED3DSPSM_X2NEG
;
2512 case WINED3DSPSM_DZ
: *need_abs
= TRUE
; return WINED3DSPSM_DZ
;
2513 case WINED3DSPSM_DW
: *need_abs
= TRUE
; return WINED3DSPSM_DW
;
2514 case WINED3DSPSM_ABS
: return WINED3DSPSM_ABS
;
2515 case WINED3DSPSM_ABSNEG
: return WINED3DSPSM_ABS
;
2516 case WINED3DSPSM_NOT
: FIXME("Unknown modifier %u\n", mod
);
2521 static void shader_hw_scalar_op(const struct wined3d_shader_instruction
*ins
)
2523 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2524 const char *instruction
;
2525 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
2526 BOOL need_abs
= FALSE
;
2531 switch(ins
->handler_idx
)
2533 case WINED3DSIH_RSQ
: instruction
= "RSQ"; break;
2534 case WINED3DSIH_RCP
: instruction
= "RCP"; break;
2535 case WINED3DSIH_EXPP
:
2536 if (ins
->ctx
->reg_maps
->shader_version
.major
< 2)
2538 instruction
= "EXP";
2542 case WINED3DSIH_EXP
:
2543 instruction
= "EX2";
2545 case WINED3DSIH_LOG
:
2546 case WINED3DSIH_LOGP
:
2547 /* The precision requirements suggest that LOGP matches ARBvp's LOG
2548 * instruction, but notice that the output of those instructions is
2550 src0_copy
.modifiers
= abs_modifier(src0_copy
.modifiers
, &need_abs
);
2551 instruction
= "LG2";
2553 default: instruction
= "";
2554 FIXME("Unhandled opcode %s.\n", debug_d3dshaderinstructionhandler(ins
->handler_idx
));
2558 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
2560 src0_copy
.swizzle
= shader_arb_select_component(src0_copy
.swizzle
, 3);
2562 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
); /* Destination */
2563 shader_arb_get_src_param(ins
, &src0_copy
, 0, src
);
2567 shader_addline(buffer
, "ABS TA.w, %s;\n", src
);
2568 shader_addline(buffer
, "%s%s %s, TA.w;\n", instruction
, shader_arb_get_modifier(ins
), dst
);
2572 shader_addline(buffer
, "%s%s %s, %s;\n", instruction
, shader_arb_get_modifier(ins
), dst
, src
);
2577 static void shader_hw_nrm(const struct wined3d_shader_instruction
*ins
)
2579 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2582 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2583 BOOL pshader
= shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2584 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2586 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2587 shader_arb_get_src_param(ins
, &ins
->src
[0], 1 /* Use TB */, src_name
);
2589 /* In D3D, NRM of a vector with length zero returns zero. Catch this situation, as
2590 * otherwise NRM or RSQ would return NaN */
2591 if(pshader
&& priv
->target_version
>= NV3
)
2593 /* GL_NV_fragment_program2's NRM needs protection against length zero vectors too
2595 * TODO: Find out if DP3+NRM+MOV is really faster than DP3+RSQ+MUL
2597 shader_addline(buffer
, "DP3C TA, %s, %s;\n", src_name
, src_name
);
2598 shader_addline(buffer
, "NRM%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
);
2599 shader_addline(buffer
, "MOV %s (EQ), %s;\n", dst_name
, zero
);
2601 else if(priv
->target_version
>= NV2
)
2603 shader_addline(buffer
, "DP3C TA.x, %s, %s;\n", src_name
, src_name
);
2604 shader_addline(buffer
, "RSQ TA.x (NE), TA.x;\n");
2605 shader_addline(buffer
, "MUL%s %s, %s, TA.x;\n", shader_arb_get_modifier(ins
), dst_name
,
2610 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2612 shader_addline(buffer
, "DP3 TA.x, %s, %s;\n", src_name
, src_name
);
2613 /* Pass any non-zero value to RSQ if the input vector has a length of zero. The
2614 * RSQ result doesn't matter, as long as multiplying it by 0 returns 0.
2616 shader_addline(buffer
, "SGE TA.y, -TA.x, %s;\n", zero
);
2617 shader_addline(buffer
, "MAD TA.x, %s, TA.y, TA.x;\n", one
);
2619 shader_addline(buffer
, "RSQ TA.x, TA.x;\n");
2620 /* dst.w = src[0].w * 1 / (src.x^2 + src.y^2 + src.z^2)^(1/2) according to msdn*/
2621 shader_addline(buffer
, "MUL%s %s, %s, TA.x;\n", shader_arb_get_modifier(ins
), dst_name
,
2626 static void shader_hw_lrp(const struct wined3d_shader_instruction
*ins
)
2628 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2630 char src_name
[3][50];
2632 /* ARB_fragment_program has a convenient LRP instruction */
2633 if(shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)) {
2634 shader_hw_map2gl(ins
);
2638 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2639 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
2640 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
2641 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
2643 shader_addline(buffer
, "SUB TA, %s, %s;\n", src_name
[1], src_name
[2]);
2644 shader_addline(buffer
, "MAD%s %s, %s, TA, %s;\n", shader_arb_get_modifier(ins
),
2645 dst_name
, src_name
[0], src_name
[2]);
2648 static void shader_hw_sincos(const struct wined3d_shader_instruction
*ins
)
2650 /* This instruction exists in ARB, but the d3d instruction takes two extra parameters which
2651 * must contain fixed constants. So we need a separate function to filter those constants and
2654 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2655 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2656 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2658 char src_name0
[50], src_name1
[50], src_name2
[50];
2661 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
2662 if(shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)) {
2663 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2664 /* No modifiers are supported on SCS */
2665 shader_addline(buffer
, "SCS %s, %s;\n", dst_name
, src_name0
);
2667 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
)
2669 shader_arb_get_register_name(ins
, &dst
->reg
, src_name0
, &is_color
);
2670 shader_addline(buffer
, "MOV_SAT %s, %s;\n", dst_name
, src_name0
);
2672 } else if(priv
->target_version
>= NV2
) {
2673 shader_arb_get_register_name(ins
, &dst
->reg
, dst_name
, &is_color
);
2675 /* Sincos writemask must be .x, .y or .xy */
2676 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
)
2677 shader_addline(buffer
, "COS%s %s.x, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name0
);
2678 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
)
2679 shader_addline(buffer
, "SIN%s %s.y, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name0
);
2681 /* Approximate sine and cosine with a taylor series, as per math textbook. The application passes 8
2682 * helper constants(D3DSINCOSCONST1 and D3DSINCOSCONST2) in src1 and src2.
2684 * sin(x) = x - x^3/3! + x^5/5! - x^7/7! + ...
2685 * cos(x) = 1 - x^2/2! + x^4/4! - x^6/6! + ...
2687 * The constants we get are:
2689 * +1 +1, -1 -1 +1 +1 -1 -1
2690 * ---- , ---- , ---- , ----- , ----- , ----- , ------
2691 * 1!*2 2!*4 3!*8 4!*16 5!*32 6!*64 7!*128
2693 * If used with x^2, x^3, x^4 etc they calculate sin(x/2) and cos(x/2):
2697 * (x/2)^4 = x^4 / 16
2698 * (x/2)^5 = x^5 / 32
2701 * To get the final result:
2702 * sin(x) = 2 * sin(x/2) * cos(x/2)
2703 * cos(x) = cos(x/2)^2 - sin(x/2)^2
2704 * (from sin(x+y) and cos(x+y) rules)
2706 * As per MSDN, dst.z is undefined after the operation, and so is
2707 * dst.x and dst.y if they're masked out by the writemask. Ie
2708 * sincos dst.y, src1, c0, c1
2709 * returns the sine in dst.y. dst.x and dst.z are undefined, dst.w is not touched. The assembler
2710 * vsa.exe also stops with an error if the dest register is the same register as the source
2711 * register. This means we can use dest.xyz as temporary storage. The assembler vsa.exe output also
2712 * indicates that sincos consumes 8 instruction slots in vs_2_0(and, strangely, in vs_3_0).
2714 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
2715 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name2
);
2716 shader_arb_get_register_name(ins
, &dst
->reg
, dst_name
, &is_color
);
2718 shader_addline(buffer
, "MUL %s.x, %s, %s;\n", dst_name
, src_name0
, src_name0
); /* x ^ 2 */
2719 shader_addline(buffer
, "MUL TA.y, %s.x, %s;\n", dst_name
, src_name0
); /* x ^ 3 */
2720 shader_addline(buffer
, "MUL %s.y, TA.y, %s;\n", dst_name
, src_name0
); /* x ^ 4 */
2721 shader_addline(buffer
, "MUL TA.z, %s.y, %s;\n", dst_name
, src_name0
); /* x ^ 5 */
2722 shader_addline(buffer
, "MUL %s.z, TA.z, %s;\n", dst_name
, src_name0
); /* x ^ 6 */
2723 shader_addline(buffer
, "MUL TA.w, %s.z, %s;\n", dst_name
, src_name0
); /* x ^ 7 */
2727 * Unfortunately we don't get the constants in a DP4-capable form. Is there a way to
2728 * properly merge that with MULs in the code above?
2729 * The swizzles .yz and xw however fit into the .yzxw swizzle added to ps_2_0. Maybe
2730 * we can merge the sine and cosine MAD rows to calculate them together.
2732 shader_addline(buffer
, "MUL TA.x, %s, %s.w;\n", src_name0
, src_name2
); /* x^1, +1/(1!*2) */
2733 shader_addline(buffer
, "MAD TA.x, TA.y, %s.x, TA.x;\n", src_name2
); /* -1/(3!*8) */
2734 shader_addline(buffer
, "MAD TA.x, TA.z, %s.w, TA.x;\n", src_name1
); /* +1/(5!*32) */
2735 shader_addline(buffer
, "MAD TA.x, TA.w, %s.x, TA.x;\n", src_name1
); /* -1/(7!*128) */
2738 shader_addline(buffer
, "MAD TA.y, %s.x, %s.y, %s.z;\n", dst_name
, src_name2
, src_name2
); /* -1/(2!*4), +1.0 */
2739 shader_addline(buffer
, "MAD TA.y, %s.y, %s.z, TA.y;\n", dst_name
, src_name1
); /* +1/(4!*16) */
2740 shader_addline(buffer
, "MAD TA.y, %s.z, %s.y, TA.y;\n", dst_name
, src_name1
); /* -1/(6!*64) */
2742 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
) {
2744 shader_addline(buffer
, "MUL TA.z, TA.y, TA.y;\n");
2745 shader_addline(buffer
, "MAD %s.x, -TA.x, TA.x, TA.z;\n", dst_name
);
2747 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
) {
2749 shader_addline(buffer
, "MUL %s.y, TA.x, TA.y;\n", dst_name
);
2750 shader_addline(buffer
, "ADD %s.y, %s.y, %s.y;\n", dst_name
, dst_name
, dst_name
);
2755 static void shader_hw_sgn(const struct wined3d_shader_instruction
*ins
)
2757 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2760 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
2762 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2763 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
);
2765 /* SGN is only valid in vertex shaders */
2766 if(ctx
->target_version
>= NV2
) {
2767 shader_addline(buffer
, "SSG%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
);
2771 /* If SRC > 0.0, -SRC < SRC = TRUE, otherwise false.
2772 * if SRC < 0.0, SRC < -SRC = TRUE. If neither is true, src = 0.0
2774 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
) {
2775 shader_addline(buffer
, "SLT %s, -%s, %s;\n", dst_name
, src_name
, src_name
);
2777 /* src contains TA? Write to the dest first. This won't overwrite our destination.
2778 * Then use TA, and calculate the final result
2780 * Not reading from TA? Store the first result in TA to avoid overwriting the
2781 * destination if src reg = dst reg
2783 if(strstr(src_name
, "TA"))
2785 shader_addline(buffer
, "SLT %s, %s, -%s;\n", dst_name
, src_name
, src_name
);
2786 shader_addline(buffer
, "SLT TA, -%s, %s;\n", src_name
, src_name
);
2787 shader_addline(buffer
, "ADD %s, %s, -TA;\n", dst_name
, dst_name
);
2791 shader_addline(buffer
, "SLT TA, -%s, %s;\n", src_name
, src_name
);
2792 shader_addline(buffer
, "SLT %s, %s, -%s;\n", dst_name
, src_name
, src_name
);
2793 shader_addline(buffer
, "ADD %s, TA, -%s;\n", dst_name
, dst_name
);
2798 static void shader_hw_dsy(const struct wined3d_shader_instruction
*ins
)
2800 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2806 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
);
2807 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src
);
2808 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2810 shader_addline(buffer
, "DDY %s, %s;\n", dst
, src
);
2811 shader_addline(buffer
, "MUL%s %s, %s, ycorrection.y;\n", shader_arb_get_modifier(ins
), dst
, dst_name
);
2814 static void shader_hw_pow(const struct wined3d_shader_instruction
*ins
)
2816 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2817 char src0
[50], src1
[50], dst
[50];
2818 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
2819 BOOL need_abs
= FALSE
;
2820 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2821 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2823 /* POW operates on the absolute value of the input */
2824 src0_copy
.modifiers
= abs_modifier(src0_copy
.modifiers
, &need_abs
);
2826 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
);
2827 shader_arb_get_src_param(ins
, &src0_copy
, 0, src0
);
2828 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src1
);
2831 shader_addline(buffer
, "ABS TA.x, %s;\n", src0
);
2833 shader_addline(buffer
, "MOV TA.x, %s;\n", src0
);
2835 if (priv
->target_version
>= NV2
)
2837 shader_addline(buffer
, "MOVC TA.y, %s;\n", src1
);
2838 shader_addline(buffer
, "POW%s %s, TA.x, TA.y;\n", shader_arb_get_modifier(ins
), dst
);
2839 shader_addline(buffer
, "MOV %s (EQ.y), %s;\n", dst
, one
);
2843 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2844 const char *flt_eps
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_EPS
);
2846 shader_addline(buffer
, "ABS TA.y, %s;\n", src1
);
2847 shader_addline(buffer
, "SGE TA.y, -TA.y, %s;\n", zero
);
2848 /* Possibly add flt_eps to avoid getting float special values */
2849 shader_addline(buffer
, "MAD TA.z, TA.y, %s, %s;\n", flt_eps
, src1
);
2850 shader_addline(buffer
, "POW%s TA.x, TA.x, TA.z;\n", shader_arb_get_modifier(ins
));
2851 shader_addline(buffer
, "MAD TA.x, -TA.x, TA.y, TA.x;\n");
2852 shader_addline(buffer
, "MAD %s, TA.y, %s, TA.x;\n", dst
, one
);
2856 static void shader_hw_loop(const struct wined3d_shader_instruction
*ins
)
2858 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2860 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2863 shader_arb_get_src_param(ins
, &ins
->src
[1], 0, src_name
);
2867 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2868 struct list
*e
= list_head(&priv
->control_frames
);
2869 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2871 if(priv
->loop_depth
> 1) shader_addline(buffer
, "PUSHA aL;\n");
2872 /* The constant loader makes sure to load -1 into iX.w */
2873 shader_addline(buffer
, "ARLC aL, %s.xywz;\n", src_name
);
2874 shader_addline(buffer
, "BRA loop_%u_end (LE.x);\n", control_frame
->no
.loop
);
2875 shader_addline(buffer
, "loop_%u_start:\n", control_frame
->no
.loop
);
2879 shader_addline(buffer
, "LOOP %s;\n", src_name
);
2883 static void shader_hw_rep(const struct wined3d_shader_instruction
*ins
)
2885 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2887 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2889 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
);
2891 /* The constant loader makes sure to load -1 into iX.w */
2894 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2895 struct list
*e
= list_head(&priv
->control_frames
);
2896 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2898 if(priv
->loop_depth
> 1) shader_addline(buffer
, "PUSHA aL;\n");
2900 shader_addline(buffer
, "ARLC aL, %s.xywz;\n", src_name
);
2901 shader_addline(buffer
, "BRA loop_%u_end (LE.x);\n", control_frame
->no
.loop
);
2902 shader_addline(buffer
, "loop_%u_start:\n", control_frame
->no
.loop
);
2906 shader_addline(buffer
, "REP %s;\n", src_name
);
2910 static void shader_hw_endloop(const struct wined3d_shader_instruction
*ins
)
2912 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2913 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2917 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2918 struct list
*e
= list_head(&priv
->control_frames
);
2919 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2921 shader_addline(buffer
, "ARAC aL.xy, aL;\n");
2922 shader_addline(buffer
, "BRA loop_%u_start (GT.x);\n", control_frame
->no
.loop
);
2923 shader_addline(buffer
, "loop_%u_end:\n", control_frame
->no
.loop
);
2925 if(priv
->loop_depth
> 1) shader_addline(buffer
, "POPA aL;\n");
2929 shader_addline(buffer
, "ENDLOOP;\n");
2933 static void shader_hw_endrep(const struct wined3d_shader_instruction
*ins
)
2935 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2936 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2940 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2941 struct list
*e
= list_head(&priv
->control_frames
);
2942 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2944 shader_addline(buffer
, "ARAC aL.xy, aL;\n");
2945 shader_addline(buffer
, "BRA loop_%u_start (GT.x);\n", control_frame
->no
.loop
);
2946 shader_addline(buffer
, "loop_%u_end:\n", control_frame
->no
.loop
);
2948 if(priv
->loop_depth
> 1) shader_addline(buffer
, "POPA aL;\n");
2952 shader_addline(buffer
, "ENDREP;\n");
2956 static const struct control_frame
*find_last_loop(const struct shader_arb_ctx_priv
*priv
)
2958 struct control_frame
*control_frame
;
2960 LIST_FOR_EACH_ENTRY(control_frame
, &priv
->control_frames
, struct control_frame
, entry
)
2962 if(control_frame
->type
== LOOP
|| control_frame
->type
== REP
) return control_frame
;
2964 ERR("Could not find loop for break\n");
2968 static void shader_hw_break(const struct wined3d_shader_instruction
*ins
)
2970 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2971 const struct control_frame
*control_frame
= find_last_loop(ins
->ctx
->backend_data
);
2972 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2976 shader_addline(buffer
, "BRA loop_%u_end;\n", control_frame
->no
.loop
);
2980 shader_addline(buffer
, "BRK;\n");
2984 static const char *get_compare(enum wined3d_shader_rel_op op
)
2988 case WINED3D_SHADER_REL_OP_GT
: return "GT";
2989 case WINED3D_SHADER_REL_OP_EQ
: return "EQ";
2990 case WINED3D_SHADER_REL_OP_GE
: return "GE";
2991 case WINED3D_SHADER_REL_OP_LT
: return "LT";
2992 case WINED3D_SHADER_REL_OP_NE
: return "NE";
2993 case WINED3D_SHADER_REL_OP_LE
: return "LE";
2995 FIXME("Unrecognized operator %#x.\n", op
);
3000 static enum wined3d_shader_rel_op
invert_compare(enum wined3d_shader_rel_op op
)
3004 case WINED3D_SHADER_REL_OP_GT
: return WINED3D_SHADER_REL_OP_LE
;
3005 case WINED3D_SHADER_REL_OP_EQ
: return WINED3D_SHADER_REL_OP_NE
;
3006 case WINED3D_SHADER_REL_OP_GE
: return WINED3D_SHADER_REL_OP_LT
;
3007 case WINED3D_SHADER_REL_OP_LT
: return WINED3D_SHADER_REL_OP_GE
;
3008 case WINED3D_SHADER_REL_OP_NE
: return WINED3D_SHADER_REL_OP_EQ
;
3009 case WINED3D_SHADER_REL_OP_LE
: return WINED3D_SHADER_REL_OP_GT
;
3011 FIXME("Unrecognized operator %#x.\n", op
);
3016 static void shader_hw_breakc(const struct wined3d_shader_instruction
*ins
)
3018 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3019 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3020 const struct control_frame
*control_frame
= find_last_loop(ins
->ctx
->backend_data
);
3023 const char *comp
= get_compare(ins
->flags
);
3025 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
3026 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
3030 /* SUBC CC, src0, src1" works only in pixel shaders, so use TA to throw
3031 * away the subtraction result
3033 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3034 shader_addline(buffer
, "BRA loop_%u_end (%s.x);\n", control_frame
->no
.loop
, comp
);
3038 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3039 shader_addline(buffer
, "BRK (%s.x);\n", comp
);
3043 static void shader_hw_ifc(const struct wined3d_shader_instruction
*ins
)
3045 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3046 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3047 struct list
*e
= list_head(&priv
->control_frames
);
3048 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3052 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3054 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
3055 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
3059 /* Invert the flag. We jump to the else label if the condition is NOT true */
3060 comp
= get_compare(invert_compare(ins
->flags
));
3061 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3062 shader_addline(buffer
, "BRA ifc_%u_else (%s.x);\n", control_frame
->no
.ifc
, comp
);
3066 comp
= get_compare(ins
->flags
);
3067 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3068 shader_addline(buffer
, "IF %s.x;\n", comp
);
3072 static void shader_hw_else(const struct wined3d_shader_instruction
*ins
)
3074 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3075 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3076 struct list
*e
= list_head(&priv
->control_frames
);
3077 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3078 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3082 shader_addline(buffer
, "BRA ifc_%u_endif;\n", control_frame
->no
.ifc
);
3083 shader_addline(buffer
, "ifc_%u_else:\n", control_frame
->no
.ifc
);
3084 control_frame
->had_else
= TRUE
;
3088 shader_addline(buffer
, "ELSE;\n");
3092 static void shader_hw_endif(const struct wined3d_shader_instruction
*ins
)
3094 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3095 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3096 struct list
*e
= list_head(&priv
->control_frames
);
3097 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3098 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3102 if(control_frame
->had_else
)
3104 shader_addline(buffer
, "ifc_%u_endif:\n", control_frame
->no
.ifc
);
3108 shader_addline(buffer
, "#No else branch. else is endif\n");
3109 shader_addline(buffer
, "ifc_%u_else:\n", control_frame
->no
.ifc
);
3114 shader_addline(buffer
, "ENDIF;\n");
3118 static void shader_hw_texldd(const struct wined3d_shader_instruction
*ins
)
3120 DWORD sampler_idx
= ins
->src
[1].reg
.idx
[0].offset
;
3122 char reg_src
[3][40];
3123 WORD flags
= TEX_DERIV
;
3125 shader_arb_get_dst_param(ins
, &ins
->dst
[0], reg_dest
);
3126 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_src
[0]);
3127 shader_arb_get_src_param(ins
, &ins
->src
[2], 1, reg_src
[1]);
3128 shader_arb_get_src_param(ins
, &ins
->src
[3], 2, reg_src
[2]);
3130 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) flags
|= TEX_PROJ
;
3131 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) flags
|= TEX_BIAS
;
3133 shader_hw_sample(ins
, sampler_idx
, reg_dest
, reg_src
[0], flags
, reg_src
[1], reg_src
[2]);
3136 static void shader_hw_texldl(const struct wined3d_shader_instruction
*ins
)
3138 DWORD sampler_idx
= ins
->src
[1].reg
.idx
[0].offset
;
3141 WORD flags
= TEX_LOD
;
3143 shader_arb_get_dst_param(ins
, &ins
->dst
[0], reg_dest
);
3144 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_coord
);
3146 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) flags
|= TEX_PROJ
;
3147 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) flags
|= TEX_BIAS
;
3149 shader_hw_sample(ins
, sampler_idx
, reg_dest
, reg_coord
, flags
, NULL
, NULL
);
3152 static void shader_hw_label(const struct wined3d_shader_instruction
*ins
)
3154 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3155 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3157 priv
->in_main_func
= FALSE
;
3158 /* Call instructions activate the NV extensions, not labels and rets. If there is an uncalled
3159 * subroutine, don't generate a label that will make GL complain
3161 if(priv
->target_version
== ARB
) return;
3163 shader_addline(buffer
, "l%u:\n", ins
->src
[0].reg
.idx
[0].offset
);
3166 static void vshader_add_footer(struct shader_arb_ctx_priv
*priv_ctx
,
3167 const struct arb_vshader_private
*shader_data
, const struct arb_vs_compile_args
*args
,
3168 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
,
3169 struct wined3d_string_buffer
*buffer
)
3173 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
3174 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
3175 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
3176 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
3178 if (args
->super
.fog_src
== VS_FOG_Z
)
3180 shader_addline(buffer
, "MOV result.fogcoord, TMP_OUT.z;\n");
3186 /* posFixup.x is always 1.0, so we can safely use it */
3187 shader_addline(buffer
, "ADD result.fogcoord, posFixup.x, -posFixup.x;\n");
3191 /* Clamp fogcoord */
3192 const char *zero
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ZERO
);
3193 const char *one
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ONE
);
3195 shader_addline(buffer
, "MIN TMP_FOGCOORD.x, TMP_FOGCOORD.x, %s;\n", one
);
3196 shader_addline(buffer
, "MAX result.fogcoord.x, TMP_FOGCOORD.x, %s;\n", zero
);
3200 /* Clipplanes are always stored without y inversion */
3201 if (use_nv_clip(gl_info
) && priv_ctx
->target_version
>= NV2
)
3203 if (args
->super
.clip_enabled
)
3205 for (i
= 0; i
< priv_ctx
->vs_clipplanes
; i
++)
3207 shader_addline(buffer
, "DP4 result.clip[%u].x, TMP_OUT, state.clip[%u].plane;\n", i
, i
);
3211 else if (args
->clip
.boolclip
.clip_texcoord
)
3213 static const char component
[4] = {'x', 'y', 'z', 'w'};
3214 unsigned int cur_clip
= 0;
3215 const char *zero
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_ZERO
);
3217 for (i
= 0; i
< gl_info
->limits
.user_clip_distances
; ++i
)
3219 if (args
->clip
.boolclip
.clipplane_mask
& (1u << i
))
3221 shader_addline(buffer
, "DP4 TA.%c, TMP_OUT, state.clip[%u].plane;\n",
3222 component
[cur_clip
++], i
);
3228 shader_addline(buffer
, "MOV TA, %s;\n", zero
);
3231 shader_addline(buffer
, "MOV TA.yzw, %s;\n", zero
);
3234 shader_addline(buffer
, "MOV TA.zw, %s;\n", zero
);
3237 shader_addline(buffer
, "MOV TA.w, %s;\n", zero
);
3240 shader_addline(buffer
, "MOV result.texcoord[%u], TA;\n",
3241 args
->clip
.boolclip
.clip_texcoord
- 1);
3244 /* Write the final position.
3246 * OpenGL coordinates specify the center of the pixel while d3d coords specify
3247 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
3248 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
3249 * contains 1.0 to allow a mad, but arb vs swizzles are too restricted for that.
3251 if (!gl_info
->supported
[ARB_CLIP_CONTROL
])
3253 shader_addline(buffer
, "MUL TA, posFixup, TMP_OUT.w;\n");
3254 shader_addline(buffer
, "ADD TMP_OUT.x, TMP_OUT.x, TA.z;\n");
3255 shader_addline(buffer
, "MAD TMP_OUT.y, TMP_OUT.y, posFixup.y, TA.w;\n");
3257 /* Z coord [0;1]->[-1;1] mapping, see comment in
3258 * get_projection_matrix() in utils.c. */
3259 if (need_helper_const(shader_data
, reg_maps
, gl_info
))
3261 const char *two
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_TWO
);
3262 shader_addline(buffer
, "MAD TMP_OUT.z, TMP_OUT.z, %s, -TMP_OUT.w;\n", two
);
3266 shader_addline(buffer
, "ADD TMP_OUT.z, TMP_OUT.z, TMP_OUT.z;\n");
3267 shader_addline(buffer
, "ADD TMP_OUT.z, TMP_OUT.z, -TMP_OUT.w;\n");
3271 shader_addline(buffer
, "MOV result.position, TMP_OUT;\n");
3273 priv_ctx
->footer_written
= TRUE
;
3276 static void shader_hw_ret(const struct wined3d_shader_instruction
*ins
)
3278 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3279 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3280 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
3281 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3283 if(priv
->target_version
== ARB
) return;
3287 if (priv
->in_main_func
) vshader_add_footer(priv
, shader
->backend_data
,
3288 priv
->cur_vs_args
, ins
->ctx
->reg_maps
, priv
->gl_info
, buffer
);
3291 shader_addline(buffer
, "RET;\n");
3294 static void shader_hw_call(const struct wined3d_shader_instruction
*ins
)
3296 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3297 shader_addline(buffer
, "CAL l%u;\n", ins
->src
[0].reg
.idx
[0].offset
);
3300 static BOOL
shader_arb_compile(const struct wined3d_gl_info
*gl_info
, GLenum target
, const char *src
)
3302 const char *ptr
, *end
, *line
;
3305 if (TRACE_ON(d3d_shader
))
3308 end
= ptr
+ strlen(ptr
);
3309 while ((line
= wined3d_get_line(&ptr
, end
)))
3311 TRACE_(d3d_shader
)(" %.*s", (int)(ptr
- line
), line
);
3315 GL_EXTCALL(glProgramStringARB(target
, GL_PROGRAM_FORMAT_ASCII_ARB
, strlen(src
), src
));
3316 checkGLcall("glProgramStringARB()");
3318 if (FIXME_ON(d3d_shader
))
3320 gl_info
->gl_ops
.gl
.p_glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
3323 FIXME_(d3d_shader
)("Program error at position %d: %s\n\n", pos
,
3324 debugstr_a((const char *)gl_info
->gl_ops
.gl
.p_glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
3326 end
= ptr
+ strlen(ptr
);
3327 while ((line
= wined3d_get_line(&ptr
, end
)))
3329 FIXME_(d3d_shader
)(" %.*s", (int)(ptr
- line
), line
);
3331 FIXME_(d3d_shader
)("\n");
3337 if (WARN_ON(d3d_perf
))
3339 GL_EXTCALL(glGetProgramivARB(target
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
3340 checkGLcall("glGetProgramivARB()");
3342 WARN_(d3d_perf
)("Program exceeds native resource limits.\n");
3348 static void arbfp_add_sRGB_correction(struct wined3d_string_buffer
*buffer
, const char *fragcolor
,
3349 const char *tmp1
, const char *tmp2
, const char *tmp3
, const char *tmp4
, BOOL condcode
)
3351 /* Perform sRGB write correction. See GLX_EXT_framebuffer_sRGB */
3355 /* Sigh. MOVC CC doesn't work, so use one of the temps as dummy dest */
3356 shader_addline(buffer
, "SUBC %s, %s.x, srgb_consts1.x;\n", tmp1
, fragcolor
);
3357 /* Calculate the > 0.0031308 case */
3358 shader_addline(buffer
, "POW %s.x (GE), %s.x, srgb_consts0.x;\n", fragcolor
, fragcolor
);
3359 shader_addline(buffer
, "POW %s.y (GE), %s.y, srgb_consts0.x;\n", fragcolor
, fragcolor
);
3360 shader_addline(buffer
, "POW %s.z (GE), %s.z, srgb_consts0.x;\n", fragcolor
, fragcolor
);
3361 shader_addline(buffer
, "MUL %s.xyz (GE), %s, srgb_consts0.y;\n", fragcolor
, fragcolor
);
3362 shader_addline(buffer
, "SUB %s.xyz (GE), %s, srgb_consts0.z;\n", fragcolor
, fragcolor
);
3363 /* Calculate the < case */
3364 shader_addline(buffer
, "MUL %s.xyz (LT), srgb_consts0.w, %s;\n", fragcolor
, fragcolor
);
3368 /* Calculate the > 0.0031308 case */
3369 shader_addline(buffer
, "POW %s.x, %s.x, srgb_consts0.x;\n", tmp1
, fragcolor
);
3370 shader_addline(buffer
, "POW %s.y, %s.y, srgb_consts0.x;\n", tmp1
, fragcolor
);
3371 shader_addline(buffer
, "POW %s.z, %s.z, srgb_consts0.x;\n", tmp1
, fragcolor
);
3372 shader_addline(buffer
, "MUL %s, %s, srgb_consts0.y;\n", tmp1
, tmp1
);
3373 shader_addline(buffer
, "SUB %s, %s, srgb_consts0.z;\n", tmp1
, tmp1
);
3374 /* Calculate the < case */
3375 shader_addline(buffer
, "MUL %s, srgb_consts0.w, %s;\n", tmp2
, fragcolor
);
3376 /* Get 1.0 / 0.0 masks for > 0.0031308 and < 0.0031308 */
3377 shader_addline(buffer
, "SLT %s, srgb_consts1.x, %s;\n", tmp3
, fragcolor
);
3378 shader_addline(buffer
, "SGE %s, srgb_consts1.x, %s;\n", tmp4
, fragcolor
);
3379 /* Store the components > 0.0031308 in the destination */
3380 shader_addline(buffer
, "MUL %s.xyz, %s, %s;\n", fragcolor
, tmp1
, tmp3
);
3381 /* Add the components that are < 0.0031308 */
3382 shader_addline(buffer
, "MAD %s.xyz, %s, %s, %s;\n", fragcolor
, tmp2
, tmp4
, fragcolor
);
3383 /* Move everything into result.color at once. Nvidia hardware cannot handle partial
3384 * result.color writes(.rgb first, then .a), or handle overwriting already written
3385 * components. The assembler uses a temporary register in this case, which is usually
3386 * not allocated from one of our registers that were used earlier.
3389 /* [0.0;1.0] clamping. Not needed, this is done implicitly */
3392 static const unsigned int *find_loop_control_values(const struct wined3d_shader
*shader
, DWORD idx
)
3394 const struct wined3d_shader_lconst
*constant
;
3396 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsI
, struct wined3d_shader_lconst
, entry
)
3398 if (constant
->idx
== idx
)
3400 return constant
->value
;
3406 static void init_ps_input(const struct wined3d_shader
*shader
,
3407 const struct arb_ps_compile_args
*args
, struct shader_arb_ctx_priv
*priv
)
3409 static const char * const texcoords
[8] =
3411 "fragment.texcoord[0]", "fragment.texcoord[1]", "fragment.texcoord[2]", "fragment.texcoord[3]",
3412 "fragment.texcoord[4]", "fragment.texcoord[5]", "fragment.texcoord[6]", "fragment.texcoord[7]"
3415 const struct wined3d_shader_signature_element
*input
;
3416 const char *semantic_name
;
3417 unsigned int semantic_idx
;
3419 if (args
->super
.vp_mode
== WINED3D_VP_MODE_SHADER
)
3421 /* That one is easy. The vertex shaders provide v0-v7 in
3422 * fragment.texcoord and v8 and v9 in fragment.color. */
3423 for (i
= 0; i
< 8; ++i
)
3425 priv
->ps_input
[i
] = texcoords
[i
];
3427 priv
->ps_input
[8] = "fragment.color.primary";
3428 priv
->ps_input
[9] = "fragment.color.secondary";
3432 /* The fragment shader has to collect the varyings on its own. In any case
3433 * properly load color0 and color1. In the case of pre-transformed
3434 * vertices also load texture coordinates. Set other attributes to 0.0.
3436 * For fixed-function this behavior is correct, according to the tests.
3437 * For pre-transformed we'd either need a replacement shader that can load
3438 * other attributes like BINORMAL, or load the texture coordinate
3439 * attribute pointers to match the fragment shader signature. */
3440 for (i
= 0; i
< shader
->input_signature
.element_count
; ++i
)
3442 input
= &shader
->input_signature
.elements
[i
];
3443 if (!(semantic_name
= input
->semantic_name
))
3445 semantic_idx
= input
->semantic_idx
;
3447 if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_COLOR
))
3450 priv
->ps_input
[input
->register_idx
] = "fragment.color.primary";
3451 else if (semantic_idx
== 1)
3452 priv
->ps_input
[input
->register_idx
] = "fragment.color.secondary";
3454 priv
->ps_input
[input
->register_idx
] = "0.0";
3456 else if (args
->super
.vp_mode
== WINED3D_VP_MODE_FF
)
3458 priv
->ps_input
[input
->register_idx
] = "0.0";
3460 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
3462 if (semantic_idx
< 8)
3463 priv
->ps_input
[input
->register_idx
] = texcoords
[semantic_idx
];
3465 priv
->ps_input
[input
->register_idx
] = "0.0";
3467 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_FOG
))
3470 priv
->ps_input
[input
->register_idx
] = "fragment.fogcoord";
3472 priv
->ps_input
[input
->register_idx
] = "0.0";
3476 priv
->ps_input
[input
->register_idx
] = "0.0";
3479 TRACE("v%u, semantic %s%u is %s\n", input
->register_idx
,
3480 semantic_name
, semantic_idx
, priv
->ps_input
[input
->register_idx
]);
3484 static void arbfp_add_linear_fog(struct wined3d_string_buffer
*buffer
,
3485 const char *fragcolor
, const char *tmp
)
3487 shader_addline(buffer
, "SUB %s.x, state.fog.params.z, fragment.fogcoord.x;\n", tmp
);
3488 shader_addline(buffer
, "MUL_SAT %s.x, %s.x, state.fog.params.w;\n", tmp
, tmp
);
3489 shader_addline(buffer
, "LRP %s.rgb, %s.x, %s, state.fog.color;\n", fragcolor
, tmp
, fragcolor
);
3492 /* Context activation is done by the caller. */
3493 static GLuint
shader_arb_generate_pshader(const struct wined3d_shader
*shader
,
3494 const struct wined3d_gl_info
*gl_info
, struct wined3d_string_buffer
*buffer
,
3495 const struct arb_ps_compile_args
*args
, struct arb_ps_compiled_shader
*compiled
)
3497 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
3500 unsigned int next_local
= 0;
3501 struct shader_arb_ctx_priv priv_ctx
;
3502 BOOL dcl_td
= FALSE
;
3503 BOOL want_nv_prog
= FALSE
;
3504 struct arb_pshader_private
*shader_priv
= shader
->backend_data
;
3505 BOOL custom_linear_fog
= FALSE
;
3510 unsigned int i
, found
= 0;
3512 map
= reg_maps
->temporary
;
3515 i
= wined3d_bit_scan(&map
);
3516 if ((shader
->u
.ps
.color0_mov
&& i
== shader
->u
.ps
.color0_reg
)
3517 || (reg_maps
->shader_version
.major
< 2 && !i
))
3520 sprintf(srgbtmp
[found
], "R%u", i
);
3522 if (found
== 4) break;
3527 sprintf(srgbtmp
[0], "TA");
3528 sprintf(srgbtmp
[1], "TB");
3529 sprintf(srgbtmp
[2], "TC");
3530 sprintf(srgbtmp
[3], "TD");
3534 sprintf(srgbtmp
[1], "TA");
3535 sprintf(srgbtmp
[2], "TB");
3536 sprintf(srgbtmp
[3], "TC");
3539 sprintf(srgbtmp
[2], "TA");
3540 sprintf(srgbtmp
[3], "TB");
3543 sprintf(srgbtmp
[3], "TA");
3549 /* Create the hw ARB shader */
3550 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
3551 priv_ctx
.gl_info
= gl_info
;
3552 priv_ctx
.cur_ps_args
= args
;
3553 priv_ctx
.compiled_fprog
= compiled
;
3554 priv_ctx
.cur_np2fixup_info
= &compiled
->np2fixup_info
;
3555 init_ps_input(shader
, args
, &priv_ctx
);
3556 list_init(&priv_ctx
.control_frames
);
3557 priv_ctx
.ps_post_process
= args
->super
.srgb_correction
;
3559 /* Avoid enabling NV_fragment_program* if we do not need it.
3561 * Enabling GL_NV_fragment_program_option causes the driver to occupy a temporary register,
3562 * and it slows down the shader execution noticeably(about 5%). Usually our instruction emulation
3563 * is faster than what we gain from using higher native instructions. There are some things though
3564 * that cannot be emulated. In that case enable the extensions.
3565 * If the extension is enabled, instruction handlers that support both ways will use it.
3567 * Testing shows no performance difference between OPTION NV_fragment_program2 and NV_fragment_program.
3568 * So enable the best we can get.
3570 if(reg_maps
->usesdsx
|| reg_maps
->usesdsy
|| reg_maps
->loop_depth
> 0 || reg_maps
->usestexldd
||
3571 reg_maps
->usestexldl
|| reg_maps
->usesfacing
|| reg_maps
->usesifc
|| reg_maps
->usescall
)
3573 want_nv_prog
= TRUE
;
3576 shader_addline(buffer
, "!!ARBfp1.0\n");
3577 if (want_nv_prog
&& gl_info
->supported
[NV_FRAGMENT_PROGRAM2
])
3579 shader_addline(buffer
, "OPTION NV_fragment_program2;\n");
3580 priv_ctx
.target_version
= NV3
;
3582 else if (want_nv_prog
&& gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
])
3584 shader_addline(buffer
, "OPTION NV_fragment_program;\n");
3585 priv_ctx
.target_version
= NV2
;
3589 /* This is an error - either we're advertising the wrong shader version, or aren't enforcing some
3592 ERR("The shader requires instructions that are not available in plain GL_ARB_fragment_program\n");
3595 priv_ctx
.target_version
= ARB
;
3598 if (reg_maps
->rt_mask
> 1)
3600 shader_addline(buffer
, "OPTION ARB_draw_buffers;\n");
3603 if (reg_maps
->shader_version
.major
< 3)
3605 switch (args
->super
.fog
)
3607 case WINED3D_FFP_PS_FOG_OFF
:
3609 case WINED3D_FFP_PS_FOG_LINEAR
:
3610 if (gl_info
->quirks
& WINED3D_QUIRK_BROKEN_ARB_FOG
)
3612 custom_linear_fog
= TRUE
;
3613 priv_ctx
.ps_post_process
= TRUE
;
3616 shader_addline(buffer
, "OPTION ARB_fog_linear;\n");
3618 case WINED3D_FFP_PS_FOG_EXP
:
3619 shader_addline(buffer
, "OPTION ARB_fog_exp;\n");
3621 case WINED3D_FFP_PS_FOG_EXP2
:
3622 shader_addline(buffer
, "OPTION ARB_fog_exp2;\n");
3627 /* For now always declare the temps. At least the Nvidia assembler optimizes completely
3628 * unused temps away(but occupies them for the whole shader if they're used once). Always
3629 * declaring them avoids tricky bookkeeping work
3631 shader_addline(buffer
, "TEMP TA;\n"); /* Used for modifiers */
3632 shader_addline(buffer
, "TEMP TB;\n"); /* Used for modifiers */
3633 shader_addline(buffer
, "TEMP TC;\n"); /* Used for modifiers */
3634 if(dcl_td
) shader_addline(buffer
, "TEMP TD;\n"); /* Used for sRGB writing */
3635 shader_addline(buffer
, "PARAM coefdiv = { 0.5, 0.25, 0.125, 0.0625 };\n");
3636 shader_addline(buffer
, "PARAM coefmul = { 2, 4, 8, 16 };\n");
3637 wined3d_ftoa(eps
, ftoa_tmp
);
3638 shader_addline(buffer
, "PARAM ps_helper_const = { 0.0, 1.0, %s, 0.0 };\n", ftoa_tmp
);
3640 if (reg_maps
->shader_version
.major
< 2)
3642 strcpy(fragcolor
, "R0");
3646 if (priv_ctx
.ps_post_process
)
3648 if (shader
->u
.ps
.color0_mov
)
3650 sprintf(fragcolor
, "R%u", shader
->u
.ps
.color0_reg
);
3654 shader_addline(buffer
, "TEMP TMP_COLOR;\n");
3655 strcpy(fragcolor
, "TMP_COLOR");
3658 strcpy(fragcolor
, "result.color");
3662 if (args
->super
.srgb_correction
)
3664 shader_addline(buffer
, "PARAM srgb_consts0 = ");
3665 shader_arb_append_imm_vec4(buffer
, &wined3d_srgb_const
[0].x
);
3666 shader_addline(buffer
, ";\n");
3667 shader_addline(buffer
, "PARAM srgb_consts1 = ");
3668 shader_arb_append_imm_vec4(buffer
, &wined3d_srgb_const
[1].x
);
3669 shader_addline(buffer
, ";\n");
3672 /* Base Declarations */
3673 shader_generate_arb_declarations(shader
, reg_maps
, buffer
, gl_info
, NULL
, &priv_ctx
);
3675 map
= reg_maps
->bumpmat
;
3678 unsigned char bump_const
;
3680 i
= wined3d_bit_scan(&map
);
3681 bump_const
= compiled
->numbumpenvmatconsts
;
3682 compiled
->bumpenvmatconst
[bump_const
].const_num
= WINED3D_CONST_NUM_UNUSED
;
3683 compiled
->bumpenvmatconst
[bump_const
].texunit
= i
;
3684 compiled
->luminanceconst
[bump_const
].const_num
= WINED3D_CONST_NUM_UNUSED
;
3685 compiled
->luminanceconst
[bump_const
].texunit
= i
;
3687 /* We can fit the constants into the constant limit for sure because texbem, texbeml, bem and beml are only supported
3688 * in 1.x shaders, and GL_ARB_fragment_program has a constant limit of 24 constants. So in the worst case we're loading
3689 * 8 shader constants, 8 bump matrices and 8 luminance parameters and are perfectly fine. (No NP2 fixup on bumpmapped
3690 * textures due to conditional NP2 restrictions)
3692 * Use local constants to load the bump env parameters, not program.env. This avoids collisions with d3d constants of
3693 * shaders in newer shader models. Since the bump env parameters have to share their space with NP2 fixup constants,
3694 * their location is shader dependent anyway and they cannot be loaded globally.
3696 compiled
->bumpenvmatconst
[bump_const
].const_num
= next_local
++;
3697 shader_addline(buffer
, "PARAM bumpenvmat%d = program.local[%d];\n",
3698 i
, compiled
->bumpenvmatconst
[bump_const
].const_num
);
3699 compiled
->numbumpenvmatconsts
= bump_const
+ 1;
3701 if (!(reg_maps
->luminanceparams
& (1u << i
)))
3704 compiled
->luminanceconst
[bump_const
].const_num
= next_local
++;
3705 shader_addline(buffer
, "PARAM luminance%d = program.local[%d];\n",
3706 i
, compiled
->luminanceconst
[bump_const
].const_num
);
3709 for (i
= 0; i
< WINED3D_MAX_CONSTS_I
; ++i
)
3711 compiled
->int_consts
[i
] = WINED3D_CONST_NUM_UNUSED
;
3712 if (reg_maps
->integer_constants
& (1u << i
) && priv_ctx
.target_version
>= NV2
)
3714 const unsigned int *control_values
= find_loop_control_values(shader
, i
);
3718 shader_addline(buffer
, "PARAM I%u = {%u, %u, %u, -1};\n", i
,
3719 control_values
[0], control_values
[1], control_values
[2]);
3723 compiled
->int_consts
[i
] = next_local
;
3724 compiled
->num_int_consts
++;
3725 shader_addline(buffer
, "PARAM I%u = program.local[%u];\n", i
, next_local
++);
3730 if(reg_maps
->vpos
|| reg_maps
->usesdsy
)
3732 compiled
->ycorrection
= next_local
;
3733 shader_addline(buffer
, "PARAM ycorrection = program.local[%u];\n", next_local
++);
3737 shader_addline(buffer
, "TEMP vpos;\n");
3738 /* ycorrection.x: Backbuffer height(onscreen) or 0(offscreen).
3739 * ycorrection.y: -1.0(onscreen), 1.0(offscreen)
3740 * ycorrection.z: 1.0
3741 * ycorrection.w: 0.0
3743 shader_addline(buffer
, "MAD vpos, fragment.position, ycorrection.zyww, ycorrection.wxww;\n");
3744 shader_addline(buffer
, "FLR vpos.xy, vpos;\n");
3749 compiled
->ycorrection
= WINED3D_CONST_NUM_UNUSED
;
3752 /* Load constants to fixup NP2 texcoords if there are still free constants left:
3753 * Constants (texture dimensions) for the NP2 fixup are loaded as local program parameters. This will consume
3754 * at most 8 (WINED3D_MAX_FRAGMENT_SAMPLERS / 2) parameters, which is highly unlikely, since the application had to
3755 * use 16 NP2 textures at the same time. In case that we run out of constants the fixup is simply not
3756 * applied / activated. This will probably result in wrong rendering of the texture, but will save us from
3757 * shader compilation errors and the subsequent errors when drawing with this shader. */
3758 if (priv_ctx
.cur_ps_args
->super
.np2_fixup
) {
3759 unsigned char cur_fixup_sampler
= 0;
3761 struct arb_ps_np2fixup_info
* const fixup
= priv_ctx
.cur_np2fixup_info
;
3762 const WORD map
= priv_ctx
.cur_ps_args
->super
.np2_fixup
;
3763 const UINT max_lconsts
= gl_info
->limits
.arb_ps_local_constants
;
3765 fixup
->offset
= next_local
;
3766 fixup
->super
.active
= 0;
3768 for (i
= 0; i
< WINED3D_MAX_FRAGMENT_SAMPLERS
; ++i
)
3770 if (!(map
& (1u << i
)))
3773 if (fixup
->offset
+ (cur_fixup_sampler
>> 1) < max_lconsts
)
3775 fixup
->super
.active
|= (1u << i
);
3776 fixup
->super
.idx
[i
] = cur_fixup_sampler
++;
3780 FIXME("No free constant found to load NP2 fixup data into shader. "
3781 "Sampling from this texture will probably look wrong.\n");
3786 fixup
->super
.num_consts
= (cur_fixup_sampler
+ 1) >> 1;
3787 if (fixup
->super
.num_consts
) {
3788 shader_addline(buffer
, "PARAM np2fixup[%u] = { program.env[%u..%u] };\n",
3789 fixup
->super
.num_consts
, fixup
->offset
, fixup
->super
.num_consts
+ fixup
->offset
- 1);
3793 if (shader_priv
->clipplane_emulation
!= ~0U && args
->clip
)
3795 shader_addline(buffer
, "KIL fragment.texcoord[%u];\n", shader_priv
->clipplane_emulation
);
3798 /* Base Shader Body */
3799 if (FAILED(shader_generate_code(shader
, buffer
, reg_maps
, &priv_ctx
, NULL
, NULL
)))
3802 if(args
->super
.srgb_correction
) {
3803 arbfp_add_sRGB_correction(buffer
, fragcolor
, srgbtmp
[0], srgbtmp
[1], srgbtmp
[2], srgbtmp
[3],
3804 priv_ctx
.target_version
>= NV2
);
3807 if (custom_linear_fog
)
3808 arbfp_add_linear_fog(buffer
, fragcolor
, "TA");
3810 if(strcmp(fragcolor
, "result.color")) {
3811 shader_addline(buffer
, "MOV result.color, %s;\n", fragcolor
);
3813 shader_addline(buffer
, "END\n");
3815 /* TODO: change to resource.glObjectHandle or something like that */
3816 GL_EXTCALL(glGenProgramsARB(1, &retval
));
3818 TRACE("Creating a hw pixel shader, prg=%d\n", retval
);
3819 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, retval
));
3821 TRACE("Created hw pixel shader, prg=%d\n", retval
);
3822 if (!shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
->buffer
))
3828 static int compare_sig(const struct wined3d_shader_signature
*sig1
, const struct wined3d_shader_signature
*sig2
)
3833 if (sig1
->element_count
!= sig2
->element_count
)
3834 return sig1
->element_count
< sig2
->element_count
? -1 : 1;
3836 for (i
= 0; i
< sig1
->element_count
; ++i
)
3838 const struct wined3d_shader_signature_element
*e1
, *e2
;
3840 e1
= &sig1
->elements
[i
];
3841 e2
= &sig2
->elements
[i
];
3843 if (!e1
->semantic_name
|| !e2
->semantic_name
)
3845 /* Compare pointers, not contents. One string is NULL (element
3846 * does not exist), the other one is not NULL. */
3847 if (e1
->semantic_name
!= e2
->semantic_name
)
3848 return e1
->semantic_name
< e2
->semantic_name
? -1 : 1;
3852 if ((ret
= strcmp(e1
->semantic_name
, e2
->semantic_name
)))
3854 if ((ret
= wined3d_uint32_compare(e1
->semantic_idx
, e2
->semantic_idx
)))
3856 if ((ret
= wined3d_uint32_compare(e1
->sysval_semantic
, e2
->sysval_semantic
)))
3858 if ((ret
= wined3d_uint32_compare(e1
->component_type
, e2
->component_type
)))
3860 if ((ret
= wined3d_uint32_compare(e1
->register_idx
, e2
->register_idx
)))
3862 if ((ret
= wined3d_uint32_compare(e1
->mask
, e2
->mask
)))
3868 static void clone_sig(struct wined3d_shader_signature
*new, const struct wined3d_shader_signature
*sig
)
3873 new->element_count
= sig
->element_count
;
3874 new->elements
= heap_calloc(new->element_count
, sizeof(*new->elements
));
3875 for (i
= 0; i
< sig
->element_count
; ++i
)
3877 new->elements
[i
] = sig
->elements
[i
];
3879 if (!new->elements
[i
].semantic_name
)
3882 /* Clone the semantic string */
3883 name
= heap_alloc(strlen(sig
->elements
[i
].semantic_name
) + 1);
3884 strcpy(name
, sig
->elements
[i
].semantic_name
);
3885 new->elements
[i
].semantic_name
= name
;
3889 static unsigned int find_input_signature(struct shader_arb_priv
*priv
, const struct wined3d_shader_signature
*sig
)
3891 struct wine_rb_entry
*entry
= wine_rb_get(&priv
->signature_tree
, sig
);
3892 struct ps_signature
*found_sig
;
3896 found_sig
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
3897 TRACE("Found existing signature %u\n", found_sig
->idx
);
3898 return found_sig
->idx
;
3900 found_sig
= heap_alloc_zero(sizeof(*found_sig
));
3901 clone_sig(&found_sig
->sig
, sig
);
3902 found_sig
->idx
= priv
->ps_sig_number
++;
3903 TRACE("New signature stored and assigned number %u\n", found_sig
->idx
);
3904 if(wine_rb_put(&priv
->signature_tree
, sig
, &found_sig
->entry
) == -1)
3906 ERR("Failed to insert program entry.\n");
3908 return found_sig
->idx
;
3911 static void init_output_registers(const struct wined3d_shader
*shader
,
3912 const struct wined3d_shader_signature
*ps_input_sig
,
3913 struct shader_arb_ctx_priv
*priv_ctx
, struct arb_vs_compiled_shader
*compiled
)
3916 static const char * const texcoords
[8] =
3918 "result.texcoord[0]", "result.texcoord[1]", "result.texcoord[2]", "result.texcoord[3]",
3919 "result.texcoord[4]", "result.texcoord[5]", "result.texcoord[6]", "result.texcoord[7]"
3921 /* Write generic input varyings 0 to 7 to result.texcoord[], varying 8 to result.color.primary
3922 * and varying 9 to result.color.secondary
3924 static const char * const decl_idx_to_string
[MAX_REG_INPUT
] =
3926 "result.texcoord[0]", "result.texcoord[1]", "result.texcoord[2]", "result.texcoord[3]",
3927 "result.texcoord[4]", "result.texcoord[5]", "result.texcoord[6]", "result.texcoord[7]",
3928 "result.color.primary", "result.color.secondary"
3933 TRACE("Pixel shader uses builtin varyings\n");
3934 /* Map builtins to builtins */
3935 for(i
= 0; i
< 8; i
++)
3937 priv_ctx
->texcrd_output
[i
] = texcoords
[i
];
3939 priv_ctx
->color_output
[0] = "result.color.primary";
3940 priv_ctx
->color_output
[1] = "result.color.secondary";
3941 priv_ctx
->fog_output
= "TMP_FOGCOORD";
3943 /* Map declared regs to builtins. Use "TA" to /dev/null unread output */
3944 for (i
= 0; i
< shader
->output_signature
.element_count
; ++i
)
3946 const struct wined3d_shader_signature_element
*output
= &shader
->output_signature
.elements
[i
];
3948 if (!output
->semantic_name
)
3951 if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_POSITION
))
3953 TRACE("o%u is TMP_OUT\n", output
->register_idx
);
3954 if (!output
->semantic_idx
)
3955 priv_ctx
->vs_output
[output
->register_idx
] = "TMP_OUT";
3957 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
3959 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_PSIZE
))
3961 TRACE("o%u is result.pointsize\n", output
->register_idx
);
3962 if (!output
->semantic_idx
)
3963 priv_ctx
->vs_output
[output
->register_idx
] = "result.pointsize";
3965 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
3967 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_COLOR
))
3969 TRACE("o%u is result.color.?, idx %u\n", output
->register_idx
, output
->semantic_idx
);
3970 if (!output
->semantic_idx
)
3971 priv_ctx
->vs_output
[output
->register_idx
] = "result.color.primary";
3972 else if (output
->semantic_idx
== 1)
3973 priv_ctx
->vs_output
[output
->register_idx
] = "result.color.secondary";
3974 else priv_ctx
->vs_output
[output
->register_idx
] = "TA";
3976 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
3978 TRACE("o%u is result.texcoord[%u]\n", output
->register_idx
, output
->semantic_idx
);
3979 if (output
->semantic_idx
>= 8)
3980 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
3982 priv_ctx
->vs_output
[output
->register_idx
] = texcoords
[output
->semantic_idx
];
3984 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_FOG
))
3986 TRACE("o%u is result.fogcoord\n", output
->register_idx
);
3987 if (output
->semantic_idx
> 0)
3988 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
3990 priv_ctx
->vs_output
[output
->register_idx
] = "result.fogcoord";
3994 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4000 TRACE("Pixel shader uses declared varyings\n");
4002 /* Map builtin to declared. /dev/null the results by default to the TA temp reg */
4003 for(i
= 0; i
< 8; i
++)
4005 priv_ctx
->texcrd_output
[i
] = "TA";
4007 priv_ctx
->color_output
[0] = "TA";
4008 priv_ctx
->color_output
[1] = "TA";
4009 priv_ctx
->fog_output
= "TA";
4011 for (i
= 0; i
< ps_input_sig
->element_count
; ++i
)
4013 const struct wined3d_shader_signature_element
*input
= &ps_input_sig
->elements
[i
];
4015 if (!input
->semantic_name
)
4018 /* If a declared input register is not written by builtin arguments, don't write to it.
4019 * GL_NV_vertex_program makes sure the input defaults to 0.0, which is correct with D3D
4021 * Don't care about POSITION and PSIZE here - this is a builtin vertex shader, position goes
4022 * to TMP_OUT in any case
4024 if (shader_match_semantic(input
->semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
4026 if (input
->semantic_idx
< 8)
4027 priv_ctx
->texcrd_output
[input
->semantic_idx
] = decl_idx_to_string
[input
->register_idx
];
4029 else if (shader_match_semantic(input
->semantic_name
, WINED3D_DECL_USAGE_COLOR
))
4031 if (input
->semantic_idx
< 2)
4032 priv_ctx
->color_output
[input
->semantic_idx
] = decl_idx_to_string
[input
->register_idx
];
4034 else if (shader_match_semantic(input
->semantic_name
, WINED3D_DECL_USAGE_FOG
))
4036 if (!input
->semantic_idx
)
4037 priv_ctx
->fog_output
= decl_idx_to_string
[input
->register_idx
];
4044 if (!strcmp(decl_idx_to_string
[input
->register_idx
], "result.color.primary")
4045 || !strcmp(decl_idx_to_string
[input
->register_idx
], "result.color.secondary"))
4047 compiled
->need_color_unclamp
= TRUE
;
4051 /* Map declared to declared */
4052 for (i
= 0; i
< shader
->output_signature
.element_count
; ++i
)
4054 const struct wined3d_shader_signature_element
*output
= &shader
->output_signature
.elements
[i
];
4056 /* Write unread output to TA to throw them away */
4057 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4059 if (!output
->semantic_name
)
4062 if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_POSITION
) && !output
->semantic_idx
)
4064 priv_ctx
->vs_output
[output
->register_idx
] = "TMP_OUT";
4067 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_PSIZE
) && !output
->semantic_idx
)
4069 priv_ctx
->vs_output
[output
->register_idx
] = "result.pointsize";
4073 for (j
= 0; j
< ps_input_sig
->element_count
; ++j
)
4075 const struct wined3d_shader_signature_element
*input
= &ps_input_sig
->elements
[j
];
4077 if (!input
->semantic_name
)
4080 if (!strcmp(input
->semantic_name
, output
->semantic_name
)
4081 && input
->semantic_idx
== output
->semantic_idx
)
4083 priv_ctx
->vs_output
[output
->register_idx
] = decl_idx_to_string
[input
->register_idx
];
4085 if (!strcmp(priv_ctx
->vs_output
[output
->register_idx
], "result.color.primary")
4086 || !strcmp(priv_ctx
->vs_output
[output
->register_idx
], "result.color.secondary"))
4088 compiled
->need_color_unclamp
= TRUE
;
4095 /* Context activation is done by the caller. */
4096 static GLuint
shader_arb_generate_vshader(const struct wined3d_shader
*shader
,
4097 const struct wined3d_gl_info
*gl_info
, struct wined3d_string_buffer
*buffer
,
4098 const struct arb_vs_compile_args
*args
, struct arb_vs_compiled_shader
*compiled
,
4099 const struct wined3d_shader_signature
*ps_input_sig
)
4101 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
4102 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
4104 unsigned int next_local
= 0;
4105 struct shader_arb_ctx_priv priv_ctx
;
4108 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
4109 priv_ctx
.gl_info
= gl_info
;
4110 priv_ctx
.cur_vs_args
= args
;
4111 list_init(&priv_ctx
.control_frames
);
4112 init_output_registers(shader
, ps_input_sig
, &priv_ctx
, compiled
);
4114 /* Create the hw ARB shader */
4115 shader_addline(buffer
, "!!ARBvp1.0\n");
4117 /* Always enable the NV extension if available. Unlike fragment shaders, there is no
4118 * mesurable performance penalty, and we can always make use of it for clipplanes.
4120 if (gl_info
->supported
[NV_VERTEX_PROGRAM3
])
4122 shader_addline(buffer
, "OPTION NV_vertex_program3;\n");
4123 priv_ctx
.target_version
= NV3
;
4124 shader_addline(buffer
, "ADDRESS aL;\n");
4126 else if (gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
])
4128 shader_addline(buffer
, "OPTION NV_vertex_program2;\n");
4129 priv_ctx
.target_version
= NV2
;
4130 shader_addline(buffer
, "ADDRESS aL;\n");
4132 priv_ctx
.target_version
= ARB
;
4135 shader_addline(buffer
, "TEMP TMP_OUT;\n");
4137 shader_addline(buffer
, "TEMP TMP_FOGCOORD;\n");
4138 if (need_helper_const(shader_data
, reg_maps
, gl_info
))
4141 wined3d_ftoa(eps
, ftoa_tmp
);
4142 shader_addline(buffer
, "PARAM helper_const = { 0.0, 1.0, 2.0, %s};\n", ftoa_tmp
);
4144 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
))
4146 shader_addline(buffer
, "PARAM rel_addr_const = { 0.5, %d.0, 0.0, 0.0 };\n", shader_data
->rel_offset
);
4147 shader_addline(buffer
, "TEMP A0_SHADOW;\n");
4150 shader_addline(buffer
, "TEMP TA;\n");
4151 shader_addline(buffer
, "TEMP TB;\n");
4153 /* Base Declarations */
4154 shader_generate_arb_declarations(shader
, reg_maps
, buffer
, gl_info
,
4155 &priv_ctx
.vs_clipplanes
, &priv_ctx
);
4157 for (i
= 0; i
< WINED3D_MAX_CONSTS_I
; ++i
)
4159 compiled
->int_consts
[i
] = WINED3D_CONST_NUM_UNUSED
;
4160 if (reg_maps
->integer_constants
& (1u << i
) && priv_ctx
.target_version
>= NV2
)
4162 const unsigned int *control_values
= find_loop_control_values(shader
, i
);
4166 shader_addline(buffer
, "PARAM I%u = {%u, %u, %u, -1};\n", i
,
4167 control_values
[0], control_values
[1], control_values
[2]);
4171 compiled
->int_consts
[i
] = next_local
;
4172 compiled
->num_int_consts
++;
4173 shader_addline(buffer
, "PARAM I%u = program.local[%u];\n", i
, next_local
++);
4178 /* We need a constant to fixup the final position */
4179 shader_addline(buffer
, "PARAM posFixup = program.local[%u];\n", next_local
);
4180 compiled
->pos_fixup
= next_local
++;
4182 /* Initialize output parameters. GL_ARB_vertex_program does not require special initialization values
4183 * for output parameters. D3D in theory does not do that either, but some applications depend on a
4184 * proper initialization of the secondary color, and programs using the fixed function pipeline without
4185 * a replacement shader depend on the texcoord.w being set properly.
4187 * GL_NV_vertex_program defines that all output values are initialized to {0.0, 0.0, 0.0, 1.0}. This
4188 * assertion is in effect even when using GL_ARB_vertex_program without any NV specific additions. So
4189 * skip this if NV_vertex_program is supported. Otherwise, initialize the secondary color. For the tex-
4190 * coords, we have a flag in the opengl caps. Many cards do not require the texcoord being set, and
4191 * this can eat a number of instructions, so skip it unless this cap is set as well
4193 if (!gl_info
->supported
[NV_VERTEX_PROGRAM
])
4195 const char *color_init
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_0001
);
4196 shader_addline(buffer
, "MOV result.color.secondary, %s;\n", color_init
);
4199 /* The shader starts with the main function */
4200 priv_ctx
.in_main_func
= TRUE
;
4201 /* Base Shader Body */
4202 if (FAILED(shader_generate_code(shader
, buffer
, reg_maps
, &priv_ctx
, NULL
, NULL
)))
4205 if (!priv_ctx
.footer_written
) vshader_add_footer(&priv_ctx
,
4206 shader_data
, args
, reg_maps
, gl_info
, buffer
);
4208 shader_addline(buffer
, "END\n");
4210 /* TODO: change to resource.glObjectHandle or something like that */
4211 GL_EXTCALL(glGenProgramsARB(1, &ret
));
4213 TRACE("Creating a hw vertex shader, prg=%d\n", ret
);
4214 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, ret
));
4216 TRACE("Created hw vertex shader, prg=%d\n", ret
);
4217 if (!shader_arb_compile(gl_info
, GL_VERTEX_PROGRAM_ARB
, buffer
->buffer
))
4223 /* Context activation is done by the caller. */
4224 static struct arb_ps_compiled_shader
*find_arb_pshader(struct wined3d_context_gl
*context_gl
,
4225 struct wined3d_shader
*shader
, const struct arb_ps_compile_args
*args
)
4227 const struct wined3d_d3d_info
*d3d_info
= context_gl
->c
.d3d_info
;
4228 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
4229 struct wined3d_device
*device
= shader
->device
;
4232 struct arb_ps_compiled_shader
*new_array
;
4233 struct wined3d_string_buffer buffer
;
4234 struct arb_pshader_private
*shader_data
;
4237 if (!shader
->backend_data
)
4239 struct shader_arb_priv
*priv
= device
->shader_priv
;
4241 shader
->backend_data
= heap_alloc_zero(sizeof(*shader_data
));
4242 shader_data
= shader
->backend_data
;
4243 shader_data
->clamp_consts
= shader
->reg_maps
.shader_version
.major
== 1;
4245 if (shader
->reg_maps
.shader_version
.major
< 3)
4246 shader_data
->input_signature_idx
= ~0U;
4248 shader_data
->input_signature_idx
= find_input_signature(priv
, &shader
->input_signature
);
4250 TRACE("Shader got assigned input signature index %u\n", shader_data
->input_signature_idx
);
4252 if (!d3d_info
->vs_clipping
)
4253 shader_data
->clipplane_emulation
= shader_find_free_input_register(&shader
->reg_maps
,
4254 d3d_info
->limits
.ffp_blend_stages
- 1);
4256 shader_data
->clipplane_emulation
= ~0U;
4258 shader_data
= shader
->backend_data
;
4260 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4261 * so a linear search is more performant than a hashmap or a binary search
4262 * (cache coherency etc)
4264 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4266 if (!memcmp(&shader_data
->gl_shaders
[i
].args
, args
, sizeof(*args
)))
4267 return &shader_data
->gl_shaders
[i
];
4270 TRACE("No matching GL shader found, compiling a new shader\n");
4271 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4272 if (shader_data
->num_gl_shaders
)
4274 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4275 new_array
= HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, shader_data
->gl_shaders
,
4276 new_size
* sizeof(*shader_data
->gl_shaders
));
4280 new_array
= heap_alloc_zero(sizeof(*shader_data
->gl_shaders
));
4285 ERR("Out of memory\n");
4288 shader_data
->gl_shaders
= new_array
;
4289 shader_data
->shader_array_size
= new_size
;
4292 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4294 if (!string_buffer_init(&buffer
))
4296 ERR("Failed to initialize shader buffer.\n");
4300 ret
= shader_arb_generate_pshader(shader
, gl_info
, &buffer
, args
,
4301 &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
]);
4302 string_buffer_free(&buffer
);
4303 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].prgId
= ret
;
4305 return &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++];
4308 static inline BOOL
vs_args_equal(const struct arb_vs_compile_args
*stored
, const struct arb_vs_compile_args
*new,
4309 const DWORD use_map
, BOOL skip_int
) {
4310 if((stored
->super
.swizzle_map
& use_map
) != new->super
.swizzle_map
) return FALSE
;
4311 if(stored
->super
.clip_enabled
!= new->super
.clip_enabled
) return FALSE
;
4312 if(stored
->super
.fog_src
!= new->super
.fog_src
) return FALSE
;
4313 if(stored
->clip
.boolclip_compare
!= new->clip
.boolclip_compare
) return FALSE
;
4314 if(stored
->ps_signature
!= new->ps_signature
) return FALSE
;
4315 if(stored
->vertex
.samplers_compare
!= new->vertex
.samplers_compare
) return FALSE
;
4316 if(skip_int
) return TRUE
;
4318 return !memcmp(stored
->loop_ctrl
, new->loop_ctrl
, sizeof(stored
->loop_ctrl
));
4321 static struct arb_vs_compiled_shader
*find_arb_vshader(struct wined3d_shader
*shader
,
4322 const struct wined3d_gl_info
*gl_info
, DWORD use_map
, const struct arb_vs_compile_args
*args
,
4323 const struct wined3d_shader_signature
*ps_input_sig
)
4327 struct arb_vs_compiled_shader
*new_array
;
4328 struct wined3d_string_buffer buffer
;
4329 struct arb_vshader_private
*shader_data
;
4332 if (!shader
->backend_data
)
4334 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
4336 shader
->backend_data
= heap_alloc_zero(sizeof(*shader_data
));
4337 shader_data
= shader
->backend_data
;
4339 if ((gl_info
->quirks
& WINED3D_QUIRK_ARB_VS_OFFSET_LIMIT
)
4340 && reg_maps
->min_rel_offset
<= reg_maps
->max_rel_offset
)
4342 if (reg_maps
->max_rel_offset
- reg_maps
->min_rel_offset
> 127)
4344 FIXME("The difference between the minimum and maximum relative offset is > 127.\n");
4345 FIXME("Which this OpenGL implementation does not support. Try using GLSL.\n");
4346 FIXME("Min: %u, Max: %u.\n", reg_maps
->min_rel_offset
, reg_maps
->max_rel_offset
);
4348 else if (reg_maps
->max_rel_offset
- reg_maps
->min_rel_offset
> 63)
4349 shader_data
->rel_offset
= reg_maps
->min_rel_offset
+ 63;
4350 else if (reg_maps
->max_rel_offset
> 63)
4351 shader_data
->rel_offset
= reg_maps
->min_rel_offset
;
4354 shader_data
= shader
->backend_data
;
4356 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4357 * so a linear search is more performant than a hashmap or a binary search
4358 * (cache coherency etc)
4360 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
4361 if (vs_args_equal(&shader_data
->gl_shaders
[i
].args
, args
,
4362 use_map
, gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]))
4364 return &shader_data
->gl_shaders
[i
];
4368 TRACE("No matching GL shader found, compiling a new shader\n");
4370 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4371 if (shader_data
->num_gl_shaders
)
4373 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4374 new_array
= HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, shader_data
->gl_shaders
,
4375 new_size
* sizeof(*shader_data
->gl_shaders
));
4379 new_array
= heap_alloc_zero(sizeof(*shader_data
->gl_shaders
));
4384 ERR("Out of memory\n");
4387 shader_data
->gl_shaders
= new_array
;
4388 shader_data
->shader_array_size
= new_size
;
4391 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4393 if (!string_buffer_init(&buffer
))
4395 ERR("Failed to initialize shader buffer.\n");
4399 ret
= shader_arb_generate_vshader(shader
, gl_info
, &buffer
, args
,
4400 &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
],
4402 string_buffer_free(&buffer
);
4403 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].prgId
= ret
;
4405 return &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++];
4408 static void find_arb_ps_compile_args(const struct wined3d_state
*state
,
4409 const struct wined3d_context_gl
*context_gl
, const struct wined3d_shader
*shader
,
4410 struct arb_ps_compile_args
*args
)
4412 const struct wined3d_d3d_info
*d3d_info
= context_gl
->c
.d3d_info
;
4413 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
4417 find_ps_compile_args(state
, shader
, context_gl
->c
.stream_info
.position_transformed
, &args
->super
, &context_gl
->c
);
4419 /* This forces all local boolean constants to 1 to make them stateblock independent */
4420 args
->bools
= shader
->reg_maps
.local_bool_consts
;
4422 for (i
= 0; i
< WINED3D_MAX_CONSTS_B
; ++i
)
4424 if (state
->ps_consts_b
[i
])
4425 args
->bools
|= ( 1u << i
);
4428 /* Only enable the clip plane emulation KIL if at least one clipplane is enabled. The KIL instruction
4429 * is quite expensive because it forces the driver to disable early Z discards. It is cheaper to
4430 * duplicate the shader than have a no-op KIL instruction in every shader
4432 if (!d3d_info
->vs_clipping
&& use_vs(state
)
4433 && state
->render_states
[WINED3D_RS_CLIPPING
]
4434 && state
->render_states
[WINED3D_RS_CLIPPLANEENABLE
])
4439 /* Skip if unused or local, or supported natively */
4440 int_skip
= ~shader
->reg_maps
.integer_constants
| shader
->reg_maps
.local_int_consts
;
4441 if (int_skip
== 0xffff || gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
])
4443 memset(args
->loop_ctrl
, 0, sizeof(args
->loop_ctrl
));
4447 for (i
= 0; i
< WINED3D_MAX_CONSTS_I
; ++i
)
4449 if (int_skip
& (1u << i
))
4451 args
->loop_ctrl
[i
][0] = 0;
4452 args
->loop_ctrl
[i
][1] = 0;
4453 args
->loop_ctrl
[i
][2] = 0;
4457 args
->loop_ctrl
[i
][0] = state
->ps_consts_i
[i
].x
;
4458 args
->loop_ctrl
[i
][1] = state
->ps_consts_i
[i
].y
;
4459 args
->loop_ctrl
[i
][2] = state
->ps_consts_i
[i
].z
;
4464 static void find_arb_vs_compile_args(const struct wined3d_state
*state
,
4465 const struct wined3d_context_gl
*context_gl
, const struct wined3d_shader
*shader
,
4466 struct arb_vs_compile_args
*args
)
4468 const struct wined3d_d3d_info
*d3d_info
= context_gl
->c
.d3d_info
;
4469 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
4470 const struct wined3d_device
*device
= shader
->device
;
4471 const struct wined3d_adapter
*adapter
= device
->adapter
;
4475 find_vs_compile_args(state
, shader
, &args
->super
, &context_gl
->c
);
4477 args
->clip
.boolclip_compare
= 0;
4480 const struct wined3d_shader
*ps
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
4481 const struct arb_pshader_private
*shader_priv
= ps
->backend_data
;
4482 args
->ps_signature
= shader_priv
->input_signature_idx
;
4484 args
->clip
.boolclip
.clip_texcoord
= shader_priv
->clipplane_emulation
+ 1;
4488 args
->ps_signature
= ~0;
4489 if (!d3d_info
->vs_clipping
&& adapter
->fragment_pipe
== &arbfp_fragment_pipeline
)
4490 args
->clip
.boolclip
.clip_texcoord
= ffp_clip_emul(&context_gl
->c
) ? d3d_info
->limits
.ffp_blend_stages
: 0;
4491 /* Otherwise: Setting boolclip_compare set clip_texcoord to 0 */
4494 if (args
->clip
.boolclip
.clip_texcoord
)
4496 if (state
->render_states
[WINED3D_RS_CLIPPING
])
4497 args
->clip
.boolclip
.clipplane_mask
= (unsigned char)state
->render_states
[WINED3D_RS_CLIPPLANEENABLE
];
4498 /* clipplane_mask was set to 0 by setting boolclip_compare to 0 */
4501 /* This forces all local boolean constants to 1 to make them stateblock independent */
4502 args
->clip
.boolclip
.bools
= shader
->reg_maps
.local_bool_consts
;
4503 /* TODO: Figure out if it would be better to store bool constants as bitmasks in the stateblock */
4504 for (i
= 0; i
< WINED3D_MAX_CONSTS_B
; ++i
)
4506 if (state
->vs_consts_b
[i
])
4507 args
->clip
.boolclip
.bools
|= (1u << i
);
4510 args
->vertex
.samplers
[0] = context_gl
->tex_unit_map
[WINED3D_MAX_FRAGMENT_SAMPLERS
+ 0];
4511 args
->vertex
.samplers
[1] = context_gl
->tex_unit_map
[WINED3D_MAX_FRAGMENT_SAMPLERS
+ 1];
4512 args
->vertex
.samplers
[2] = context_gl
->tex_unit_map
[WINED3D_MAX_FRAGMENT_SAMPLERS
+ 2];
4513 args
->vertex
.samplers
[3] = 0;
4515 /* Skip if unused or local */
4516 int_skip
= ~shader
->reg_maps
.integer_constants
| shader
->reg_maps
.local_int_consts
;
4517 /* This is about flow control, not clipping. */
4518 if (int_skip
== 0xffff || gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
])
4520 memset(args
->loop_ctrl
, 0, sizeof(args
->loop_ctrl
));
4524 for (i
= 0; i
< WINED3D_MAX_CONSTS_I
; ++i
)
4526 if (int_skip
& (1u << i
))
4528 args
->loop_ctrl
[i
][0] = 0;
4529 args
->loop_ctrl
[i
][1] = 0;
4530 args
->loop_ctrl
[i
][2] = 0;
4534 args
->loop_ctrl
[i
][0] = state
->vs_consts_i
[i
].x
;
4535 args
->loop_ctrl
[i
][1] = state
->vs_consts_i
[i
].y
;
4536 args
->loop_ctrl
[i
][2] = state
->vs_consts_i
[i
].z
;
4541 /* Context activation is done by the caller. */
4542 static void shader_arb_select(void *shader_priv
, struct wined3d_context
*context
,
4543 const struct wined3d_state
*state
)
4545 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
4546 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
4547 struct shader_arb_priv
*priv
= shader_priv
;
4550 /* Deal with pixel shaders first so the vertex shader arg function has the input signature ready */
4553 struct wined3d_shader
*ps
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
4554 struct arb_ps_compile_args compile_args
;
4555 struct arb_ps_compiled_shader
*compiled
;
4557 TRACE("Using pixel shader %p.\n", ps
);
4558 find_arb_ps_compile_args(state
, context_gl
, ps
, &compile_args
);
4559 compiled
= find_arb_pshader(context_gl
, ps
, &compile_args
);
4560 priv
->current_fprogram_id
= compiled
->prgId
;
4561 priv
->compiled_fprog
= compiled
;
4563 /* Bind the fragment program */
4564 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, priv
->current_fprogram_id
));
4565 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->current_fprogram_id);");
4567 if (!priv
->use_arbfp_fixed_func
)
4568 priv
->fragment_pipe
->fp_enable(context
, FALSE
);
4570 /* Enable OpenGL fragment programs. */
4571 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
4572 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
4574 TRACE("Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", priv
->current_fprogram_id
);
4576 /* Pixel Shader 1.x constants are clamped to [-1;1], Pixel Shader 2.0 constants are not. If switching between
4577 * a 1.x and newer shader, reload the first 8 constants
4579 if (priv
->last_ps_const_clamped
!= ((struct arb_pshader_private
*)ps
->backend_data
)->clamp_consts
)
4581 priv
->last_ps_const_clamped
= ((struct arb_pshader_private
*)ps
->backend_data
)->clamp_consts
;
4582 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, 8);
4583 for(i
= 0; i
< 8; i
++)
4585 priv
->pshader_const_dirty
[i
] = 1;
4587 /* Also takes care of loading local constants */
4588 shader_arb_load_constants_internal(shader_priv
, context_gl
, state
, TRUE
, FALSE
, TRUE
);
4592 UINT rt_height
= state
->fb
.render_targets
[0]->height
;
4593 shader_arb_ps_local_constants(compiled
, context_gl
, state
, rt_height
);
4596 /* Force constant reloading for the NP2 fixup (see comment in shader_glsl_select for more info) */
4597 if (compiled
->np2fixup_info
.super
.active
)
4598 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_NP2_FIXUP
;
4600 if (ps
->load_local_constsF
)
4601 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_F
;
4605 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
] && !priv
->use_arbfp_fixed_func
)
4607 /* Disable only if we're not using arbfp fixed function fragment
4608 * processing. If this is used, keep GL_FRAGMENT_PROGRAM_ARB
4609 * enabled, and the fixed function pipeline will bind the fixed
4610 * function replacement shader. */
4611 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4612 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4613 priv
->current_fprogram_id
= 0;
4615 priv
->fragment_pipe
->fp_enable(context
, TRUE
);
4620 struct wined3d_shader
*vs
= state
->shader
[WINED3D_SHADER_TYPE_VERTEX
];
4621 struct arb_vs_compile_args compile_args
;
4622 struct arb_vs_compiled_shader
*compiled
;
4623 const struct wined3d_shader_signature
*ps_input_sig
;
4625 TRACE("Using vertex shader %p\n", vs
);
4626 find_arb_vs_compile_args(state
, context_gl
, vs
, &compile_args
);
4628 /* Instead of searching for the signature in the signature list, read the one from the
4629 * current pixel shader. It's maybe not the shader where the signature came from, but it
4630 * is the same signature and faster to find. */
4631 if (compile_args
.ps_signature
== ~0U)
4632 ps_input_sig
= NULL
;
4634 ps_input_sig
= &state
->shader
[WINED3D_SHADER_TYPE_PIXEL
]->input_signature
;
4636 compiled
= find_arb_vshader(vs
, gl_info
, context
->stream_info
.use_map
,
4637 &compile_args
, ps_input_sig
);
4638 priv
->current_vprogram_id
= compiled
->prgId
;
4639 priv
->compiled_vprog
= compiled
;
4641 /* Bind the vertex program */
4642 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->current_vprogram_id
));
4643 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->current_vprogram_id);");
4645 priv
->vertex_pipe
->vp_enable(context
, FALSE
);
4647 /* Enable OpenGL vertex programs */
4648 gl_info
->gl_ops
.gl
.p_glEnable(GL_VERTEX_PROGRAM_ARB
);
4649 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
4650 TRACE("Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", priv
->current_vprogram_id
);
4651 shader_arb_vs_local_constants(compiled
, context_gl
, state
);
4653 if(priv
->last_vs_color_unclamp
!= compiled
->need_color_unclamp
) {
4654 priv
->last_vs_color_unclamp
= compiled
->need_color_unclamp
;
4656 if (gl_info
->supported
[ARB_COLOR_BUFFER_FLOAT
])
4658 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, !compiled
->need_color_unclamp
));
4659 checkGLcall("glClampColorARB");
4661 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
4665 if (vs
->load_local_constsF
)
4666 context
->constant_update_mask
|= WINED3D_SHADER_CONST_VS_F
;
4670 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
4672 priv
->current_vprogram_id
= 0;
4673 gl_info
->gl_ops
.gl
.p_glDisable(GL_VERTEX_PROGRAM_ARB
);
4674 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4676 priv
->vertex_pipe
->vp_enable(context
, TRUE
);
4680 static void shader_arb_select_compute(void *shader_priv
, struct wined3d_context
*context
,
4681 const struct wined3d_state
*state
)
4683 ERR("Compute pipeline not supported by the ARB shader backend.\n");
4686 /* Context activation is done by the caller. */
4687 static void shader_arb_disable(void *shader_priv
, struct wined3d_context
*context
)
4689 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
4690 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
4691 struct shader_arb_priv
*priv
= shader_priv
;
4693 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
4695 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4696 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4697 priv
->current_fprogram_id
= 0;
4699 priv
->fragment_pipe
->fp_enable(context
, FALSE
);
4701 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
4703 priv
->current_vprogram_id
= 0;
4704 gl_info
->gl_ops
.gl
.p_glDisable(GL_VERTEX_PROGRAM_ARB
);
4705 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4707 priv
->vertex_pipe
->vp_enable(context
, FALSE
);
4709 if (gl_info
->supported
[ARB_COLOR_BUFFER_FLOAT
] && priv
->last_vs_color_unclamp
)
4711 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, GL_FIXED_ONLY_ARB
));
4712 checkGLcall("glClampColorARB");
4713 priv
->last_vs_color_unclamp
= FALSE
;
4716 context
->shader_update_mask
= (1u << WINED3D_SHADER_TYPE_PIXEL
)
4717 | (1u << WINED3D_SHADER_TYPE_VERTEX
)
4718 | (1u << WINED3D_SHADER_TYPE_GEOMETRY
)
4719 | (1u << WINED3D_SHADER_TYPE_HULL
)
4720 | (1u << WINED3D_SHADER_TYPE_DOMAIN
)
4721 | (1u << WINED3D_SHADER_TYPE_COMPUTE
);
4724 static void shader_arb_destroy(struct wined3d_shader
*shader
)
4726 struct wined3d_device
*device
= shader
->device
;
4727 const struct wined3d_gl_info
*gl_info
;
4728 struct wined3d_context
*context
;
4731 /* This can happen if a shader was never compiled */
4732 if (!shader
->backend_data
)
4735 context
= context_acquire(device
, NULL
, 0);
4736 gl_info
= wined3d_context_gl(context
)->gl_info
;
4738 if (shader_is_pshader_version(shader
->reg_maps
.shader_version
.type
))
4740 struct arb_pshader_private
*shader_data
= shader
->backend_data
;
4742 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4743 GL_EXTCALL(glDeleteProgramsARB(1, &shader_data
->gl_shaders
[i
].prgId
));
4745 heap_free(shader_data
->gl_shaders
);
4749 struct arb_vshader_private
*shader_data
= shader
->backend_data
;
4751 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4752 GL_EXTCALL(glDeleteProgramsARB(1, &shader_data
->gl_shaders
[i
].prgId
));
4754 heap_free(shader_data
->gl_shaders
);
4757 checkGLcall("delete programs");
4759 context_release(context
);
4761 heap_free(shader
->backend_data
);
4762 shader
->backend_data
= NULL
;
4765 static int sig_tree_compare(const void *key
, const struct wine_rb_entry
*entry
)
4767 struct ps_signature
*e
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
4768 return compare_sig(key
, &e
->sig
);
4771 static HRESULT
shader_arb_alloc(struct wined3d_device
*device
, const struct wined3d_vertex_pipe_ops
*vertex_pipe
,
4772 const struct wined3d_fragment_pipe_ops
*fragment_pipe
)
4774 const struct wined3d_d3d_info
*d3d_info
= &device
->adapter
->d3d_info
;
4775 struct fragment_caps fragment_caps
;
4776 void *vertex_priv
, *fragment_priv
;
4777 struct shader_arb_priv
*priv
;
4779 if (!(priv
= heap_alloc_zero(sizeof(*priv
))))
4780 return E_OUTOFMEMORY
;
4782 if (!(vertex_priv
= vertex_pipe
->vp_alloc(&arb_program_shader_backend
, priv
)))
4784 ERR("Failed to initialize vertex pipe.\n");
4789 if (!(fragment_priv
= fragment_pipe
->alloc_private(&arb_program_shader_backend
, priv
)))
4791 ERR("Failed to initialize fragment pipe.\n");
4792 vertex_pipe
->vp_free(device
, NULL
);
4797 memset(priv
->vshader_const_dirty
, 1,
4798 sizeof(*priv
->vshader_const_dirty
) * d3d_info
->limits
.vs_uniform_count
);
4799 memset(priv
->pshader_const_dirty
, 1,
4800 sizeof(*priv
->pshader_const_dirty
) * d3d_info
->limits
.ps_uniform_count
);
4802 wine_rb_init(&priv
->signature_tree
, sig_tree_compare
);
4804 priv
->vertex_pipe
= vertex_pipe
;
4805 priv
->fragment_pipe
= fragment_pipe
;
4806 fragment_pipe
->get_caps(device
->adapter
, &fragment_caps
);
4807 priv
->ffp_proj_control
= fragment_caps
.wined3d_caps
& WINED3D_FRAGMENT_CAP_PROJ_CONTROL
;
4809 device
->vertex_priv
= vertex_priv
;
4810 device
->fragment_priv
= fragment_priv
;
4811 device
->shader_priv
= priv
;
4816 static void release_signature(struct wine_rb_entry
*entry
, void *context
)
4818 struct ps_signature
*sig
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
4821 for (i
= 0; i
< sig
->sig
.element_count
; ++i
)
4823 heap_free((char *)sig
->sig
.elements
[i
].semantic_name
);
4825 heap_free(sig
->sig
.elements
);
4829 /* Context activation is done by the caller. */
4830 static void shader_arb_free(struct wined3d_device
*device
, struct wined3d_context
*context
)
4832 struct shader_arb_priv
*priv
= device
->shader_priv
;
4834 wine_rb_destroy(&priv
->signature_tree
, release_signature
, NULL
);
4835 priv
->fragment_pipe
->free_private(device
, context
);
4836 priv
->vertex_pipe
->vp_free(device
, context
);
4837 heap_free(device
->shader_priv
);
4840 static BOOL
shader_arb_allocate_context_data(struct wined3d_context
*context
)
4845 static void shader_arb_free_context_data(struct wined3d_context
*context
)
4847 struct shader_arb_priv
*priv
;
4849 priv
= context
->device
->shader_priv
;
4850 if (priv
->last_context
== context
)
4851 priv
->last_context
= NULL
;
4854 static void shader_arb_init_context_state(struct wined3d_context
*context
) {}
4856 static void shader_arb_get_caps(const struct wined3d_adapter
*adapter
, struct shader_caps
*caps
)
4858 const struct wined3d_gl_info
*gl_info
= &adapter
->gl_info
;
4860 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
4865 /* 96 is the minimum allowed value of MAX_PROGRAM_ENV_PARAMETERS_ARB
4866 * for vertex programs. If the native limit is less than that it's
4867 * not very useful, and e.g. Mesa swrast returns 0, probably to
4868 * indicate it's a software implementation. */
4869 if (gl_info
->limits
.arb_vs_native_constants
< 96)
4870 vs_consts
= gl_info
->limits
.arb_vs_float_constants
;
4872 vs_consts
= min(gl_info
->limits
.arb_vs_float_constants
, gl_info
->limits
.arb_vs_native_constants
);
4874 if (gl_info
->supported
[NV_VERTEX_PROGRAM3
])
4877 TRACE("Hardware vertex shader version 3.0 enabled (NV_VERTEX_PROGRAM3)\n");
4879 else if (vs_consts
>= 256)
4881 /* Shader Model 2.0 requires at least 256 vertex shader constants */
4883 TRACE("Hardware vertex shader version 2.0 enabled (ARB_PROGRAM)\n");
4888 TRACE("Hardware vertex shader version 1.1 enabled (ARB_PROGRAM)\n");
4890 caps
->vs_version
= min(wined3d_settings
.max_sm_vs
, vs_version
);
4891 caps
->vs_uniform_count
= min(WINED3D_MAX_VS_CONSTS_F
, vs_consts
);
4895 caps
->vs_version
= 0;
4896 caps
->vs_uniform_count
= 0;
4899 caps
->hs_version
= 0;
4900 caps
->ds_version
= 0;
4901 caps
->gs_version
= 0;
4902 caps
->cs_version
= 0;
4904 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
4909 /* Similar as above for vertex programs, but the minimum for fragment
4910 * programs is 24. */
4911 if (gl_info
->limits
.arb_ps_native_constants
< 24)
4912 ps_consts
= gl_info
->limits
.arb_ps_float_constants
;
4914 ps_consts
= min(gl_info
->limits
.arb_ps_float_constants
, gl_info
->limits
.arb_ps_native_constants
);
4916 if (gl_info
->supported
[NV_FRAGMENT_PROGRAM2
])
4919 TRACE("Hardware pixel shader version 3.0 enabled (NV_FRAGMENT_PROGRAM2)\n");
4921 else if (ps_consts
>= 32)
4923 /* Shader Model 2.0 requires at least 32 pixel shader constants */
4925 TRACE("Hardware pixel shader version 2.0 enabled (ARB_PROGRAM)\n");
4930 TRACE("Hardware pixel shader version 1.4 enabled (ARB_PROGRAM)\n");
4932 caps
->ps_version
= min(wined3d_settings
.max_sm_ps
, ps_version
);
4933 caps
->ps_uniform_count
= min(WINED3D_MAX_PS_CONSTS_F
, ps_consts
);
4934 caps
->ps_1x_max_value
= 8.0f
;
4938 caps
->ps_version
= 0;
4939 caps
->ps_uniform_count
= 0;
4940 caps
->ps_1x_max_value
= 0.0f
;
4943 caps
->varying_count
= 0;
4944 caps
->wined3d_caps
= WINED3D_SHADER_CAP_SRGB_WRITE
;
4945 if (use_nv_clip(gl_info
))
4946 caps
->wined3d_caps
|= WINED3D_SHADER_CAP_VS_CLIPPING
;
4949 static BOOL
shader_arb_color_fixup_supported(struct color_fixup_desc fixup
)
4951 /* We support everything except complex conversions. */
4952 return !is_complex_fixup(fixup
);
4955 static void shader_arb_add_instruction_modifiers(const struct wined3d_shader_instruction
*ins
) {
4957 char write_mask
[20], regstr
[50];
4958 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
4959 BOOL is_color
= FALSE
;
4960 const struct wined3d_shader_dst_param
*dst
;
4962 if (!ins
->dst_count
) return;
4966 if (!shift
) return; /* Saturate alone is handled by the instructions */
4968 shader_arb_get_write_mask(ins
, dst
, write_mask
);
4969 shader_arb_get_register_name(ins
, &dst
->reg
, regstr
, &is_color
);
4971 /* Generate a line that does the output modifier computation
4972 * FIXME: _SAT vs shift? _SAT alone is already handled in the instructions, if this
4973 * maps problems in e.g. _d4_sat modify shader_arb_get_modifier
4975 shader_addline(buffer
, "MUL%s %s%s, %s, %s;\n", shader_arb_get_modifier(ins
),
4976 regstr
, write_mask
, regstr
, shift_tab
[shift
]);
4979 static const SHADER_HANDLER shader_arb_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
4981 /* WINED3DSIH_ABS */ shader_hw_map2gl
,
4982 /* WINED3DSIH_ADD */ shader_hw_map2gl
,
4983 /* WINED3DSIH_AND */ NULL
,
4984 /* WINED3DSIH_ATOMIC_AND */ NULL
,
4985 /* WINED3DSIH_ATOMIC_CMP_STORE */ NULL
,
4986 /* WINED3DSIH_ATOMIC_IADD */ NULL
,
4987 /* WINED3DSIH_ATOMIC_IMAX */ NULL
,
4988 /* WINED3DSIH_ATOMIC_IMIN */ NULL
,
4989 /* WINED3DSIH_ATOMIC_OR */ NULL
,
4990 /* WINED3DSIH_ATOMIC_UMAX */ NULL
,
4991 /* WINED3DSIH_ATOMIC_UMIN */ NULL
,
4992 /* WINED3DSIH_ATOMIC_XOR */ NULL
,
4993 /* WINED3DSIH_BEM */ pshader_hw_bem
,
4994 /* WINED3DSIH_BFI */ NULL
,
4995 /* WINED3DSIH_BFREV */ NULL
,
4996 /* WINED3DSIH_BREAK */ shader_hw_break
,
4997 /* WINED3DSIH_BREAKC */ shader_hw_breakc
,
4998 /* WINED3DSIH_BREAKP */ NULL
,
4999 /* WINED3DSIH_BUFINFO */ NULL
,
5000 /* WINED3DSIH_CALL */ shader_hw_call
,
5001 /* WINED3DSIH_CALLNZ */ NULL
,
5002 /* WINED3DSIH_CASE */ NULL
,
5003 /* WINED3DSIH_CMP */ pshader_hw_cmp
,
5004 /* WINED3DSIH_CND */ pshader_hw_cnd
,
5005 /* WINED3DSIH_CONTINUE */ NULL
,
5006 /* WINED3DSIH_CONTINUEP */ NULL
,
5007 /* WINED3DSIH_COUNTBITS */ NULL
,
5008 /* WINED3DSIH_CRS */ shader_hw_map2gl
,
5009 /* WINED3DSIH_CUT */ NULL
,
5010 /* WINED3DSIH_CUT_STREAM */ NULL
,
5011 /* WINED3DSIH_DCL */ shader_hw_nop
,
5012 /* WINED3DSIH_DCL_CONSTANT_BUFFER */ shader_hw_nop
,
5013 /* WINED3DSIH_DCL_FUNCTION_BODY */ NULL
,
5014 /* WINED3DSIH_DCL_FUNCTION_TABLE */ NULL
,
5015 /* WINED3DSIH_DCL_GLOBAL_FLAGS */ NULL
,
5016 /* WINED3DSIH_DCL_GS_INSTANCES */ NULL
,
5017 /* WINED3DSIH_DCL_HS_FORK_PHASE_INSTANCE_COUNT */ NULL
,
5018 /* WINED3DSIH_DCL_HS_JOIN_PHASE_INSTANCE_COUNT */ NULL
,
5019 /* WINED3DSIH_DCL_HS_MAX_TESSFACTOR */ NULL
,
5020 /* WINED3DSIH_DCL_IMMEDIATE_CONSTANT_BUFFER */ NULL
,
5021 /* WINED3DSIH_DCL_INDEX_RANGE */ NULL
,
5022 /* WINED3DSIH_DCL_INDEXABLE_TEMP */ NULL
,
5023 /* WINED3DSIH_DCL_INPUT */ NULL
,
5024 /* WINED3DSIH_DCL_INPUT_CONTROL_POINT_COUNT */ NULL
,
5025 /* WINED3DSIH_DCL_INPUT_PRIMITIVE */ shader_hw_nop
,
5026 /* WINED3DSIH_DCL_INPUT_PS */ NULL
,
5027 /* WINED3DSIH_DCL_INPUT_PS_SGV */ NULL
,
5028 /* WINED3DSIH_DCL_INPUT_PS_SIV */ NULL
,
5029 /* WINED3DSIH_DCL_INPUT_SGV */ NULL
,
5030 /* WINED3DSIH_DCL_INPUT_SIV */ NULL
,
5031 /* WINED3DSIH_DCL_INTERFACE */ NULL
,
5032 /* WINED3DSIH_DCL_OUTPUT */ NULL
,
5033 /* WINED3DSIH_DCL_OUTPUT_CONTROL_POINT_COUNT */ NULL
,
5034 /* WINED3DSIH_DCL_OUTPUT_SIV */ NULL
,
5035 /* WINED3DSIH_DCL_OUTPUT_TOPOLOGY */ shader_hw_nop
,
5036 /* WINED3DSIH_DCL_RESOURCE_RAW */ NULL
,
5037 /* WINED3DSIH_DCL_RESOURCE_STRUCTURED */ NULL
,
5038 /* WINED3DSIH_DCL_SAMPLER */ NULL
,
5039 /* WINED3DSIH_DCL_STREAM */ NULL
,
5040 /* WINED3DSIH_DCL_TEMPS */ NULL
,
5041 /* WINED3DSIH_DCL_TESSELLATOR_DOMAIN */ NULL
,
5042 /* WINED3DSIH_DCL_TESSELLATOR_OUTPUT_PRIMITIVE */ NULL
,
5043 /* WINED3DSIH_DCL_TESSELLATOR_PARTITIONING */ NULL
,
5044 /* WINED3DSIH_DCL_TGSM_RAW */ NULL
,
5045 /* WINED3DSIH_DCL_TGSM_STRUCTURED */ NULL
,
5046 /* WINED3DSIH_DCL_THREAD_GROUP */ NULL
,
5047 /* WINED3DSIH_DCL_UAV_RAW */ NULL
,
5048 /* WINED3DSIH_DCL_UAV_STRUCTURED */ NULL
,
5049 /* WINED3DSIH_DCL_UAV_TYPED */ NULL
,
5050 /* WINED3DSIH_DCL_VERTICES_OUT */ shader_hw_nop
,
5051 /* WINED3DSIH_DEF */ shader_hw_nop
,
5052 /* WINED3DSIH_DEFAULT */ NULL
,
5053 /* WINED3DSIH_DEFB */ shader_hw_nop
,
5054 /* WINED3DSIH_DEFI */ shader_hw_nop
,
5055 /* WINED3DSIH_DIV */ NULL
,
5056 /* WINED3DSIH_DP2 */ NULL
,
5057 /* WINED3DSIH_DP2ADD */ pshader_hw_dp2add
,
5058 /* WINED3DSIH_DP3 */ shader_hw_map2gl
,
5059 /* WINED3DSIH_DP4 */ shader_hw_map2gl
,
5060 /* WINED3DSIH_DST */ shader_hw_map2gl
,
5061 /* WINED3DSIH_DSX */ shader_hw_map2gl
,
5062 /* WINED3DSIH_DSX_COARSE */ NULL
,
5063 /* WINED3DSIH_DSX_FINE */ NULL
,
5064 /* WINED3DSIH_DSY */ shader_hw_dsy
,
5065 /* WINED3DSIH_DSY_COARSE */ NULL
,
5066 /* WINED3DSIH_DSY_FINE */ NULL
,
5067 /* WINED3DSIH_ELSE */ shader_hw_else
,
5068 /* WINED3DSIH_EMIT */ NULL
,
5069 /* WINED3DSIH_EMIT_STREAM */ NULL
,
5070 /* WINED3DSIH_ENDIF */ shader_hw_endif
,
5071 /* WINED3DSIH_ENDLOOP */ shader_hw_endloop
,
5072 /* WINED3DSIH_ENDREP */ shader_hw_endrep
,
5073 /* WINED3DSIH_ENDSWITCH */ NULL
,
5074 /* WINED3DSIH_EQ */ NULL
,
5075 /* WINED3DSIH_EVAL_CENTROID */ NULL
,
5076 /* WINED3DSIH_EVAL_SAMPLE_INDEX */ NULL
,
5077 /* WINED3DSIH_EXP */ shader_hw_scalar_op
,
5078 /* WINED3DSIH_EXPP */ shader_hw_scalar_op
,
5079 /* WINED3DSIH_F16TOF32 */ NULL
,
5080 /* WINED3DSIH_F32TOF16 */ NULL
,
5081 /* WINED3DSIH_FCALL */ NULL
,
5082 /* WINED3DSIH_FIRSTBIT_HI */ NULL
,
5083 /* WINED3DSIH_FIRSTBIT_LO */ NULL
,
5084 /* WINED3DSIH_FIRSTBIT_SHI */ NULL
,
5085 /* WINED3DSIH_FRC */ shader_hw_map2gl
,
5086 /* WINED3DSIH_FTOI */ NULL
,
5087 /* WINED3DSIH_FTOU */ NULL
,
5088 /* WINED3DSIH_GATHER4 */ NULL
,
5089 /* WINED3DSIH_GATHER4_C */ NULL
,
5090 /* WINED3DSIH_GATHER4_PO */ NULL
,
5091 /* WINED3DSIH_GATHER4_PO_C */ NULL
,
5092 /* WINED3DSIH_GE */ NULL
,
5093 /* WINED3DSIH_HS_CONTROL_POINT_PHASE */ NULL
,
5094 /* WINED3DSIH_HS_DECLS */ NULL
,
5095 /* WINED3DSIH_HS_FORK_PHASE */ NULL
,
5096 /* WINED3DSIH_HS_JOIN_PHASE */ NULL
,
5097 /* WINED3DSIH_IADD */ NULL
,
5098 /* WINED3DSIH_IBFE */ NULL
,
5099 /* WINED3DSIH_IEQ */ NULL
,
5100 /* WINED3DSIH_IF */ NULL
/* Hardcoded into the shader */,
5101 /* WINED3DSIH_IFC */ shader_hw_ifc
,
5102 /* WINED3DSIH_IGE */ NULL
,
5103 /* WINED3DSIH_ILT */ NULL
,
5104 /* WINED3DSIH_IMAD */ NULL
,
5105 /* WINED3DSIH_IMAX */ NULL
,
5106 /* WINED3DSIH_IMIN */ NULL
,
5107 /* WINED3DSIH_IMM_ATOMIC_ALLOC */ NULL
,
5108 /* WINED3DSIH_IMM_ATOMIC_AND */ NULL
,
5109 /* WINED3DSIH_IMM_ATOMIC_CMP_EXCH */ NULL
,
5110 /* WINED3DSIH_IMM_ATOMIC_CONSUME */ NULL
,
5111 /* WINED3DSIH_IMM_ATOMIC_EXCH */ NULL
,
5112 /* WINED3DSIH_IMM_ATOMIC_IADD */ NULL
,
5113 /* WINED3DSIH_IMM_ATOMIC_IMAX */ NULL
,
5114 /* WINED3DSIH_IMM_ATOMIC_IMIN */ NULL
,
5115 /* WINED3DSIH_IMM_ATOMIC_OR */ NULL
,
5116 /* WINED3DSIH_IMM_ATOMIC_UMAX */ NULL
,
5117 /* WINED3DSIH_IMM_ATOMIC_UMIN */ NULL
,
5118 /* WINED3DSIH_IMM_ATOMIC_XOR */ NULL
,
5119 /* WINED3DSIH_IMUL */ NULL
,
5120 /* WINED3DSIH_INE */ NULL
,
5121 /* WINED3DSIH_INEG */ NULL
,
5122 /* WINED3DSIH_ISHL */ NULL
,
5123 /* WINED3DSIH_ISHR */ NULL
,
5124 /* WINED3DSIH_ITOF */ NULL
,
5125 /* WINED3DSIH_LABEL */ shader_hw_label
,
5126 /* WINED3DSIH_LD */ NULL
,
5127 /* WINED3DSIH_LD2DMS */ NULL
,
5128 /* WINED3DSIH_LD_RAW */ NULL
,
5129 /* WINED3DSIH_LD_STRUCTURED */ NULL
,
5130 /* WINED3DSIH_LD_UAV_TYPED */ NULL
,
5131 /* WINED3DSIH_LIT */ shader_hw_map2gl
,
5132 /* WINED3DSIH_LOD */ NULL
,
5133 /* WINED3DSIH_LOG */ shader_hw_scalar_op
,
5134 /* WINED3DSIH_LOGP */ shader_hw_scalar_op
,
5135 /* WINED3DSIH_LOOP */ shader_hw_loop
,
5136 /* WINED3DSIH_LRP */ shader_hw_lrp
,
5137 /* WINED3DSIH_LT */ NULL
,
5138 /* WINED3DSIH_M3x2 */ shader_hw_mnxn
,
5139 /* WINED3DSIH_M3x3 */ shader_hw_mnxn
,
5140 /* WINED3DSIH_M3x4 */ shader_hw_mnxn
,
5141 /* WINED3DSIH_M4x3 */ shader_hw_mnxn
,
5142 /* WINED3DSIH_M4x4 */ shader_hw_mnxn
,
5143 /* WINED3DSIH_MAD */ shader_hw_map2gl
,
5144 /* WINED3DSIH_MAX */ shader_hw_map2gl
,
5145 /* WINED3DSIH_MIN */ shader_hw_map2gl
,
5146 /* WINED3DSIH_MOV */ shader_hw_mov
,
5147 /* WINED3DSIH_MOVA */ shader_hw_mov
,
5148 /* WINED3DSIH_MOVC */ NULL
,
5149 /* WINED3DSIH_MUL */ shader_hw_map2gl
,
5150 /* WINED3DSIH_NE */ NULL
,
5151 /* WINED3DSIH_NOP */ shader_hw_nop
,
5152 /* WINED3DSIH_NOT */ NULL
,
5153 /* WINED3DSIH_NRM */ shader_hw_nrm
,
5154 /* WINED3DSIH_OR */ NULL
,
5155 /* WINED3DSIH_PHASE */ shader_hw_nop
,
5156 /* WINED3DSIH_POW */ shader_hw_pow
,
5157 /* WINED3DSIH_RCP */ shader_hw_scalar_op
,
5158 /* WINED3DSIH_REP */ shader_hw_rep
,
5159 /* WINED3DSIH_RESINFO */ NULL
,
5160 /* WINED3DSIH_RET */ shader_hw_ret
,
5161 /* WINED3DSIH_RETP */ NULL
,
5162 /* WINED3DSIH_ROUND_NE */ NULL
,
5163 /* WINED3DSIH_ROUND_NI */ NULL
,
5164 /* WINED3DSIH_ROUND_PI */ NULL
,
5165 /* WINED3DSIH_ROUND_Z */ NULL
,
5166 /* WINED3DSIH_RSQ */ shader_hw_scalar_op
,
5167 /* WINED3DSIH_SAMPLE */ NULL
,
5168 /* WINED3DSIH_SAMPLE_B */ NULL
,
5169 /* WINED3DSIH_SAMPLE_C */ NULL
,
5170 /* WINED3DSIH_SAMPLE_C_LZ */ NULL
,
5171 /* WINED3DSIH_SAMPLE_GRAD */ NULL
,
5172 /* WINED3DSIH_SAMPLE_INFO */ NULL
,
5173 /* WINED3DSIH_SAMPLE_LOD */ NULL
,
5174 /* WINED3DSIH_SAMPLE_POS */ NULL
,
5175 /* WINED3DSIH_SETP */ NULL
,
5176 /* WINED3DSIH_SGE */ shader_hw_map2gl
,
5177 /* WINED3DSIH_SGN */ shader_hw_sgn
,
5178 /* WINED3DSIH_SINCOS */ shader_hw_sincos
,
5179 /* WINED3DSIH_SLT */ shader_hw_map2gl
,
5180 /* WINED3DSIH_SQRT */ NULL
,
5181 /* WINED3DSIH_STORE_RAW */ NULL
,
5182 /* WINED3DSIH_STORE_STRUCTURED */ NULL
,
5183 /* WINED3DSIH_STORE_UAV_TYPED */ NULL
,
5184 /* WINED3DSIH_SUB */ shader_hw_map2gl
,
5185 /* WINED3DSIH_SWAPC */ NULL
,
5186 /* WINED3DSIH_SWITCH */ NULL
,
5187 /* WINED3DSIH_SYNC */ NULL
,
5188 /* WINED3DSIH_TEX */ pshader_hw_tex
,
5189 /* WINED3DSIH_TEXBEM */ pshader_hw_texbem
,
5190 /* WINED3DSIH_TEXBEML */ pshader_hw_texbem
,
5191 /* WINED3DSIH_TEXCOORD */ pshader_hw_texcoord
,
5192 /* WINED3DSIH_TEXDEPTH */ pshader_hw_texdepth
,
5193 /* WINED3DSIH_TEXDP3 */ pshader_hw_texdp3
,
5194 /* WINED3DSIH_TEXDP3TEX */ pshader_hw_texdp3tex
,
5195 /* WINED3DSIH_TEXKILL */ pshader_hw_texkill
,
5196 /* WINED3DSIH_TEXLDD */ shader_hw_texldd
,
5197 /* WINED3DSIH_TEXLDL */ shader_hw_texldl
,
5198 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_hw_texm3x2depth
,
5199 /* WINED3DSIH_TEXM3x2PAD */ pshader_hw_texm3x2pad
,
5200 /* WINED3DSIH_TEXM3x2TEX */ pshader_hw_texm3x2tex
,
5201 /* WINED3DSIH_TEXM3x3 */ pshader_hw_texm3x3
,
5202 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
5203 /* WINED3DSIH_TEXM3x3PAD */ pshader_hw_texm3x3pad
,
5204 /* WINED3DSIH_TEXM3x3SPEC */ pshader_hw_texm3x3spec
,
5205 /* WINED3DSIH_TEXM3x3TEX */ pshader_hw_texm3x3tex
,
5206 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_hw_texm3x3vspec
,
5207 /* WINED3DSIH_TEXREG2AR */ pshader_hw_texreg2ar
,
5208 /* WINED3DSIH_TEXREG2GB */ pshader_hw_texreg2gb
,
5209 /* WINED3DSIH_TEXREG2RGB */ pshader_hw_texreg2rgb
,
5210 /* WINED3DSIH_UBFE */ NULL
,
5211 /* WINED3DSIH_UDIV */ NULL
,
5212 /* WINED3DSIH_UGE */ NULL
,
5213 /* WINED3DSIH_ULT */ NULL
,
5214 /* WINED3DSIH_UMAX */ NULL
,
5215 /* WINED3DSIH_UMIN */ NULL
,
5216 /* WINED3DSIH_UMUL */ NULL
,
5217 /* WINED3DSIH_USHR */ NULL
,
5218 /* WINED3DSIH_UTOF */ NULL
,
5219 /* WINED3DSIH_XOR */ NULL
,
5222 static BOOL
get_bool_const(const struct wined3d_shader_instruction
*ins
,
5223 const struct wined3d_shader
*shader
, DWORD idx
)
5225 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
5226 BOOL vshader
= shader_is_vshader_version(reg_maps
->shader_version
.type
);
5227 const struct wined3d_shader_lconst
*constant
;
5229 WORD flag
= (1u << idx
);
5230 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5232 if (reg_maps
->local_bool_consts
& flag
)
5234 /* What good is an if(bool) with a hardcoded local constant? I don't know, but handle it */
5235 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsB
, struct wined3d_shader_lconst
, entry
)
5237 if (constant
->idx
== idx
)
5239 return constant
->value
[0];
5242 ERR("Local constant not found\n");
5247 if(vshader
) bools
= priv
->cur_vs_args
->clip
.boolclip
.bools
;
5248 else bools
= priv
->cur_ps_args
->bools
;
5249 return bools
& flag
;
5253 static void get_loop_control_const(const struct wined3d_shader_instruction
*ins
,
5254 const struct wined3d_shader
*shader
, UINT idx
, struct wined3d_shader_loop_control
*loop_control
)
5256 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
5257 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5259 /* Integer constants can either be a local constant, or they can be stored in the shader
5260 * type specific compile args. */
5261 if (reg_maps
->local_int_consts
& (1u << idx
))
5263 const struct wined3d_shader_lconst
*constant
;
5265 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsI
, struct wined3d_shader_lconst
, entry
)
5267 if (constant
->idx
== idx
)
5269 loop_control
->count
= constant
->value
[0];
5270 loop_control
->start
= constant
->value
[1];
5271 /* Step is signed. */
5272 loop_control
->step
= (int)constant
->value
[2];
5276 /* If this happens the flag was set incorrectly */
5277 ERR("Local constant not found\n");
5278 loop_control
->count
= 0;
5279 loop_control
->start
= 0;
5280 loop_control
->step
= 0;
5284 switch (reg_maps
->shader_version
.type
)
5286 case WINED3D_SHADER_TYPE_VERTEX
:
5287 /* Count and aL start value are unsigned */
5288 loop_control
->count
= priv
->cur_vs_args
->loop_ctrl
[idx
][0];
5289 loop_control
->start
= priv
->cur_vs_args
->loop_ctrl
[idx
][1];
5290 /* Step is signed. */
5291 loop_control
->step
= ((char)priv
->cur_vs_args
->loop_ctrl
[idx
][2]);
5294 case WINED3D_SHADER_TYPE_PIXEL
:
5295 loop_control
->count
= priv
->cur_ps_args
->loop_ctrl
[idx
][0];
5296 loop_control
->start
= priv
->cur_ps_args
->loop_ctrl
[idx
][1];
5297 loop_control
->step
= ((char)priv
->cur_ps_args
->loop_ctrl
[idx
][2]);
5301 FIXME("Unhandled shader type %#x.\n", reg_maps
->shader_version
.type
);
5306 static void record_instruction(struct list
*list
, const struct wined3d_shader_instruction
*ins
)
5308 struct wined3d_shader_src_param
*src_param
= NULL
, *rel_addr
;
5309 struct wined3d_shader_dst_param
*dst_param
;
5310 struct recorded_instruction
*rec
;
5313 if (!(rec
= heap_alloc_zero(sizeof(*rec
))))
5315 ERR("Out of memory\n");
5320 if (!(dst_param
= heap_alloc(sizeof(*dst_param
))))
5322 *dst_param
= *ins
->dst
;
5323 if (ins
->dst
->reg
.idx
[0].rel_addr
)
5325 if (!(rel_addr
= heap_alloc(sizeof(*rel_addr
))))
5327 *rel_addr
= *ins
->dst
->reg
.idx
[0].rel_addr
;
5328 dst_param
->reg
.idx
[0].rel_addr
= rel_addr
;
5330 rec
->ins
.dst
= dst_param
;
5332 if (!(src_param
= heap_calloc(ins
->src_count
, sizeof(*src_param
))))
5334 for (i
= 0; i
< ins
->src_count
; ++i
)
5336 src_param
[i
] = ins
->src
[i
];
5337 if (ins
->src
[i
].reg
.idx
[0].rel_addr
)
5339 if (!(rel_addr
= heap_alloc(sizeof(*rel_addr
))))
5341 *rel_addr
= *ins
->src
[i
].reg
.idx
[0].rel_addr
;
5342 src_param
[i
].reg
.idx
[0].rel_addr
= rel_addr
;
5345 rec
->ins
.src
= src_param
;
5346 list_add_tail(list
, &rec
->entry
);
5350 ERR("Out of memory\n");
5353 heap_free((void *)dst_param
->reg
.idx
[0].rel_addr
);
5354 heap_free(dst_param
);
5358 for (i
= 0; i
< ins
->src_count
; ++i
)
5360 heap_free((void *)src_param
[i
].reg
.idx
[0].rel_addr
);
5362 heap_free(src_param
);
5367 static void free_recorded_instruction(struct list
*list
)
5369 struct recorded_instruction
*rec_ins
, *entry2
;
5372 LIST_FOR_EACH_ENTRY_SAFE(rec_ins
, entry2
, list
, struct recorded_instruction
, entry
)
5374 list_remove(&rec_ins
->entry
);
5375 if (rec_ins
->ins
.dst
)
5377 heap_free((void *)rec_ins
->ins
.dst
->reg
.idx
[0].rel_addr
);
5378 heap_free((void *)rec_ins
->ins
.dst
);
5380 if (rec_ins
->ins
.src
)
5382 for (i
= 0; i
< rec_ins
->ins
.src_count
; ++i
)
5384 heap_free((void *)rec_ins
->ins
.src
[i
].reg
.idx
[0].rel_addr
);
5386 heap_free((void *)rec_ins
->ins
.src
);
5392 static void pop_control_frame(const struct wined3d_shader_instruction
*ins
)
5394 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5395 struct control_frame
*control_frame
;
5397 if (ins
->handler_idx
== WINED3DSIH_ENDLOOP
|| ins
->handler_idx
== WINED3DSIH_ENDREP
)
5399 struct list
*e
= list_head(&priv
->control_frames
);
5400 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5401 list_remove(&control_frame
->entry
);
5402 heap_free(control_frame
);
5405 else if (ins
->handler_idx
== WINED3DSIH_ENDIF
)
5407 /* Non-ifc ENDIFs were already handled previously. */
5408 struct list
*e
= list_head(&priv
->control_frames
);
5409 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5410 list_remove(&control_frame
->entry
);
5411 heap_free(control_frame
);
5415 static void shader_arb_handle_instruction(const struct wined3d_shader_instruction
*ins
) {
5416 SHADER_HANDLER hw_fct
;
5417 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5418 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
5419 struct control_frame
*control_frame
;
5420 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
5423 if(ins
->handler_idx
== WINED3DSIH_LOOP
|| ins
->handler_idx
== WINED3DSIH_REP
)
5425 control_frame
= heap_alloc_zero(sizeof(*control_frame
));
5426 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5428 if(ins
->handler_idx
== WINED3DSIH_LOOP
) control_frame
->type
= LOOP
;
5429 if(ins
->handler_idx
== WINED3DSIH_REP
) control_frame
->type
= REP
;
5431 if(priv
->target_version
>= NV2
)
5433 control_frame
->no
.loop
= priv
->num_loops
++;
5438 /* Don't bother recording when we're in a not used if branch */
5444 if(!priv
->recording
)
5446 list_init(&priv
->record
);
5447 priv
->recording
= TRUE
;
5448 control_frame
->outer_loop
= TRUE
;
5449 get_loop_control_const(ins
, shader
, ins
->src
[0].reg
.idx
[0].offset
, &control_frame
->loop_control
);
5450 return; /* Instruction is handled */
5452 /* Record this loop in the outer loop's recording */
5455 else if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
|| ins
->handler_idx
== WINED3DSIH_ENDREP
)
5457 if(priv
->target_version
>= NV2
)
5459 /* Nothing to do. The control frame is popped after the HW instr handler */
5463 struct list
*e
= list_head(&priv
->control_frames
);
5464 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5465 list_remove(&control_frame
->entry
);
5467 if(control_frame
->outer_loop
)
5469 unsigned int iteration
;
5473 /* Turn off recording before playback */
5474 priv
->recording
= FALSE
;
5476 /* Move the recorded instructions to a separate list and get them out of the private data
5477 * structure. If there are nested loops, the shader_arb_handle_instruction below will
5478 * be recorded again, thus priv->record might be overwritten
5481 list_move_tail(©
, &priv
->record
);
5482 list_init(&priv
->record
);
5484 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5486 shader_addline(buffer
, "#unrolling loop: %u iterations, aL=%u, inc %d\n",
5487 control_frame
->loop_control
.count
, control_frame
->loop_control
.start
,
5488 control_frame
->loop_control
.step
);
5489 aL
= control_frame
->loop_control
.start
;
5493 shader_addline(buffer
, "#unrolling rep: %u iterations\n", control_frame
->loop_control
.count
);
5496 for (iteration
= 0; iteration
< control_frame
->loop_control
.count
; ++iteration
)
5498 struct recorded_instruction
*rec_ins
;
5499 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5502 shader_addline(buffer
, "#Iteration %u, aL=%d\n", iteration
, aL
);
5506 shader_addline(buffer
, "#Iteration %u\n", iteration
);
5509 LIST_FOR_EACH_ENTRY(rec_ins
, ©
, struct recorded_instruction
, entry
)
5511 shader_arb_handle_instruction(&rec_ins
->ins
);
5514 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5516 aL
+= control_frame
->loop_control
.step
;
5519 shader_addline(buffer
, "#end loop/rep\n");
5521 free_recorded_instruction(©
);
5522 heap_free(control_frame
);
5523 return; /* Instruction is handled */
5527 /* This is a nested loop. Proceed to the normal recording function */
5528 heap_free(control_frame
);
5535 record_instruction(&priv
->record
, ins
);
5540 if(ins
->handler_idx
== WINED3DSIH_IF
)
5542 control_frame
= heap_alloc_zero(sizeof(*control_frame
));
5543 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5544 control_frame
->type
= IF
;
5546 bool_const
= get_bool_const(ins
, shader
, ins
->src
[0].reg
.idx
[0].offset
);
5547 if (ins
->src
[0].modifiers
== WINED3DSPSM_NOT
)
5548 bool_const
= !bool_const
;
5549 if (!priv
->muted
&& !bool_const
)
5551 shader_addline(buffer
, "#if(FALSE){\n");
5553 control_frame
->muting
= TRUE
;
5555 else shader_addline(buffer
, "#if(TRUE) {\n");
5557 return; /* Instruction is handled */
5559 else if(ins
->handler_idx
== WINED3DSIH_IFC
)
5561 /* IF(bool) and if_cond(a, b) use the same ELSE and ENDIF tokens */
5562 control_frame
= heap_alloc_zero(sizeof(*control_frame
));
5563 control_frame
->type
= IFC
;
5564 control_frame
->no
.ifc
= priv
->num_ifcs
++;
5565 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5567 else if(ins
->handler_idx
== WINED3DSIH_ELSE
)
5569 struct list
*e
= list_head(&priv
->control_frames
);
5570 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5572 if(control_frame
->type
== IF
)
5574 shader_addline(buffer
, "#} else {\n");
5575 if(!priv
->muted
&& !control_frame
->muting
)
5578 control_frame
->muting
= TRUE
;
5580 else if(control_frame
->muting
) priv
->muted
= FALSE
;
5581 return; /* Instruction is handled. */
5583 /* In case of an ifc, generate a HW shader instruction */
5584 if (control_frame
->type
!= IFC
)
5585 ERR("Control frame does not match.\n");
5587 else if(ins
->handler_idx
== WINED3DSIH_ENDIF
)
5589 struct list
*e
= list_head(&priv
->control_frames
);
5590 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5592 if(control_frame
->type
== IF
)
5594 shader_addline(buffer
, "#} endif\n");
5595 if(control_frame
->muting
) priv
->muted
= FALSE
;
5596 list_remove(&control_frame
->entry
);
5597 heap_free(control_frame
);
5598 return; /* Instruction is handled */
5600 /* In case of an ifc, generate a HW shader instruction */
5601 if (control_frame
->type
!= IFC
)
5602 ERR("Control frame does not match.\n");
5607 pop_control_frame(ins
);
5611 /* Select handler */
5612 hw_fct
= shader_arb_instruction_handler_table
[ins
->handler_idx
];
5614 /* Unhandled opcode */
5617 FIXME("Backend can't handle opcode %s.\n", debug_d3dshaderinstructionhandler(ins
->handler_idx
));
5622 pop_control_frame(ins
);
5624 shader_arb_add_instruction_modifiers(ins
);
5627 static BOOL
shader_arb_has_ffp_proj_control(void *shader_priv
)
5629 struct shader_arb_priv
*priv
= shader_priv
;
5631 return priv
->ffp_proj_control
;
5634 static void shader_arb_precompile(void *shader_priv
, struct wined3d_shader
*shader
) {}
5636 static uint64_t shader_arb_shader_compile(struct wined3d_context
*context
, const struct wined3d_shader_desc
*shader_desc
,
5637 enum wined3d_shader_type shader_type
)
5639 ERR("Not implemented.\n");
5643 const struct wined3d_shader_backend_ops arb_program_shader_backend
=
5645 shader_arb_handle_instruction
,
5646 shader_arb_precompile
,
5648 shader_arb_select_compute
,
5650 shader_arb_update_float_vertex_constants
,
5651 shader_arb_update_float_pixel_constants
,
5652 shader_arb_load_constants
,
5656 shader_arb_allocate_context_data
,
5657 shader_arb_free_context_data
,
5658 shader_arb_init_context_state
,
5659 shader_arb_get_caps
,
5660 shader_arb_color_fixup_supported
,
5661 shader_arb_has_ffp_proj_control
,
5662 shader_arb_shader_compile
,
5665 /* ARB_fragment_program fixed function pipeline replacement definitions */
5666 #define ARB_FFP_CONST_TFACTOR 0
5667 #define ARB_FFP_CONST_COLOR_KEY_LOW ((ARB_FFP_CONST_TFACTOR) + 1)
5668 #define ARB_FFP_CONST_COLOR_KEY_HIGH ((ARB_FFP_CONST_COLOR_KEY_LOW) + 1)
5669 #define ARB_FFP_CONST_SPECULAR_ENABLE ((ARB_FFP_CONST_COLOR_KEY_HIGH) + 1)
5670 #define ARB_FFP_CONST_CONSTANT(i) ((ARB_FFP_CONST_SPECULAR_ENABLE) + 1 + i)
5671 #define ARB_FFP_CONST_BUMPMAT(i) ((ARB_FFP_CONST_CONSTANT(7)) + 1 + i)
5672 #define ARB_FFP_CONST_LUMINANCE(i) ((ARB_FFP_CONST_BUMPMAT(7)) + 1 + i)
5674 struct arbfp_ffp_desc
5676 struct ffp_frag_desc parent
;
5680 /* Context activation is done by the caller. */
5681 static void arbfp_enable(const struct wined3d_context
*context
, BOOL enable
)
5683 const struct wined3d_gl_info
*gl_info
= wined3d_context_gl_const(context
)->gl_info
;
5687 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
5688 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
5692 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
5693 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
5697 static void *arbfp_alloc(const struct wined3d_shader_backend_ops
*shader_backend
, void *shader_priv
)
5699 struct shader_arb_priv
*priv
;
5701 /* Share private data between the shader backend and the pipeline
5702 * replacement, if both are the arb implementation. This is needed to
5703 * figure out whether ARBfp should be disabled if no pixel shader is bound
5705 if (shader_backend
== &arb_program_shader_backend
)
5707 else if (!(priv
= heap_alloc_zero(sizeof(*priv
))))
5710 wine_rb_init(&priv
->fragment_shaders
, wined3d_ffp_frag_program_key_compare
);
5711 priv
->use_arbfp_fixed_func
= TRUE
;
5716 /* Context activation is done by the caller. */
5717 static void arbfp_free_ffpshader(struct wine_rb_entry
*entry
, void *param
)
5719 struct arbfp_ffp_desc
*entry_arb
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_ffp_desc
, parent
.entry
);
5720 struct wined3d_context_gl
*context_gl
= param
;
5721 const struct wined3d_gl_info
*gl_info
;
5723 gl_info
= context_gl
->gl_info
;
5724 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb
->shader
));
5725 checkGLcall("delete ffp program");
5726 heap_free(entry_arb
);
5729 /* Context activation is done by the caller. */
5730 static void arbfp_free(struct wined3d_device
*device
, struct wined3d_context
*context
)
5732 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
5733 struct shader_arb_priv
*priv
= device
->fragment_priv
;
5735 wine_rb_destroy(&priv
->fragment_shaders
, arbfp_free_ffpshader
, context_gl
);
5736 priv
->use_arbfp_fixed_func
= FALSE
;
5738 if (device
->shader_backend
!= &arb_program_shader_backend
)
5739 heap_free(device
->fragment_priv
);
5742 static void arbfp_get_caps(const struct wined3d_adapter
*adapter
, struct fragment_caps
*caps
)
5744 const struct wined3d_gl_info
*gl_info
= &adapter
->gl_info
;
5746 caps
->wined3d_caps
= WINED3D_FRAGMENT_CAP_PROJ_CONTROL
5747 | WINED3D_FRAGMENT_CAP_SRGB_WRITE
5748 | WINED3D_FRAGMENT_CAP_COLOR_KEY
;
5749 caps
->PrimitiveMiscCaps
= WINED3DPMISCCAPS_TSSARGTEMP
;
5750 caps
->TextureOpCaps
= WINED3DTEXOPCAPS_DISABLE
|
5751 WINED3DTEXOPCAPS_SELECTARG1
|
5752 WINED3DTEXOPCAPS_SELECTARG2
|
5753 WINED3DTEXOPCAPS_MODULATE4X
|
5754 WINED3DTEXOPCAPS_MODULATE2X
|
5755 WINED3DTEXOPCAPS_MODULATE
|
5756 WINED3DTEXOPCAPS_ADDSIGNED2X
|
5757 WINED3DTEXOPCAPS_ADDSIGNED
|
5758 WINED3DTEXOPCAPS_ADD
|
5759 WINED3DTEXOPCAPS_SUBTRACT
|
5760 WINED3DTEXOPCAPS_ADDSMOOTH
|
5761 WINED3DTEXOPCAPS_BLENDCURRENTALPHA
|
5762 WINED3DTEXOPCAPS_BLENDFACTORALPHA
|
5763 WINED3DTEXOPCAPS_BLENDTEXTUREALPHA
|
5764 WINED3DTEXOPCAPS_BLENDDIFFUSEALPHA
|
5765 WINED3DTEXOPCAPS_BLENDTEXTUREALPHAPM
|
5766 WINED3DTEXOPCAPS_MODULATEALPHA_ADDCOLOR
|
5767 WINED3DTEXOPCAPS_MODULATECOLOR_ADDALPHA
|
5768 WINED3DTEXOPCAPS_MODULATEINVCOLOR_ADDALPHA
|
5769 WINED3DTEXOPCAPS_MODULATEINVALPHA_ADDCOLOR
|
5770 WINED3DTEXOPCAPS_DOTPRODUCT3
|
5771 WINED3DTEXOPCAPS_MULTIPLYADD
|
5772 WINED3DTEXOPCAPS_LERP
|
5773 WINED3DTEXOPCAPS_BUMPENVMAP
|
5774 WINED3DTEXOPCAPS_BUMPENVMAPLUMINANCE
;
5776 /* TODO: Implement WINED3DTEXOPCAPS_PREMODULATE */
5778 caps
->MaxTextureBlendStages
= WINED3D_MAX_TEXTURES
;
5779 caps
->MaxSimultaneousTextures
= min(gl_info
->limits
.samplers
[WINED3D_SHADER_TYPE_PIXEL
], WINED3D_MAX_TEXTURES
);
5782 static unsigned int arbfp_get_emul_mask(const struct wined3d_gl_info
*gl_info
)
5784 return GL_EXT_EMUL_ARB_MULTITEXTURE
| GL_EXT_EMUL_EXT_FOG_COORD
;
5787 static void state_texfactor_arbfp(struct wined3d_context
*context
,
5788 const struct wined3d_state
*state
, DWORD state_id
)
5790 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
5791 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
5792 struct wined3d_device
*device
= context
->device
;
5793 struct wined3d_color color
;
5795 if (device
->shader_backend
== &arb_program_shader_backend
)
5797 struct shader_arb_priv
*priv
;
5799 /* Don't load the parameter if we're using an arbfp pixel shader,
5800 * otherwise we'll overwrite application provided constants. */
5804 priv
= device
->shader_priv
;
5805 priv
->pshader_const_dirty
[ARB_FFP_CONST_TFACTOR
] = 1;
5806 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_TFACTOR
+ 1);
5809 wined3d_color_from_d3dcolor(&color
, state
->render_states
[WINED3D_RS_TEXTUREFACTOR
]);
5810 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_TFACTOR
, &color
.r
));
5811 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_TFACTOR, &color.r)");
5814 static void state_tss_constant_arbfp(struct wined3d_context
*context
,
5815 const struct wined3d_state
*state
, DWORD state_id
)
5817 DWORD stage
= (state_id
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
5818 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
5819 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
5820 struct wined3d_device
*device
= context
->device
;
5821 struct wined3d_color color
;
5823 if (device
->shader_backend
== &arb_program_shader_backend
)
5825 struct shader_arb_priv
*priv
;
5827 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
5828 * application provided constants.
5833 priv
= device
->shader_priv
;
5834 priv
->pshader_const_dirty
[ARB_FFP_CONST_CONSTANT(stage
)] = 1;
5835 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_CONSTANT(stage
) + 1);
5838 wined3d_color_from_d3dcolor(&color
, state
->texture_states
[stage
][WINED3D_TSS_CONSTANT
]);
5839 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_CONSTANT(stage
), &color
.r
));
5840 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_CONSTANT(stage), &color.r)");
5843 static void state_arb_specularenable(struct wined3d_context
*context
,
5844 const struct wined3d_state
*state
, DWORD state_id
)
5846 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
5847 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
5848 struct wined3d_device
*device
= context
->device
;
5851 if (device
->shader_backend
== &arb_program_shader_backend
)
5853 struct shader_arb_priv
*priv
;
5855 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
5856 * application provided constants.
5861 priv
= device
->shader_priv
;
5862 priv
->pshader_const_dirty
[ARB_FFP_CONST_SPECULAR_ENABLE
] = 1;
5863 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_SPECULAR_ENABLE
+ 1);
5866 if (state
->render_states
[WINED3D_RS_SPECULARENABLE
])
5868 /* The specular color has no alpha */
5869 col
[0] = 1.0f
; col
[1] = 1.0f
;
5870 col
[2] = 1.0f
; col
[3] = 0.0f
;
5872 col
[0] = 0.0f
; col
[1] = 0.0f
;
5873 col
[2] = 0.0f
; col
[3] = 0.0f
;
5875 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_SPECULAR_ENABLE
, col
));
5876 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_SPECULAR_ENABLE, col)");
5879 static void set_bumpmat_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
5881 DWORD stage
= (state_id
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
5882 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
5883 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
5884 struct wined3d_device
*device
= context
->device
;
5887 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_BUMP_ENV
;
5889 if (device
->shader_backend
== &arb_program_shader_backend
)
5891 struct shader_arb_priv
*priv
= device
->shader_priv
;
5893 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants. */
5897 priv
->pshader_const_dirty
[ARB_FFP_CONST_BUMPMAT(stage
)] = 1;
5898 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_BUMPMAT(stage
) + 1);
5901 mat
[0][0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT00
]);
5902 mat
[0][1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT01
]);
5903 mat
[1][0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT10
]);
5904 mat
[1][1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT11
]);
5906 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_BUMPMAT(stage
), &mat
[0][0]));
5907 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_BUMPMAT(stage), &mat[0][0])");
5910 static void tex_bumpenvlum_arbfp(struct wined3d_context
*context
,
5911 const struct wined3d_state
*state
, DWORD state_id
)
5913 DWORD stage
= (state_id
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
5914 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
5915 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
5916 struct wined3d_device
*device
= context
->device
;
5919 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_BUMP_ENV
;
5921 if (device
->shader_backend
== &arb_program_shader_backend
)
5923 struct shader_arb_priv
*priv
= device
->shader_priv
;
5925 /* Exit now, don't set the luminance below, otherwise we may overwrite pixel shader constants. */
5929 priv
->pshader_const_dirty
[ARB_FFP_CONST_LUMINANCE(stage
)] = 1;
5930 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_LUMINANCE(stage
) + 1);
5933 param
[0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_LSCALE
]);
5934 param
[1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_LOFFSET
]);
5938 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_LUMINANCE(stage
), param
));
5939 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_LUMINANCE(stage), param)");
5942 static void alpha_test_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
5944 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
5945 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
5949 TRACE("context %p, state %p, state_id %#lx.\n", context
, state
, state_id
);
5951 if (state
->render_states
[WINED3D_RS_ALPHATESTENABLE
])
5953 gl_info
->gl_ops
.gl
.p_glEnable(GL_ALPHA_TEST
);
5954 checkGLcall("glEnable GL_ALPHA_TEST");
5958 gl_info
->gl_ops
.gl
.p_glDisable(GL_ALPHA_TEST
);
5959 checkGLcall("glDisable GL_ALPHA_TEST");
5963 ref
= wined3d_alpha_ref(state
);
5964 glParm
= wined3d_gl_compare_func(state
->render_states
[WINED3D_RS_ALPHAFUNC
]);
5968 gl_info
->gl_ops
.gl
.p_glAlphaFunc(glParm
, ref
);
5969 checkGLcall("glAlphaFunc");
5973 static void color_key_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
5975 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
5976 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
5977 const struct wined3d_texture
*texture
= state
->textures
[0];
5978 struct wined3d_device
*device
= context
->device
;
5979 struct wined3d_color float_key
[2];
5984 if (device
->shader_backend
== &arb_program_shader_backend
)
5986 struct shader_arb_priv
*priv
;
5988 /* Don't load the parameter if we're using an arbfp pixel shader,
5989 * otherwise we'll overwrite application provided constants. */
5993 priv
= device
->shader_priv
;
5994 priv
->pshader_const_dirty
[ARB_FFP_CONST_COLOR_KEY_LOW
] = 1;
5995 priv
->pshader_const_dirty
[ARB_FFP_CONST_COLOR_KEY_HIGH
] = 1;
5996 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_COLOR_KEY_HIGH
+ 1);
5999 wined3d_format_get_float_color_key(texture
->resource
.format
, &texture
->async
.src_blt_color_key
, float_key
);
6001 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_COLOR_KEY_LOW
, &float_key
[0].r
));
6002 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_COLOR_KEY_LOW, &float_key[0].r)");
6003 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_COLOR_KEY_HIGH
, &float_key
[1].r
));
6004 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_COLOR_KEY_HIGH, &float_key[1].r)");
6007 static const char *get_argreg(struct wined3d_string_buffer
*buffer
, unsigned int argnum
, unsigned int stage
, DWORD arg
)
6011 if(arg
== ARG_UNUSED
) return "unused"; /* This is the marker for unused registers */
6013 switch(arg
& WINED3DTA_SELECTMASK
) {
6014 case WINED3DTA_DIFFUSE
:
6015 ret
= "fragment.color.primary"; break;
6017 case WINED3DTA_CURRENT
:
6021 case WINED3DTA_TEXTURE
:
6023 case 0: ret
= "tex0"; break;
6024 case 1: ret
= "tex1"; break;
6025 case 2: ret
= "tex2"; break;
6026 case 3: ret
= "tex3"; break;
6027 case 4: ret
= "tex4"; break;
6028 case 5: ret
= "tex5"; break;
6029 case 6: ret
= "tex6"; break;
6030 case 7: ret
= "tex7"; break;
6031 default: ret
= "unknown texture";
6035 case WINED3DTA_TFACTOR
:
6036 ret
= "tfactor"; break;
6038 case WINED3DTA_SPECULAR
:
6039 ret
= "fragment.color.secondary"; break;
6041 case WINED3DTA_TEMP
:
6042 ret
= "tempreg"; break;
6044 case WINED3DTA_CONSTANT
:
6046 case 0: ret
= "const0"; break;
6047 case 1: ret
= "const1"; break;
6048 case 2: ret
= "const2"; break;
6049 case 3: ret
= "const3"; break;
6050 case 4: ret
= "const4"; break;
6051 case 5: ret
= "const5"; break;
6052 case 6: ret
= "const6"; break;
6053 case 7: ret
= "const7"; break;
6054 default: ret
= "unknown constant";
6062 if(arg
& WINED3DTA_COMPLEMENT
) {
6063 shader_addline(buffer
, "SUB arg%u, const.x, %s;\n", argnum
, ret
);
6064 if(argnum
== 0) ret
= "arg0";
6065 if(argnum
== 1) ret
= "arg1";
6066 if(argnum
== 2) ret
= "arg2";
6068 if(arg
& WINED3DTA_ALPHAREPLICATE
) {
6069 shader_addline(buffer
, "MOV arg%u, %s.w;\n", argnum
, ret
);
6070 if(argnum
== 0) ret
= "arg0";
6071 if(argnum
== 1) ret
= "arg1";
6072 if(argnum
== 2) ret
= "arg2";
6077 static void gen_ffp_instr(struct wined3d_string_buffer
*buffer
, unsigned int stage
, BOOL color
,
6078 BOOL alpha
, BOOL tmp_dst
, unsigned int op
, unsigned int dw_arg0
, unsigned int dw_arg1
, unsigned int dw_arg2
)
6080 const char *dstmask
, *dstreg
, *arg0
, *arg1
, *arg2
;
6081 unsigned int mul
= 1;
6090 dstreg
= tmp_dst
? "tempreg" : "ret";
6092 arg0
= get_argreg(buffer
, 0, stage
, dw_arg0
);
6093 arg1
= get_argreg(buffer
, 1, stage
, dw_arg1
);
6094 arg2
= get_argreg(buffer
, 2, stage
, dw_arg2
);
6098 case WINED3D_TOP_DISABLE
:
6101 case WINED3D_TOP_SELECT_ARG2
:
6104 case WINED3D_TOP_SELECT_ARG1
:
6105 shader_addline(buffer
, "MOV %s%s, %s;\n", dstreg
, dstmask
, arg1
);
6108 case WINED3D_TOP_MODULATE_4X
:
6111 case WINED3D_TOP_MODULATE_2X
:
6114 case WINED3D_TOP_MODULATE
:
6115 shader_addline(buffer
, "MUL %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6118 case WINED3D_TOP_ADD_SIGNED_2X
:
6121 case WINED3D_TOP_ADD_SIGNED
:
6122 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
6125 case WINED3D_TOP_ADD
:
6126 shader_addline(buffer
, "ADD_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6129 case WINED3D_TOP_SUBTRACT
:
6130 shader_addline(buffer
, "SUB_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6133 case WINED3D_TOP_ADD_SMOOTH
:
6134 shader_addline(buffer
, "SUB arg1, const.x, %s;\n", arg1
);
6135 shader_addline(buffer
, "MAD_SAT %s%s, arg1, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6138 case WINED3D_TOP_BLEND_CURRENT_ALPHA
:
6139 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_CURRENT
);
6140 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6142 case WINED3D_TOP_BLEND_FACTOR_ALPHA
:
6143 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TFACTOR
);
6144 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6146 case WINED3D_TOP_BLEND_TEXTURE_ALPHA
:
6147 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
6148 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6150 case WINED3D_TOP_BLEND_DIFFUSE_ALPHA
:
6151 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_DIFFUSE
);
6152 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6155 case WINED3D_TOP_BLEND_TEXTURE_ALPHA_PM
:
6156 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
6157 shader_addline(buffer
, "SUB arg0.w, const.x, %s.w;\n", arg0
);
6158 shader_addline(buffer
, "MAD_SAT %s%s, %s, arg0.w, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6161 /* D3DTOP_PREMODULATE ???? */
6163 case WINED3D_TOP_MODULATE_INVALPHA_ADD_COLOR
:
6164 shader_addline(buffer
, "SUB arg0.w, const.x, %s;\n", arg1
);
6165 shader_addline(buffer
, "MAD_SAT %s%s, arg0.w, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6167 case WINED3D_TOP_MODULATE_ALPHA_ADD_COLOR
:
6168 shader_addline(buffer
, "MAD_SAT %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
6170 case WINED3D_TOP_MODULATE_INVCOLOR_ADD_ALPHA
:
6171 shader_addline(buffer
, "SUB arg0, const.x, %s;\n", arg1
);
6172 shader_addline(buffer
, "MAD_SAT %s%s, arg0, %s, %s.w;\n", dstreg
, dstmask
, arg2
, arg1
);
6174 case WINED3D_TOP_MODULATE_COLOR_ADD_ALPHA
:
6175 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s.w;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
6178 case WINED3D_TOP_DOTPRODUCT3
:
6180 shader_addline(buffer
, "SUB arg1, %s, const.w;\n", arg1
);
6181 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
6182 shader_addline(buffer
, "DP3_SAT %s%s, arg1, arg2;\n", dstreg
, dstmask
);
6185 case WINED3D_TOP_MULTIPLY_ADD
:
6186 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg0
);
6189 case WINED3D_TOP_LERP
:
6190 /* The msdn is not quite right here */
6191 shader_addline(buffer
, "LRP %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6194 case WINED3D_TOP_BUMPENVMAP
:
6195 case WINED3D_TOP_BUMPENVMAP_LUMINANCE
:
6196 /* Those are handled in the first pass of the shader(generation pass 1 and 2) already */
6200 FIXME("Unhandled texture op %08x\n", op
);
6204 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.y;\n", dstreg
, dstmask
, dstreg
);
6206 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.z;\n", dstreg
, dstmask
, dstreg
);
6209 static const char *arbfp_texture_target(enum wined3d_gl_resource_type type
)
6213 case WINED3D_GL_RES_TYPE_TEX_1D
:
6215 case WINED3D_GL_RES_TYPE_TEX_2D
:
6217 case WINED3D_GL_RES_TYPE_TEX_3D
:
6219 case WINED3D_GL_RES_TYPE_TEX_CUBE
:
6221 case WINED3D_GL_RES_TYPE_TEX_RECT
:
6224 return "unexpected_resource_type";
6228 static GLuint
gen_arbfp_ffp_shader(const struct ffp_frag_settings
*settings
, const struct wined3d_gl_info
*gl_info
)
6230 BYTE tex_read
= 0, bump_used
= 0, luminance_used
= 0, constant_used
= 0;
6231 BOOL tempreg_used
= FALSE
, tfactor_used
= FALSE
;
6232 unsigned int stage
, lowest_disabled_stage
;
6233 struct wined3d_string_buffer buffer
;
6234 struct color_fixup_masks masks
;
6235 BOOL custom_linear_fog
= FALSE
;
6236 const char *textype
, *instr
;
6237 DWORD arg0
, arg1
, arg2
;
6238 char colorcor_dst
[8];
6242 if (!string_buffer_init(&buffer
))
6244 ERR("Failed to initialize shader buffer.\n");
6248 shader_addline(&buffer
, "!!ARBfp1.0\n");
6250 if (settings
->color_key_enabled
)
6252 shader_addline(&buffer
, "PARAM color_key_low = program.env[%u];\n", ARB_FFP_CONST_COLOR_KEY_LOW
);
6253 shader_addline(&buffer
, "PARAM color_key_high = program.env[%u];\n", ARB_FFP_CONST_COLOR_KEY_HIGH
);
6257 /* Find out which textures are read */
6258 for (stage
= 0; stage
< WINED3D_MAX_TEXTURES
; ++stage
)
6260 if (settings
->op
[stage
].cop
== WINED3D_TOP_DISABLE
)
6263 arg0
= settings
->op
[stage
].carg0
& WINED3DTA_SELECTMASK
;
6264 arg1
= settings
->op
[stage
].carg1
& WINED3DTA_SELECTMASK
;
6265 arg2
= settings
->op
[stage
].carg2
& WINED3DTA_SELECTMASK
;
6267 if (arg0
== WINED3DTA_TEXTURE
|| arg1
== WINED3DTA_TEXTURE
|| arg2
== WINED3DTA_TEXTURE
)
6268 tex_read
|= 1u << stage
;
6269 if (settings
->op
[stage
].tmp_dst
)
6270 tempreg_used
= TRUE
;
6271 if (arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
)
6272 tempreg_used
= TRUE
;
6273 if (arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
)
6274 tfactor_used
= TRUE
;
6275 if (arg0
== WINED3DTA_CONSTANT
|| arg1
== WINED3DTA_CONSTANT
|| arg2
== WINED3DTA_CONSTANT
)
6276 constant_used
|= 1u << stage
;
6278 switch (settings
->op
[stage
].cop
)
6280 case WINED3D_TOP_BUMPENVMAP_LUMINANCE
:
6281 luminance_used
|= 1u << stage
;
6283 case WINED3D_TOP_BUMPENVMAP
:
6284 bump_used
|= 1u << stage
;
6286 case WINED3D_TOP_BLEND_TEXTURE_ALPHA
:
6287 case WINED3D_TOP_BLEND_TEXTURE_ALPHA_PM
:
6288 tex_read
|= 1u << stage
;
6291 case WINED3D_TOP_BLEND_FACTOR_ALPHA
:
6292 tfactor_used
= TRUE
;
6299 if (settings
->op
[stage
].aop
== WINED3D_TOP_DISABLE
)
6302 arg0
= settings
->op
[stage
].aarg0
& WINED3DTA_SELECTMASK
;
6303 arg1
= settings
->op
[stage
].aarg1
& WINED3DTA_SELECTMASK
;
6304 arg2
= settings
->op
[stage
].aarg2
& WINED3DTA_SELECTMASK
;
6306 if (arg0
== WINED3DTA_TEXTURE
|| arg1
== WINED3DTA_TEXTURE
|| arg2
== WINED3DTA_TEXTURE
)
6307 tex_read
|= 1u << stage
;
6308 if (arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
)
6309 tempreg_used
= TRUE
;
6310 if (arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
)
6311 tfactor_used
= TRUE
;
6312 if (arg0
== WINED3DTA_CONSTANT
|| arg1
== WINED3DTA_CONSTANT
|| arg2
== WINED3DTA_CONSTANT
)
6313 constant_used
|= 1u << stage
;
6315 lowest_disabled_stage
= stage
;
6317 switch (settings
->fog
)
6319 case WINED3D_FFP_PS_FOG_OFF
: break;
6320 case WINED3D_FFP_PS_FOG_LINEAR
:
6321 if (gl_info
->quirks
& WINED3D_QUIRK_BROKEN_ARB_FOG
)
6323 custom_linear_fog
= TRUE
;
6326 shader_addline(&buffer
, "OPTION ARB_fog_linear;\n");
6329 case WINED3D_FFP_PS_FOG_EXP
: shader_addline(&buffer
, "OPTION ARB_fog_exp;\n"); break;
6330 case WINED3D_FFP_PS_FOG_EXP2
: shader_addline(&buffer
, "OPTION ARB_fog_exp2;\n"); break;
6331 default: FIXME("Unexpected fog setting %d\n", settings
->fog
);
6334 shader_addline(&buffer
, "PARAM const = {1, 2, 4, 0.5};\n");
6335 shader_addline(&buffer
, "TEMP TMP;\n");
6336 shader_addline(&buffer
, "TEMP ret;\n");
6337 if (tempreg_used
|| settings
->sRGB_write
) shader_addline(&buffer
, "TEMP tempreg;\n");
6338 shader_addline(&buffer
, "TEMP arg0;\n");
6339 shader_addline(&buffer
, "TEMP arg1;\n");
6340 shader_addline(&buffer
, "TEMP arg2;\n");
6341 for (stage
= 0; stage
< WINED3D_MAX_TEXTURES
; ++stage
)
6343 if (constant_used
& (1u << stage
))
6344 shader_addline(&buffer
, "PARAM const%u = program.env[%u];\n", stage
, ARB_FFP_CONST_CONSTANT(stage
));
6346 if (!(tex_read
& (1u << stage
)))
6349 shader_addline(&buffer
, "TEMP tex%u;\n", stage
);
6351 if (!(bump_used
& (1u << stage
)))
6353 shader_addline(&buffer
, "PARAM bumpmat%u = program.env[%u];\n", stage
, ARB_FFP_CONST_BUMPMAT(stage
));
6355 if (!(luminance_used
& (1u << stage
)))
6357 shader_addline(&buffer
, "PARAM luminance%u = program.env[%u];\n", stage
, ARB_FFP_CONST_LUMINANCE(stage
));
6360 shader_addline(&buffer
, "PARAM tfactor = program.env[%u];\n", ARB_FFP_CONST_TFACTOR
);
6361 shader_addline(&buffer
, "PARAM specular_enable = program.env[%u];\n", ARB_FFP_CONST_SPECULAR_ENABLE
);
6363 if (settings
->sRGB_write
)
6365 shader_addline(&buffer
, "PARAM srgb_consts0 = ");
6366 shader_arb_append_imm_vec4(&buffer
, &wined3d_srgb_const
[0].x
);
6367 shader_addline(&buffer
, ";\n");
6368 shader_addline(&buffer
, "PARAM srgb_consts1 = ");
6369 shader_arb_append_imm_vec4(&buffer
, &wined3d_srgb_const
[1].x
);
6370 shader_addline(&buffer
, ";\n");
6373 if (lowest_disabled_stage
< 7 && settings
->emul_clipplanes
)
6374 shader_addline(&buffer
, "KIL fragment.texcoord[7];\n");
6376 if (tempreg_used
|| settings
->sRGB_write
)
6377 shader_addline(&buffer
, "MOV tempreg, 0.0;\n");
6379 /* Generate texture sampling instructions */
6380 for (stage
= 0; stage
< WINED3D_MAX_TEXTURES
&& settings
->op
[stage
].cop
!= WINED3D_TOP_DISABLE
; ++stage
)
6382 if (!(tex_read
& (1u << stage
)))
6385 textype
= arbfp_texture_target(settings
->op
[stage
].tex_type
);
6387 if (settings
->op
[stage
].projected
== WINED3D_PROJECTION_NONE
)
6391 else if (settings
->op
[stage
].projected
== WINED3D_PROJECTION_COUNT4
6392 || settings
->op
[stage
].projected
== WINED3D_PROJECTION_COUNT3
)
6398 FIXME("Unexpected projection mode %d\n", settings
->op
[stage
].projected
);
6403 && (settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP
6404 || settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
))
6406 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, x, z, 0, 0;\n", stage
- 1);
6407 shader_addline(&buffer
, "DP3 ret.x, arg1, tex%u;\n", stage
- 1);
6408 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, y, w, 0, 0;\n", stage
- 1);
6409 shader_addline(&buffer
, "DP3 ret.y, arg1, tex%u;\n", stage
- 1);
6411 /* With projective textures, texbem only divides the static
6412 * texture coordinate, not the displacement, so multiply the
6413 * displacement with the dividing parameter before passing it to
6415 if (settings
->op
[stage
].projected
!= WINED3D_PROJECTION_NONE
)
6417 if (settings
->op
[stage
].projected
== WINED3D_PROJECTION_COUNT4
)
6419 shader_addline(&buffer
, "MOV ret.w, fragment.texcoord[%u].w;\n", stage
);
6420 shader_addline(&buffer
, "MUL ret.xyz, ret, fragment.texcoord[%u].w, fragment.texcoord[%u];\n",
6425 shader_addline(&buffer
, "MOV ret.w, fragment.texcoord[%u].z;\n", stage
);
6426 shader_addline(&buffer
, "MAD ret.xyz, ret, fragment.texcoord[%u].z, fragment.texcoord[%u];\n",
6432 shader_addline(&buffer
, "ADD ret, ret, fragment.texcoord[%u];\n", stage
);
6435 shader_addline(&buffer
, "%s tex%u, ret, texture[%u], %s;\n",
6436 instr
, stage
, stage
, textype
);
6437 if (settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
)
6439 shader_addline(&buffer
, "MAD_SAT ret.x, tex%u.z, luminance%u.x, luminance%u.y;\n",
6440 stage
- 1, stage
- 1, stage
- 1);
6441 shader_addline(&buffer
, "MUL tex%u, tex%u, ret.x;\n", stage
, stage
);
6444 else if (settings
->op
[stage
].projected
== WINED3D_PROJECTION_COUNT3
)
6446 shader_addline(&buffer
, "MOV ret, fragment.texcoord[%u];\n", stage
);
6447 shader_addline(&buffer
, "MOV ret.w, ret.z;\n");
6448 shader_addline(&buffer
, "%s tex%u, ret, texture[%u], %s;\n",
6449 instr
, stage
, stage
, textype
);
6453 shader_addline(&buffer
, "%s tex%u, fragment.texcoord[%u], texture[%u], %s;\n",
6454 instr
, stage
, stage
, stage
, textype
);
6457 sprintf(colorcor_dst
, "tex%u", stage
);
6458 masks
= calc_color_correction(settings
->op
[stage
].color_fixup
, WINED3DSP_WRITEMASK_ALL
);
6459 gen_color_correction(&buffer
, colorcor_dst
, colorcor_dst
, "const.x", "const.y",
6460 settings
->op
[stage
].color_fixup
, masks
);
6463 if (settings
->color_key_enabled
)
6465 shader_addline(&buffer
, "SLT TMP, tex0, color_key_low;\n"); /* below low key */
6466 shader_addline(&buffer
, "SGE ret, tex0, color_key_high;\n"); /* above high key */
6467 shader_addline(&buffer
, "ADD TMP, TMP, ret;\n"); /* or */
6468 shader_addline(&buffer
, "DP4 TMP.b, TMP, TMP;\n"); /* on any channel */
6469 shader_addline(&buffer
, "SGE TMP, -TMP.b, 0.0;\n"); /* logical not */
6470 shader_addline(&buffer
, "KIL -TMP;\n"); /* discard if true */
6473 shader_addline(&buffer
, "MOV ret, fragment.color.primary;\n");
6475 /* Generate the main shader */
6476 for (stage
= 0; stage
< WINED3D_MAX_TEXTURES
; ++stage
)
6478 if (settings
->op
[stage
].cop
== WINED3D_TOP_DISABLE
)
6481 if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG1
6482 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG1
)
6483 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
;
6484 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG1
6485 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG2
)
6486 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg2
;
6487 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG2
6488 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG1
)
6489 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg1
;
6490 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG2
6491 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG2
)
6492 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
6494 op_equal
= settings
->op
[stage
].aop
== settings
->op
[stage
].cop
6495 && settings
->op
[stage
].carg0
== settings
->op
[stage
].aarg0
6496 && settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
6497 && settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
6499 if (settings
->op
[stage
].aop
== WINED3D_TOP_DISABLE
)
6501 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].tmp_dst
,
6502 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6503 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6507 gen_ffp_instr(&buffer
, stage
, TRUE
, TRUE
, settings
->op
[stage
].tmp_dst
,
6508 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6509 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6511 else if (settings
->op
[stage
].cop
!= WINED3D_TOP_BUMPENVMAP
6512 && settings
->op
[stage
].cop
!= WINED3D_TOP_BUMPENVMAP_LUMINANCE
)
6514 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].tmp_dst
,
6515 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6516 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6517 gen_ffp_instr(&buffer
, stage
, FALSE
, TRUE
, settings
->op
[stage
].tmp_dst
,
6518 settings
->op
[stage
].aop
, settings
->op
[stage
].aarg0
,
6519 settings
->op
[stage
].aarg1
, settings
->op
[stage
].aarg2
);
6523 if (settings
->sRGB_write
|| custom_linear_fog
)
6525 shader_addline(&buffer
, "MAD ret, fragment.color.secondary, specular_enable, ret;\n");
6526 if (settings
->sRGB_write
)
6527 arbfp_add_sRGB_correction(&buffer
, "ret", "arg0", "arg1", "arg2", "tempreg", FALSE
);
6528 if (custom_linear_fog
)
6529 arbfp_add_linear_fog(&buffer
, "ret", "arg0");
6530 shader_addline(&buffer
, "MOV result.color, ret;\n");
6534 shader_addline(&buffer
, "MAD result.color, fragment.color.secondary, specular_enable, ret;\n");
6538 shader_addline(&buffer
, "END\n");
6540 /* Generate the shader */
6541 GL_EXTCALL(glGenProgramsARB(1, &ret
));
6542 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, ret
));
6543 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
6545 string_buffer_free(&buffer
);
6549 static void fragment_prog_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6551 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
6552 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
6553 const struct wined3d_device
*device
= context
->device
;
6554 struct shader_arb_priv
*priv
= device
->fragment_priv
;
6555 BOOL use_pshader
= use_ps(state
);
6556 struct ffp_frag_settings settings
;
6557 const struct arbfp_ffp_desc
*desc
;
6560 TRACE("context %p, state %p, state_id %#lx.\n", context
, state
, state_id
);
6562 if (isStateDirty(context
, STATE_RENDER(WINED3D_RS_FOGENABLE
)))
6564 if (!use_pshader
&& device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
)
6566 /* Reload fixed function constants since they collide with the
6567 * pixel shader constants. */
6568 for (i
= 0; i
< WINED3D_MAX_TEXTURES
; ++i
)
6570 set_bumpmat_arbfp(context
, state
, STATE_TEXTURESTAGE(i
, WINED3D_TSS_BUMPENV_MAT00
));
6571 state_tss_constant_arbfp(context
, state
, STATE_TEXTURESTAGE(i
, WINED3D_TSS_CONSTANT
));
6573 state_texfactor_arbfp(context
, state
, STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
));
6574 state_arb_specularenable(context
, state
, STATE_RENDER(WINED3D_RS_SPECULARENABLE
));
6575 color_key_arbfp(context
, state
, STATE_COLOR_KEY
);
6577 else if (use_pshader
)
6579 context
->shader_update_mask
|= 1u << WINED3D_SHADER_TYPE_PIXEL
;
6586 /* Find or create a shader implementing the fixed function pipeline
6587 * settings, then activate it. */
6588 wined3d_ffp_get_fs_settings(context
, state
, &settings
, FALSE
);
6589 desc
= (const struct arbfp_ffp_desc
*)find_ffp_frag_shader(&priv
->fragment_shaders
, &settings
);
6592 struct arbfp_ffp_desc
*new_desc
;
6594 if (!(new_desc
= heap_alloc(sizeof(*new_desc
))))
6596 ERR("Out of memory\n");
6600 new_desc
->parent
.settings
= settings
;
6601 new_desc
->shader
= gen_arbfp_ffp_shader(&settings
, gl_info
);
6602 add_ffp_frag_shader(&priv
->fragment_shaders
, &new_desc
->parent
);
6603 TRACE("Allocated fixed function replacement shader descriptor %p\n", new_desc
);
6607 /* Now activate the replacement program. GL_FRAGMENT_PROGRAM_ARB is already active (however, note the
6608 * comment above the shader_select call below). If e.g. GLSL is active, the shader_select call will
6611 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, desc
->shader
));
6612 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, desc->shader)");
6613 priv
->current_fprogram_id
= desc
->shader
;
6615 if (device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
)
6617 /* Reload fixed function constants since they collide with the
6618 * pixel shader constants. */
6619 for (i
= 0; i
< WINED3D_MAX_TEXTURES
; ++i
)
6621 set_bumpmat_arbfp(context
, state
, STATE_TEXTURESTAGE(i
, WINED3D_TSS_BUMPENV_MAT00
));
6622 state_tss_constant_arbfp(context
, state
, STATE_TEXTURESTAGE(i
, WINED3D_TSS_CONSTANT
));
6624 state_texfactor_arbfp(context
, state
, STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
));
6625 state_arb_specularenable(context
, state
, STATE_RENDER(WINED3D_RS_SPECULARENABLE
));
6626 color_key_arbfp(context
, state
, STATE_COLOR_KEY
);
6628 context
->last_was_pshader
= FALSE
;
6630 else if (!context
->last_was_pshader
)
6632 if (device
->shader_backend
== &arb_program_shader_backend
)
6633 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_F
;
6634 context
->last_was_pshader
= TRUE
;
6637 context
->shader_update_mask
|= 1u << WINED3D_SHADER_TYPE_PIXEL
;
6640 /* We can't link the fog states to the fragment state directly since the
6641 * vertex pipeline links them to FOGENABLE. A different linking in different
6642 * pipeline parts can't be expressed in the combined state table, so we need
6643 * to handle that with a forwarding function. The other invisible side effect
6644 * is that changing the fog start and fog end (which links to FOGENABLE in
6645 * vertex) results in the fragment_prog_arbfp function being called because
6646 * FOGENABLE is dirty, which calls this function here. */
6647 static void state_arbfp_fog(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6649 enum fogsource new_source
;
6650 DWORD fogstart
= state
->render_states
[WINED3D_RS_FOGSTART
];
6651 DWORD fogend
= state
->render_states
[WINED3D_RS_FOGEND
];
6653 TRACE("context %p, state %p, state_id %#lx.\n", context
, state
, state_id
);
6655 if (!isStateDirty(context
, STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
)))
6656 fragment_prog_arbfp(context
, state
, state_id
);
6658 if (!state
->render_states
[WINED3D_RS_FOGENABLE
])
6661 if (state
->render_states
[WINED3D_RS_FOGTABLEMODE
] == WINED3D_FOG_NONE
)
6665 new_source
= FOGSOURCE_VS
;
6669 if (state
->render_states
[WINED3D_RS_FOGVERTEXMODE
] == WINED3D_FOG_NONE
|| context
->last_was_rhw
)
6670 new_source
= FOGSOURCE_COORD
;
6672 new_source
= FOGSOURCE_FFP
;
6677 new_source
= FOGSOURCE_FFP
;
6680 if (new_source
!= context
->fog_source
|| fogstart
== fogend
)
6682 context
->fog_source
= new_source
;
6683 state_fogstartend(context
, state
, STATE_RENDER(WINED3D_RS_FOGSTART
));
6687 static void textransform(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6689 if (!isStateDirty(context
, STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
)))
6690 fragment_prog_arbfp(context
, state
, state_id
);
6693 static const struct wined3d_state_entry_template arbfp_fragmentstate_template
[] =
6695 {STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
), { STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
), state_texfactor_arbfp
}, WINED3D_GL_EXT_NONE
},
6696 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6697 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6698 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6699 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6700 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6701 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6702 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6703 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6704 {STATE_TEXTURESTAGE(0, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6705 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6706 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6707 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6708 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6709 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6710 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6711 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6712 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6713 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6714 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6715 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6716 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6717 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6718 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6719 {STATE_TEXTURESTAGE(1, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6720 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6721 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6722 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6723 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6724 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6725 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6726 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6727 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6728 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6729 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6730 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6731 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6732 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6733 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6734 {STATE_TEXTURESTAGE(2, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6735 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6736 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6737 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6738 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6739 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6740 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6741 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6742 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6743 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6744 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6745 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6746 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6747 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6748 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6749 {STATE_TEXTURESTAGE(3, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6750 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6751 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6752 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6753 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6754 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6755 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6756 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6757 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6758 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6759 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6760 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6761 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6762 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6763 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6764 {STATE_TEXTURESTAGE(4, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6765 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6766 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6767 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6768 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6769 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6770 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6771 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6772 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6773 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6774 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6775 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6776 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6777 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6778 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6779 {STATE_TEXTURESTAGE(5, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6780 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6781 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6782 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6783 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6784 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6785 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6786 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6787 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6788 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6789 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6790 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6791 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6792 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6793 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6794 {STATE_TEXTURESTAGE(6, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6795 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6796 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6797 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6798 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6799 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6800 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6801 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6802 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6803 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6804 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6805 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6806 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6807 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6808 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6809 {STATE_TEXTURESTAGE(7, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6810 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6811 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6812 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6813 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6814 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6815 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6816 {STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
6817 {STATE_RENDER(WINED3D_RS_ALPHAFUNC
), { STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6818 {STATE_RENDER(WINED3D_RS_ALPHAREF
), { STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6819 {STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), { STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), alpha_test_arbfp
}, WINED3D_GL_EXT_NONE
},
6820 {STATE_RENDER(WINED3D_RS_COLORKEYENABLE
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6821 {STATE_COLOR_KEY
, { STATE_COLOR_KEY
, color_key_arbfp
}, WINED3D_GL_EXT_NONE
},
6822 {STATE_RENDER(WINED3D_RS_FOGENABLE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), state_arbfp_fog
}, WINED3D_GL_EXT_NONE
},
6823 {STATE_RENDER(WINED3D_RS_FOGTABLEMODE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6824 {STATE_RENDER(WINED3D_RS_FOGVERTEXMODE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6825 {STATE_RENDER(WINED3D_RS_FOGSTART
), { STATE_RENDER(WINED3D_RS_FOGSTART
), state_fogstartend
}, WINED3D_GL_EXT_NONE
},
6826 {STATE_RENDER(WINED3D_RS_FOGEND
), { STATE_RENDER(WINED3D_RS_FOGSTART
), NULL
}, WINED3D_GL_EXT_NONE
},
6827 {STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), { STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), state_srgbwrite
}, ARB_FRAMEBUFFER_SRGB
},
6828 {STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6829 {STATE_RENDER(WINED3D_RS_FOGCOLOR
), { STATE_RENDER(WINED3D_RS_FOGCOLOR
), state_fogcolor
}, WINED3D_GL_EXT_NONE
},
6830 {STATE_RENDER(WINED3D_RS_FOGDENSITY
), { STATE_RENDER(WINED3D_RS_FOGDENSITY
), state_fogdensity
}, WINED3D_GL_EXT_NONE
},
6831 {STATE_TEXTURESTAGE(0,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(0,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6832 {STATE_TEXTURESTAGE(1,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(1,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6833 {STATE_TEXTURESTAGE(2,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(2,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6834 {STATE_TEXTURESTAGE(3,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(3,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6835 {STATE_TEXTURESTAGE(4,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(4,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6836 {STATE_TEXTURESTAGE(5,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(5,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6837 {STATE_TEXTURESTAGE(6,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(6,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6838 {STATE_TEXTURESTAGE(7,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(7,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6839 {STATE_TEXTURESTAGE(0, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6840 {STATE_TEXTURESTAGE(1, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6841 {STATE_TEXTURESTAGE(2, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6842 {STATE_TEXTURESTAGE(3, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6843 {STATE_TEXTURESTAGE(4, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6844 {STATE_TEXTURESTAGE(5, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6845 {STATE_TEXTURESTAGE(6, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6846 {STATE_TEXTURESTAGE(7, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6847 {STATE_RENDER(WINED3D_RS_SPECULARENABLE
), { STATE_RENDER(WINED3D_RS_SPECULARENABLE
), state_arb_specularenable
}, WINED3D_GL_EXT_NONE
},
6848 {STATE_RENDER(WINED3D_RS_SHADEMODE
), { STATE_RENDER(WINED3D_RS_SHADEMODE
), state_shademode
}, WINED3D_GL_EXT_NONE
},
6849 {0 /* Terminate */, { 0, 0 }, WINED3D_GL_EXT_NONE
},
6852 static BOOL
arbfp_alloc_context_data(struct wined3d_context
*context
)
6857 static void arbfp_free_context_data(struct wined3d_context
*context
)
6861 const struct wined3d_fragment_pipe_ops arbfp_fragment_pipeline
=
6865 arbfp_get_emul_mask
,
6868 arbfp_alloc_context_data
,
6869 arbfp_free_context_data
,
6870 shader_arb_color_fixup_supported
,
6871 arbfp_fragmentstate_template
,
6874 struct arbfp_blit_type
6876 enum complex_fixup fixup
: 4;
6877 enum wined3d_gl_resource_type res_type
: 3;
6878 DWORD use_color_key
: 1;
6882 struct arbfp_blit_desc
6885 struct arbfp_blit_type type
;
6886 struct wine_rb_entry entry
;
6889 #define ARBFP_BLIT_PARAM_SIZE 0
6890 #define ARBFP_BLIT_PARAM_COLOR_KEY_LOW 1
6891 #define ARBFP_BLIT_PARAM_COLOR_KEY_HIGH 2
6893 struct wined3d_arbfp_blitter
6895 struct wined3d_blitter blitter
;
6896 struct wine_rb_tree shaders
;
6897 GLuint palette_texture
;
6900 static int arbfp_blit_type_compare(const void *key
, const struct wine_rb_entry
*entry
)
6902 const struct arbfp_blit_type
*ka
= key
;
6903 const struct arbfp_blit_type
*kb
= &WINE_RB_ENTRY_VALUE(entry
, const struct arbfp_blit_desc
, entry
)->type
;
6905 return memcmp(ka
, kb
, sizeof(*ka
));
6908 /* Context activation is done by the caller. */
6909 static void arbfp_free_blit_shader(struct wine_rb_entry
*entry
, void *ctx
)
6911 struct arbfp_blit_desc
*entry_arb
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_blit_desc
, entry
);
6912 const struct wined3d_gl_info
*gl_info
;
6913 struct wined3d_context_gl
*context_gl
;
6916 gl_info
= context_gl
->gl_info
;
6918 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb
->shader
));
6919 checkGLcall("glDeleteProgramsARB(1, &entry_arb->shader)");
6920 heap_free(entry_arb
);
6923 /* Context activation is done by the caller. */
6924 static void arbfp_blitter_destroy(struct wined3d_blitter
*blitter
, struct wined3d_context
*context
)
6926 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
6927 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
6928 struct wined3d_arbfp_blitter
*arbfp_blitter
;
6929 struct wined3d_blitter
*next
;
6931 if ((next
= blitter
->next
))
6932 next
->ops
->blitter_destroy(next
, &context_gl
->c
);
6934 arbfp_blitter
= CONTAINING_RECORD(blitter
, struct wined3d_arbfp_blitter
, blitter
);
6936 wine_rb_destroy(&arbfp_blitter
->shaders
, arbfp_free_blit_shader
, context_gl
);
6937 checkGLcall("Delete blit programs");
6939 if (arbfp_blitter
->palette_texture
)
6940 gl_info
->gl_ops
.gl
.p_glDeleteTextures(1, &arbfp_blitter
->palette_texture
);
6942 heap_free(arbfp_blitter
);
6945 static void gen_packed_yuv_read(struct wined3d_string_buffer
*buffer
,
6946 const struct arbfp_blit_type
*type
, char *luminance
)
6949 const char *tex
, *texinstr
= "TXP";
6951 if (type
->fixup
== COMPLEX_FIXUP_UYVY
)
6962 tex
= arbfp_texture_target(type
->res_type
);
6963 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_RECT
)
6966 /* First we have to read the chroma values. This means we need at least two pixels(no filtering),
6967 * or 4 pixels(with filtering). To get the unmodified chromas, we have to rid ourselves of the
6968 * filtering when we sample the texture.
6970 * These are the rules for reading the chroma:
6976 * So we have to get the sampling x position in non-normalized coordinates in integers
6978 if (type
->res_type
!= WINED3D_GL_RES_TYPE_TEX_RECT
)
6980 shader_addline(buffer
, "MUL texcrd.xy, fragment.texcoord[0], size.x;\n");
6981 shader_addline(buffer
, "MOV texcrd.w, size.x;\n");
6985 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
6987 /* We must not allow filtering between pixel x and x+1, this would mix U and V
6988 * Vertical filtering is ok. However, bear in mind that the pixel center is at
6991 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
6992 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, coef.y;\n");
6994 /* Multiply the x coordinate by 0.5 and get the fraction. This gives 0.25
6995 * and 0.75 for the even and odd pixels respectively. */
6996 shader_addline(buffer
, "MUL texcrd2, texcrd, coef.y;\n");
6997 shader_addline(buffer
, "FRC texcrd2, texcrd2;\n");
6999 /* Sample Pixel 1. */
7000 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
7002 /* Put the value into either of the chroma values */
7003 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
7004 shader_addline(buffer
, "MUL chroma.x, luminance.%c, temp.x;\n", chroma
);
7005 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
7006 shader_addline(buffer
, "MUL chroma.y, luminance.%c, temp.x;\n", chroma
);
7008 /* Sample pixel 2. If we read an even pixel(SLT above returned 1), sample
7009 * the pixel right to the current one. Otherwise, sample the left pixel.
7010 * Bias and scale the SLT result to -1;1 and add it to the texcrd.x.
7012 shader_addline(buffer
, "MAD temp.x, temp.x, coef.z, -coef.x;\n");
7013 shader_addline(buffer
, "ADD texcrd.x, texcrd, temp.x;\n");
7014 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
7016 /* Put the value into the other chroma */
7017 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
7018 shader_addline(buffer
, "MAD chroma.y, luminance.%c, temp.x, chroma.y;\n", chroma
);
7019 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
7020 shader_addline(buffer
, "MAD chroma.x, luminance.%c, temp.x, chroma.x;\n", chroma
);
7022 /* TODO: If filtering is enabled, sample a 2nd pair of pixels left or right of
7023 * the current one and lerp the two U and V values
7026 /* This gives the correctly filtered luminance value */
7027 shader_addline(buffer
, "TEX luminance, fragment.texcoord[0], texture[0], %s;\n", tex
);
7030 static void gen_yv12_read(struct wined3d_string_buffer
*buffer
,
7031 const struct arbfp_blit_type
*type
, char *luminance
)
7034 static const float yv12_coef
[]
7035 = {2.0f
/ 3.0f
, 1.0f
/ 6.0f
, (2.0f
/ 3.0f
) + (1.0f
/ 6.0f
), 1.0f
/ 3.0f
};
7037 tex
= arbfp_texture_target(type
->res_type
);
7039 /* YV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
7040 * V and a (W/2)x(H/2) U plane, each with 8 bit per pixel. So the effective
7041 * bitdepth is 12 bits per pixel. Since the U and V planes have only half the
7042 * pitch of the luminance plane, the packing into the gl texture is a bit
7043 * unfortunate. If the whole texture is interpreted as luminance data it looks
7044 * approximately like this:
7046 * +----------------------------------+----
7058 * +----------------+-----------------+----
7060 * | V even rows | V odd rows |
7062 * +----------------+------------------ -
7064 * | U even rows | U odd rows |
7066 * +----------------+-----------------+----
7070 * So it appears as if there are 4 chroma images, but in fact the odd rows
7071 * in the chroma images are in the same row as the even ones. So it is
7072 * kinda tricky to read
7074 * When reading from rectangle textures, keep in mind that the input y coordinates
7075 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height
7077 shader_addline(buffer
, "PARAM yv12_coef = ");
7078 shader_arb_append_imm_vec4(buffer
, yv12_coef
);
7079 shader_addline(buffer
, ";\n");
7081 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7082 /* the chroma planes have only half the width */
7083 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, coef.y;\n");
7085 /* The first value is between 2/3 and 5/6th of the texture's height, so scale+bias
7086 * the coordinate. Also read the right side of the image when reading odd lines
7088 * Don't forget to clamp the y values in into the range, otherwise we'll get filtering
7091 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7093 shader_addline(buffer
, "RCP chroma.w, size.y;\n");
7095 shader_addline(buffer
, "MUL texcrd2.y, texcrd.y, size.y;\n");
7097 shader_addline(buffer
, "FLR texcrd2.y, texcrd2.y;\n");
7098 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, yv12_coef.y, yv12_coef.x;\n");
7100 /* Read odd lines from the right side (add size * 0.5 to the x coordinate). */
7101 shader_addline(buffer
, "ADD texcrd2.x, texcrd2.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
7102 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
7103 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
7104 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
7106 /* clamp, keep the half pixel origin in mind */
7107 shader_addline(buffer
, "MAD temp.y, coef.y, chroma.w, yv12_coef.x;\n");
7108 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7109 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.z;\n");
7110 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7114 /* The y coordinate for V is in the range [size, size + size / 4). */
7115 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
7116 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, coef.w, size.y;\n");
7118 /* Read odd lines from the right side (add size * 0.5 to the x coordinate). */
7119 shader_addline(buffer
, "ADD texcrd2.x, texcrd.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
7120 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
7121 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
7122 shader_addline(buffer
, "MUL texcrd2.x, texcrd2.x, size.x;\n");
7123 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
7125 /* Make sure to read exactly from the pixel center */
7126 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
7127 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, coef.y;\n");
7130 shader_addline(buffer
, "MAD temp.y, size.y, coef.w, size.y;\n");
7131 shader_addline(buffer
, "ADD temp.y, temp.y, -coef.y;\n");
7132 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7133 shader_addline(buffer
, "ADD temp.y, size.y, coef.y;\n");
7134 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7136 /* Read the texture, put the result into the output register */
7137 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7138 shader_addline(buffer
, "MOV chroma.x, temp.w;\n");
7140 /* The other chroma value is 1/6th of the texture lower, from 5/6th to 6/6th
7141 * No need to clamp because we're just reusing the already clamped value from above
7143 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7144 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, yv12_coef.y;\n");
7146 shader_addline(buffer
, "MAD texcrd.y, size.y, coef.w, texcrd.y;\n");
7147 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7148 shader_addline(buffer
, "MOV chroma.y, temp.w;\n");
7150 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
7151 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
7152 * values due to filtering
7154 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7155 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7157 /* Multiply the y coordinate by 2/3 and clamp it */
7158 shader_addline(buffer
, "MUL texcrd.y, texcrd.y, yv12_coef.x;\n");
7159 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.x;\n");
7160 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7161 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7165 /* Reading from texture_rectangles is pretty straightforward, just use the unmodified
7166 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
7169 shader_addline(buffer
, "ADD temp.x, size.y, -coef.y;\n");
7170 shader_addline(buffer
, "MIN texcrd.y, texcrd.y, size.x;\n");
7171 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7176 static void gen_nv12_read(struct wined3d_string_buffer
*buffer
,
7177 const struct arbfp_blit_type
*type
, char *luminance
)
7180 static const float nv12_coef
[]
7181 = {2.0f
/ 3.0f
, 1.0f
/ 3.0f
, 1.0f
, 1.0f
};
7183 tex
= arbfp_texture_target(type
->res_type
);
7185 /* NV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
7186 * sized plane where each component is an UV pair. So the effective
7187 * bitdepth is 12 bits per pixel If the whole texture is interpreted as luminance
7188 * data it looks approximately like this:
7190 * +----------------------------------+----
7202 * +----------------------------------+----
7203 * |UVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUV|
7204 * |UVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUV|
7210 * +----------------------------------+----
7212 * When reading from rectangle textures, keep in mind that the input y coordinates
7213 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height. */
7215 shader_addline(buffer
, "PARAM nv12_coef = ");
7216 shader_arb_append_imm_vec4(buffer
, nv12_coef
);
7217 shader_addline(buffer
, ";\n");
7219 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7220 /* We only have half the number of chroma pixels. */
7221 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, coef.y;\n");
7223 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7225 shader_addline(buffer
, "RCP chroma.w, size.x;\n");
7226 shader_addline(buffer
, "RCP chroma.z, size.y;\n");
7228 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, nv12_coef.y, nv12_coef.x;\n");
7230 /* We must not allow filtering horizontally, this would mix U and V.
7231 * Vertical filtering is ok. However, bear in mind that the pixel center is at
7232 * 0.5, so add 0.5. */
7234 /* Convert to non-normalized coordinates so we can find the
7235 * individual pixel. */
7236 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, size.x;\n");
7237 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
7238 /* Multiply by 2 since chroma components are stored in UV pixel pairs,
7239 * add 0.5 to hit the center of the pixel. */
7240 shader_addline(buffer
, "MAD texcrd.x, texcrd.x, coef.z, coef.y;\n");
7242 /* Convert back to normalized coordinates. */
7243 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, chroma.w;\n");
7245 /* Clamp, keep the half pixel origin in mind. */
7246 shader_addline(buffer
, "MAD temp.y, coef.y, chroma.z, nv12_coef.x;\n");
7247 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7248 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.z, nv12_coef.z;\n");
7249 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7253 /* The y coordinate for chroma is in the range [size, size + size / 2). */
7254 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, coef.y, size.y;\n");
7256 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
7257 /* Multiply by 2 since chroma components are stored in UV pixel pairs,
7258 * add 0.5 to hit the center of the pixel. */
7259 shader_addline(buffer
, "MAD texcrd.x, texcrd.x, coef.z, coef.y;\n");
7262 shader_addline(buffer
, "MAD temp.y, size.y, coef.y, size.y;\n");
7263 shader_addline(buffer
, "ADD temp.y, temp.y, -coef.y;\n");
7264 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7265 shader_addline(buffer
, "ADD temp.y, size.y, coef.y;\n");
7266 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7268 /* Read the texture, put the result into the output register. */
7269 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7270 shader_addline(buffer
, "MOV chroma.y, temp.w;\n");
7272 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7275 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, chroma.w;\n");
7280 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, coef.x;\n");
7282 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7283 shader_addline(buffer
, "MOV chroma.x, temp.w;\n");
7285 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
7286 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
7287 * values due to filtering. */
7288 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7289 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7291 /* Multiply the y coordinate by 2/3 and clamp it */
7292 shader_addline(buffer
, "MUL texcrd.y, texcrd.y, nv12_coef.x;\n");
7293 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, nv12_coef.x;\n");
7294 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7295 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7299 /* Reading from texture_rectangles is pretty straightforward, just use the unmodified
7300 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
7303 shader_addline(buffer
, "ADD temp.x, size.y, -coef.y;\n");
7304 shader_addline(buffer
, "MIN texcrd.y, texcrd.y, size.x;\n");
7305 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7310 /* Context activation is done by the caller. */
7311 static GLuint
gen_p8_shader(const struct wined3d_gl_info
*gl_info
, const struct arbfp_blit_type
*type
)
7314 struct wined3d_string_buffer buffer
;
7315 const char *tex_target
= arbfp_texture_target(type
->res_type
);
7317 /* This should not happen because we only use this conversion for
7318 * present blits which don't use color keying. */
7319 if (type
->use_color_key
)
7320 FIXME("Implement P8 color keying.\n");
7323 if (!string_buffer_init(&buffer
))
7325 ERR("Failed to initialize shader buffer.\n");
7329 GL_EXTCALL(glGenProgramsARB(1, &shader
));
7330 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7333 string_buffer_free(&buffer
);
7337 shader_addline(&buffer
, "!!ARBfp1.0\n");
7338 shader_addline(&buffer
, "TEMP index;\n");
7340 /* { 255/256, 0.5/255*255/256, 0, 0 } */
7341 shader_addline(&buffer
, "PARAM constants = { 0.996, 0.00195, 0, 0 };\n");
7343 /* The alpha-component contains the palette index */
7344 shader_addline(&buffer
, "TEX index, fragment.texcoord[0], texture[0], %s;\n", tex_target
);
7346 /* Scale the index by 255/256 and add a bias of '0.5' in order to sample in the middle */
7347 shader_addline(&buffer
, "MAD index.a, index.a, constants.x, constants.y;\n");
7349 /* Use the alpha-component as an index in the palette to get the final color */
7350 shader_addline(&buffer
, "TEX result.color, index.a, texture[1], 1D;\n");
7351 shader_addline(&buffer
, "END\n");
7353 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
7355 string_buffer_free(&buffer
);
7360 /* Context activation is done by the caller. */
7361 static void arbfp_blitter_upload_palette(struct wined3d_arbfp_blitter
*blitter
,
7362 const struct wined3d_texture_gl
*texture_gl
, struct wined3d_context_gl
*context_gl
)
7364 const struct wined3d_palette
*palette
= texture_gl
->t
.swapchain
? texture_gl
->t
.swapchain
->palette
: NULL
;
7365 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
7367 if (!blitter
->palette_texture
)
7368 gl_info
->gl_ops
.gl
.p_glGenTextures(1, &blitter
->palette_texture
);
7370 GL_EXTCALL(glActiveTexture(GL_TEXTURE1
));
7371 gl_info
->gl_ops
.gl
.p_glBindTexture(GL_TEXTURE_1D
, blitter
->palette_texture
);
7373 gl_info
->gl_ops
.gl
.p_glTexEnvi(GL_TEXTURE_ENV
, GL_TEXTURE_ENV_MODE
, GL_REPLACE
);
7375 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_MIN_FILTER
, GL_NEAREST
);
7376 /* Make sure we have discrete color levels. */
7377 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_MAG_FILTER
, GL_NEAREST
);
7378 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_WRAP_S
, GL_CLAMP_TO_EDGE
);
7379 /* TODO: avoid unneeded uploads in the future by adding some SFLAG_PALETTE_DIRTY mechanism */
7382 gl_info
->gl_ops
.gl
.p_glTexImage1D(GL_TEXTURE_1D
, 0, GL_RGB
, 256, 0, GL_BGRA
,
7383 GL_UNSIGNED_INT_8_8_8_8_REV
, palette
->colors
);
7387 static const DWORD black
;
7388 FIXME("P8 surface loaded without a palette.\n");
7389 gl_info
->gl_ops
.gl
.p_glTexImage1D(GL_TEXTURE_1D
, 0, GL_RGB
, 1, 0, GL_BGRA
,
7390 GL_UNSIGNED_INT_8_8_8_8_REV
, &black
);
7393 /* Switch back to unit 0 in which the 2D texture will be stored. */
7394 wined3d_context_gl_active_texture(context_gl
, gl_info
, 0);
7397 /* Context activation is done by the caller. */
7398 static GLuint
gen_yuv_shader(const struct wined3d_gl_info
*gl_info
, const struct arbfp_blit_type
*type
)
7401 struct wined3d_string_buffer buffer
;
7402 char luminance_component
;
7404 if (type
->use_color_key
)
7405 FIXME("Implement YUV color keying.\n");
7408 if (!string_buffer_init(&buffer
))
7410 ERR("Failed to initialize shader buffer.\n");
7414 GL_EXTCALL(glGenProgramsARB(1, &shader
));
7415 checkGLcall("GL_EXTCALL(glGenProgramsARB(1, &shader))");
7416 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7417 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
7420 string_buffer_free(&buffer
);
7424 /* The YUY2 and UYVY formats contain two pixels packed into a 32 bit macropixel,
7425 * giving effectively 16 bit per pixel. The color consists of a luminance(Y) and
7426 * two chroma(U and V) values. Each macropixel has two luminance values, one for
7427 * each single pixel it contains, and one U and one V value shared between both
7430 * The data is loaded into an A8L8 texture. With YUY2, the luminance component
7431 * contains the luminance and alpha the chroma. With UYVY it is vice versa. Thus
7432 * take the format into account when generating the read swizzles
7434 * Reading the Y value is straightforward - just sample the texture. The hardware
7435 * takes care of filtering in the horizontal and vertical direction.
7437 * Reading the U and V values is harder. We have to avoid filtering horizontally,
7438 * because that would mix the U and V values of one pixel or two adjacent pixels.
7439 * Thus floor the texture coordinate and add 0.5 to get an unfiltered read,
7440 * regardless of the filtering setting. Vertical filtering works automatically
7441 * though - the U and V values of two rows are mixed nicely.
7443 * Apart of avoiding filtering issues, the code has to know which value it just
7444 * read, and where it can find the other one. To determine this, it checks if
7445 * it sampled an even or odd pixel, and shifts the 2nd read accordingly.
7447 * Handling horizontal filtering of U and V values requires reading a 2nd pair
7448 * of pixels, extracting U and V and mixing them. This is not implemented yet.
7450 * An alternative implementation idea is to load the texture as A8R8G8B8 texture,
7451 * with width / 2. This way one read gives all 3 values, finding U and V is easy
7452 * in an unfiltered situation. Finding the luminance on the other hand requires
7453 * finding out if it is an odd or even pixel. The real drawback of this approach
7454 * is filtering. This would have to be emulated completely in the shader, reading
7455 * up two 2 packed pixels in up to 2 rows and interpolating both horizontally and
7456 * vertically. Beyond that it would require adjustments to the texture handling
7457 * code to deal with the width scaling
7459 shader_addline(&buffer
, "!!ARBfp1.0\n");
7460 shader_addline(&buffer
, "TEMP luminance;\n");
7461 shader_addline(&buffer
, "TEMP temp;\n");
7462 shader_addline(&buffer
, "TEMP chroma;\n");
7463 shader_addline(&buffer
, "TEMP texcrd;\n");
7464 shader_addline(&buffer
, "TEMP texcrd2;\n");
7465 shader_addline(&buffer
, "PARAM coef = {1.0, 0.5, 2.0, 0.25};\n");
7466 shader_addline(&buffer
, "PARAM yuv_coef = {1.403, 0.344, 0.714, 1.770};\n");
7467 shader_addline(&buffer
, "PARAM size = program.local[%u];\n", ARBFP_BLIT_PARAM_SIZE
);
7469 switch (type
->fixup
)
7471 case COMPLEX_FIXUP_UYVY
:
7472 case COMPLEX_FIXUP_YUY2
:
7473 gen_packed_yuv_read(&buffer
, type
, &luminance_component
);
7476 case COMPLEX_FIXUP_YV12
:
7477 gen_yv12_read(&buffer
, type
, &luminance_component
);
7480 case COMPLEX_FIXUP_NV12
:
7481 gen_nv12_read(&buffer
, type
, &luminance_component
);
7485 FIXME("Unsupported YUV fixup %#x\n", type
->fixup
);
7486 string_buffer_free(&buffer
);
7490 /* Calculate the final result. Formula is taken from
7491 * http://www.fourcc.org/fccyvrgb.php. Note that the chroma
7492 * ranges from -0.5 to 0.5
7494 shader_addline(&buffer
, "SUB chroma.xy, chroma, coef.y;\n");
7496 shader_addline(&buffer
, "MAD result.color.x, chroma.x, yuv_coef.x, luminance.%c;\n", luminance_component
);
7497 shader_addline(&buffer
, "MAD temp.x, -chroma.y, yuv_coef.y, luminance.%c;\n", luminance_component
);
7498 shader_addline(&buffer
, "MAD result.color.y, -chroma.x, yuv_coef.z, temp.x;\n");
7499 shader_addline(&buffer
, "MAD result.color.z, chroma.y, yuv_coef.w, luminance.%c;\n", luminance_component
);
7500 shader_addline(&buffer
, "END\n");
7502 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
7504 string_buffer_free(&buffer
);
7509 /* Context activation is done by the caller. */
7510 static GLuint
arbfp_gen_plain_shader(const struct wined3d_gl_info
*gl_info
, const struct arbfp_blit_type
*type
)
7513 struct wined3d_string_buffer buffer
;
7514 const char *tex_target
= arbfp_texture_target(type
->res_type
);
7517 if (!string_buffer_init(&buffer
))
7519 ERR("Failed to initialize shader buffer.\n");
7523 GL_EXTCALL(glGenProgramsARB(1, &shader
));
7526 string_buffer_free(&buffer
);
7529 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7531 shader_addline(&buffer
, "!!ARBfp1.0\n");
7533 if (type
->use_color_key
)
7535 shader_addline(&buffer
, "TEMP color;\n");
7536 shader_addline(&buffer
, "TEMP less, greater;\n");
7537 shader_addline(&buffer
, "PARAM color_key_low = program.local[%u];\n", ARBFP_BLIT_PARAM_COLOR_KEY_LOW
);
7538 shader_addline(&buffer
, "PARAM color_key_high = program.local[%u];\n", ARBFP_BLIT_PARAM_COLOR_KEY_HIGH
);
7539 shader_addline(&buffer
, "TEX color, fragment.texcoord[0], texture[0], %s;\n", tex_target
);
7540 shader_addline(&buffer
, "SLT less, color, color_key_low;\n"); /* below low key */
7541 shader_addline(&buffer
, "SGE greater, color, color_key_high;\n"); /* above high key */
7542 shader_addline(&buffer
, "ADD less, less, greater;\n"); /* or */
7543 shader_addline(&buffer
, "DP4 less.b, less, less;\n"); /* on any channel */
7544 shader_addline(&buffer
, "SGE less, -less.b, 0.0;\n"); /* logical not */
7545 shader_addline(&buffer
, "KIL -less;\n"); /* discard if true */
7546 shader_addline(&buffer
, "MOV result.color, color;\n");
7550 shader_addline(&buffer
, "TEX result.color, fragment.texcoord[0], texture[0], %s;\n", tex_target
);
7553 shader_addline(&buffer
, "END\n");
7555 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
7557 string_buffer_free(&buffer
);
7562 /* Context activation is done by the caller. */
7563 static HRESULT
arbfp_blit_set(struct wined3d_arbfp_blitter
*blitter
, struct wined3d_context_gl
*context_gl
,
7564 const struct wined3d_texture_gl
*texture_gl
, unsigned int sub_resource_idx
,
7565 const struct wined3d_color_key
*color_key
)
7567 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
7568 enum complex_fixup fixup
;
7569 struct wine_rb_entry
*entry
;
7570 struct arbfp_blit_type type
;
7571 struct arbfp_blit_desc
*desc
;
7572 struct wined3d_color float_color_key
[2];
7573 struct wined3d_vec4 size
;
7577 level
= sub_resource_idx
% texture_gl
->t
.level_count
;
7578 size
.x
= wined3d_texture_get_level_pow2_width(&texture_gl
->t
, level
);
7579 size
.y
= wined3d_texture_get_level_pow2_height(&texture_gl
->t
, level
);
7583 if (is_complex_fixup(texture_gl
->t
.resource
.format
->color_fixup
))
7584 fixup
= get_complex_fixup(texture_gl
->t
.resource
.format
->color_fixup
);
7586 fixup
= COMPLEX_FIXUP_NONE
;
7588 switch (texture_gl
->target
)
7591 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_1D
;
7595 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_2D
;
7599 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_3D
;
7602 case GL_TEXTURE_CUBE_MAP_ARB
:
7603 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_CUBE
;
7606 case GL_TEXTURE_RECTANGLE_ARB
:
7607 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_RECT
;
7611 ERR("Unexpected GL texture type %#x.\n", texture_gl
->target
);
7612 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_2D
;
7615 type
.use_color_key
= !!color_key
;
7618 if ((entry
= wine_rb_get(&blitter
->shaders
, &type
)))
7620 desc
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_blit_desc
, entry
);
7621 shader
= desc
->shader
;
7627 case COMPLEX_FIXUP_NONE
:
7628 if (!is_identity_fixup(texture_gl
->t
.resource
.format
->color_fixup
))
7629 FIXME("Implement support for sign or swizzle fixups.\n");
7630 shader
= arbfp_gen_plain_shader(gl_info
, &type
);
7633 case COMPLEX_FIXUP_P8
:
7634 shader
= gen_p8_shader(gl_info
, &type
);
7637 case COMPLEX_FIXUP_YUY2
:
7638 case COMPLEX_FIXUP_UYVY
:
7639 case COMPLEX_FIXUP_YV12
:
7640 case COMPLEX_FIXUP_NV12
:
7641 shader
= gen_yuv_shader(gl_info
, &type
);
7645 FIXME("Unsupported fixup %#x.\n", fixup
);
7651 ERR("Failed to get shader for fixup %#x.\n", fixup
);
7655 if (!(desc
= heap_alloc(sizeof(*desc
))))
7659 desc
->shader
= shader
;
7660 if (wine_rb_put(&blitter
->shaders
, &desc
->type
, &desc
->entry
) == -1)
7663 ERR("Out of memory\n");
7664 GL_EXTCALL(glDeleteProgramsARB(1, &shader
));
7665 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader))");
7666 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, 0));
7667 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, 0)");
7669 return E_OUTOFMEMORY
;
7673 if (fixup
== COMPLEX_FIXUP_P8
)
7674 arbfp_blitter_upload_palette(blitter
, texture_gl
, context_gl
);
7676 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
7677 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
7678 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7679 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
7680 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARBFP_BLIT_PARAM_SIZE
, &size
.x
));
7681 checkGLcall("glProgramLocalParameter4fvARB");
7682 if (type
.use_color_key
)
7684 wined3d_format_get_float_color_key(texture_gl
->t
.resource
.format
, color_key
, float_color_key
);
7685 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
7686 ARBFP_BLIT_PARAM_COLOR_KEY_LOW
, &float_color_key
[0].r
));
7687 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
7688 ARBFP_BLIT_PARAM_COLOR_KEY_HIGH
, &float_color_key
[1].r
));
7689 checkGLcall("glProgramLocalParameter4fvARB");
7695 /* Context activation is done by the caller. */
7696 static void arbfp_blit_unset(const struct wined3d_gl_info
*gl_info
)
7698 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
7699 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
7702 static BOOL
arbfp_blit_supported(enum wined3d_blit_op blit_op
, const struct wined3d_context
*context
,
7703 const struct wined3d_resource
*src_resource
, DWORD src_location
,
7704 const struct wined3d_resource
*dst_resource
, DWORD dst_location
)
7706 const struct wined3d_format
*src_format
= src_resource
->format
;
7707 const struct wined3d_format
*dst_format
= dst_resource
->format
;
7708 enum complex_fixup src_fixup
;
7711 if (src_resource
->type
!= WINED3D_RTYPE_TEXTURE_2D
)
7714 if (blit_op
== WINED3D_BLIT_OP_RAW_BLIT
&& dst_format
->id
== src_format
->id
)
7716 if (dst_format
->depth_size
|| dst_format
->stencil_size
)
7717 blit_op
= WINED3D_BLIT_OP_DEPTH_BLIT
;
7719 blit_op
= WINED3D_BLIT_OP_COLOR_BLIT
;
7724 case WINED3D_BLIT_OP_COLOR_BLIT_CKEY
:
7725 if (!context
->d3d_info
->shader_color_key
)
7727 /* The conversion modifies the alpha channel so the color key might no longer match. */
7728 TRACE("Color keying not supported with converted textures.\n");
7731 case WINED3D_BLIT_OP_COLOR_BLIT_ALPHATEST
:
7732 case WINED3D_BLIT_OP_COLOR_BLIT
:
7736 TRACE("Unsupported blit_op=%d\n", blit_op
);
7740 decompress
= (src_format
->attrs
& WINED3D_FORMAT_ATTR_COMPRESSED
)
7741 && !(dst_format
->attrs
& WINED3D_FORMAT_ATTR_COMPRESSED
);
7742 if (!decompress
&& !(src_resource
->access
& dst_resource
->access
& WINED3D_RESOURCE_ACCESS_GPU
))
7745 src_fixup
= get_complex_fixup(src_format
->color_fixup
);
7746 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
7748 TRACE("Checking support for fixup:\n");
7749 dump_color_fixup_desc(src_format
->color_fixup
);
7752 if (!is_identity_fixup(dst_format
->color_fixup
)
7753 && (dst_format
->id
!= src_format
->id
|| dst_location
!= WINED3D_LOCATION_DRAWABLE
))
7755 TRACE("Destination fixups are not supported\n");
7759 if (is_identity_fixup(src_format
->color_fixup
))
7765 /* We only support YUV conversions. */
7766 if (!is_complex_fixup(src_format
->color_fixup
))
7768 if (wined3d_settings
.offscreen_rendering_mode
== ORM_BACKBUFFER
)
7770 WARN("Claiming fixup support because of ORM_BACKBUFFER.\n");
7774 TRACE("[FAILED]\n");
7780 case COMPLEX_FIXUP_YUY2
:
7781 case COMPLEX_FIXUP_UYVY
:
7782 case COMPLEX_FIXUP_YV12
:
7783 case COMPLEX_FIXUP_NV12
:
7784 case COMPLEX_FIXUP_P8
:
7789 FIXME("Unsupported YUV fixup %#x\n", src_fixup
);
7790 TRACE("[FAILED]\n");
7795 static DWORD
arbfp_blitter_blit(struct wined3d_blitter
*blitter
, enum wined3d_blit_op op
,
7796 struct wined3d_context
*context
, struct wined3d_texture
*src_texture
, unsigned int src_sub_resource_idx
,
7797 DWORD src_location
, const RECT
*src_rect
, struct wined3d_texture
*dst_texture
,
7798 unsigned int dst_sub_resource_idx
, DWORD dst_location
, const RECT
*dst_rect
,
7799 const struct wined3d_color_key
*color_key
, enum wined3d_texture_filter_type filter
,
7800 const struct wined3d_format
*resolve_format
)
7802 struct wined3d_texture_gl
*src_texture_gl
= wined3d_texture_gl(src_texture
);
7803 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
7804 struct wined3d_device
*device
= dst_texture
->resource
.device
;
7805 struct wined3d_texture
*staging_texture
= NULL
;
7806 struct wined3d_arbfp_blitter
*arbfp_blitter
;
7807 struct wined3d_color_key alpha_test_key
;
7808 struct wined3d_blitter
*next
;
7809 unsigned int src_level
;
7812 TRACE("blitter %p, op %#x, context %p, src_texture %p, src_sub_resource_idx %u, src_location %s, "
7813 "src_rect %s, dst_texture %p, dst_sub_resource_idx %u, dst_location %s, dst_rect %s, "
7814 "colour_key %p, filter %s, resolve format %p.\n",
7815 blitter
, op
, context
, src_texture
, src_sub_resource_idx
, wined3d_debug_location(src_location
),
7816 wine_dbgstr_rect(src_rect
), dst_texture
, dst_sub_resource_idx
, wined3d_debug_location(dst_location
),
7817 wine_dbgstr_rect(dst_rect
), color_key
, debug_d3dtexturefiltertype(filter
), resolve_format
);
7819 if (!arbfp_blit_supported(op
, context
, &src_texture
->resource
, src_location
,
7820 &dst_texture
->resource
, dst_location
))
7822 if (!(next
= blitter
->next
))
7824 ERR("No blitter to handle blit op %#x.\n", op
);
7825 return dst_location
;
7828 TRACE("Forwarding to blitter %p.\n", next
);
7829 return next
->ops
->blitter_blit(next
, op
, context
, src_texture
, src_sub_resource_idx
, src_location
,
7830 src_rect
, dst_texture
, dst_sub_resource_idx
, dst_location
, dst_rect
, color_key
, filter
, resolve_format
);
7833 arbfp_blitter
= CONTAINING_RECORD(blitter
, struct wined3d_arbfp_blitter
, blitter
);
7835 if (!(src_texture
->resource
.access
& WINED3D_RESOURCE_ACCESS_GPU
))
7837 struct wined3d_resource_desc desc
;
7838 struct wined3d_box upload_box
;
7841 TRACE("Source texture is not GPU accessible, creating a staging texture.\n");
7843 src_level
= src_sub_resource_idx
% src_texture
->level_count
;
7844 desc
.resource_type
= WINED3D_RTYPE_TEXTURE_2D
;
7845 desc
.format
= src_texture
->resource
.format
->id
;
7846 desc
.multisample_type
= src_texture
->resource
.multisample_type
;
7847 desc
.multisample_quality
= src_texture
->resource
.multisample_quality
;
7848 desc
.usage
= WINED3DUSAGE_PRIVATE
;
7849 desc
.bind_flags
= 0;
7850 desc
.access
= WINED3D_RESOURCE_ACCESS_GPU
;
7851 desc
.width
= wined3d_texture_get_level_width(src_texture
, src_level
);
7852 desc
.height
= wined3d_texture_get_level_height(src_texture
, src_level
);
7856 if (FAILED(hr
= wined3d_texture_create(device
, &desc
, 1, 1, 0,
7857 NULL
, NULL
, &wined3d_null_parent_ops
, &staging_texture
)))
7859 ERR("Failed to create staging texture, hr %#lx.\n", hr
);
7860 return dst_location
;
7863 wined3d_box_set(&upload_box
, 0, 0, desc
.width
, desc
.height
, 0, desc
.depth
);
7864 wined3d_texture_upload_from_texture(staging_texture
, 0, 0, 0, 0,
7865 src_texture
, src_sub_resource_idx
, &upload_box
);
7867 src_texture
= staging_texture
;
7868 src_texture_gl
= wined3d_texture_gl(src_texture
);
7869 src_sub_resource_idx
= 0;
7871 else if (wined3d_settings
.offscreen_rendering_mode
!= ORM_FBO
7872 && (src_texture
->sub_resources
[src_sub_resource_idx
].locations
7873 & (WINED3D_LOCATION_TEXTURE_RGB
| WINED3D_LOCATION_DRAWABLE
)) == WINED3D_LOCATION_DRAWABLE
7874 && !wined3d_resource_is_offscreen(&src_texture
->resource
))
7877 /* Without FBO blits transferring from the drawable to the texture is
7878 * expensive, because we have to flip the data in sysmem. Since we can
7879 * flip in the blitter, we don't actually need that flip anyway. So we
7880 * use the surface's texture as scratch texture, and flip the source
7881 * rectangle instead. */
7882 texture2d_load_fb_texture(src_texture_gl
, src_sub_resource_idx
, FALSE
, context
);
7885 src_level
= src_sub_resource_idx
% src_texture
->level_count
;
7886 s
.top
= wined3d_texture_get_level_height(src_texture
, src_level
) - s
.top
;
7887 s
.bottom
= wined3d_texture_get_level_height(src_texture
, src_level
) - s
.bottom
;
7892 wined3d_texture_load(src_texture
, context
, FALSE
);
7895 wined3d_context_gl_apply_ffp_blit_state(context_gl
, device
);
7897 if (dst_location
== WINED3D_LOCATION_DRAWABLE
)
7900 wined3d_texture_translate_drawable_coords(dst_texture
, context_gl
->window
, &d
);
7904 context_gl_apply_texture_draw_state(context_gl
, dst_texture
, dst_sub_resource_idx
, dst_location
);
7906 if (op
== WINED3D_BLIT_OP_COLOR_BLIT_ALPHATEST
)
7908 const struct wined3d_format
*fmt
= src_texture
->resource
.format
;
7909 alpha_test_key
.color_space_low_value
= 0;
7910 alpha_test_key
.color_space_high_value
= ~(wined3d_mask_from_size(fmt
->alpha_size
) << fmt
->alpha_offset
);
7911 color_key
= &alpha_test_key
;
7914 arbfp_blit_set(arbfp_blitter
, context_gl
, src_texture_gl
, src_sub_resource_idx
, color_key
);
7916 /* Draw a textured quad */
7917 wined3d_context_gl_draw_textured_quad(context_gl
, src_texture_gl
,
7918 src_sub_resource_idx
, src_rect
, dst_rect
, filter
);
7920 /* Leave the opengl state valid for blitting */
7921 arbfp_blit_unset(context_gl
->gl_info
);
7923 if (dst_texture
->swapchain
&& (dst_texture
->swapchain
->front_buffer
== dst_texture
))
7924 context_gl
->gl_info
->gl_ops
.gl
.p_glFlush();
7926 if (staging_texture
)
7927 wined3d_texture_decref(staging_texture
);
7929 return dst_location
;
7932 static void arbfp_blitter_clear(struct wined3d_blitter
*blitter
, struct wined3d_device
*device
,
7933 unsigned int rt_count
, const struct wined3d_fb_state
*fb
, unsigned int rect_count
, const RECT
*clear_rects
,
7934 const RECT
*draw_rect
, uint32_t flags
, const struct wined3d_color
*colour
, float depth
, unsigned int stencil
)
7936 struct wined3d_blitter
*next
;
7938 if ((next
= blitter
->next
))
7939 next
->ops
->blitter_clear(next
, device
, rt_count
, fb
, rect_count
,
7940 clear_rects
, draw_rect
, flags
, colour
, depth
, stencil
);
7943 static const struct wined3d_blitter_ops arbfp_blitter_ops
=
7945 arbfp_blitter_destroy
,
7946 arbfp_blitter_clear
,
7950 void wined3d_arbfp_blitter_create(struct wined3d_blitter
**next
, const struct wined3d_device
*device
)
7952 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
7953 struct wined3d_arbfp_blitter
*blitter
;
7955 if (device
->shader_backend
!= &arb_program_shader_backend
7956 && device
->shader_backend
!= &glsl_shader_backend
)
7959 if (!gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
7962 if (!gl_info
->supported
[WINED3D_GL_LEGACY_CONTEXT
])
7965 if (!(blitter
= heap_alloc(sizeof(*blitter
))))
7967 ERR("Failed to allocate blitter.\n");
7971 TRACE("Created blitter %p.\n", blitter
);
7973 blitter
->blitter
.ops
= &arbfp_blitter_ops
;
7974 blitter
->blitter
.next
= *next
;
7975 wine_rb_init(&blitter
->shaders
, arbfp_blit_type_compare
);
7976 blitter
->palette_texture
= 0;
7977 *next
= &blitter
->blitter
;