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 2007-2008 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
35 #include "wined3d_private.h"
37 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader
);
38 WINE_DECLARE_DEBUG_CHANNEL(d3d_constants
);
39 WINE_DECLARE_DEBUG_CHANNEL(d3d_caps
);
40 WINE_DECLARE_DEBUG_CHANNEL(d3d
);
42 /* Extract a line. Note that this modifies the source string. */
43 static char *get_line(char **ptr
)
48 if (!(q
= strstr(p
, "\n")))
60 static void shader_arb_dump_program_source(const char *source
)
63 char *ptr
, *line
, *tmp
;
65 source_size
= strlen(source
) + 1;
66 tmp
= HeapAlloc(GetProcessHeap(), 0, source_size
);
69 ERR("Failed to allocate %u bytes for shader source.\n", source_size
);
72 memcpy(tmp
, source
, source_size
);
75 while ((line
= get_line(&ptr
))) FIXME(" %s\n", line
);
78 HeapFree(GetProcessHeap(), 0, tmp
);
92 static const char *arb_get_helper_value(enum wined3d_shader_type shader
, enum arb_helper_value value
)
94 if (shader
== WINED3D_SHADER_TYPE_GEOMETRY
)
96 ERR("Geometry shaders are unsupported\n");
100 if (shader
== WINED3D_SHADER_TYPE_PIXEL
)
104 case ARB_ZERO
: return "ps_helper_const.x";
105 case ARB_ONE
: return "ps_helper_const.y";
106 case ARB_TWO
: return "coefmul.x";
107 case ARB_0001
: return "ps_helper_const.xxxy";
108 case ARB_EPS
: return "ps_helper_const.z";
116 case ARB_ZERO
: return "helper_const.x";
117 case ARB_ONE
: return "helper_const.y";
118 case ARB_TWO
: return "helper_const.z";
119 case ARB_EPS
: return "helper_const.w";
120 case ARB_0001
: return "helper_const.xxxy";
121 case ARB_VS_REL_OFFSET
: return "rel_addr_const.y";
124 FIXME("Unmanaged %s shader helper constant requested: %u\n",
125 shader
== WINED3D_SHADER_TYPE_PIXEL
? "pixel" : "vertex", value
);
128 case ARB_ZERO
: return "0.0";
129 case ARB_ONE
: return "1.0";
130 case ARB_TWO
: return "2.0";
131 case ARB_0001
: return "{0.0, 0.0, 0.0, 1.0}";
132 case ARB_EPS
: return "1e-8";
133 default: return "bad";
137 static inline BOOL
ffp_clip_emul(const struct wined3d_state
*state
)
139 return state
->lowest_disabled_stage
< 7;
142 /* ARB_program_shader private data */
161 struct wined3d_shader_loop_control loop_control
;
165 struct arb_ps_np2fixup_info
167 struct ps_np2fixup_info super
;
168 /* For ARB we need a offset value:
169 * With both GLSL and ARB mode the NP2 fixup information (the texture dimensions) are stored in a
170 * consecutive way (GLSL uses a uniform array). Since ARB doesn't know the notion of a "standalone"
171 * array we need an offset to the index inside the program local parameter array. */
175 struct arb_ps_compile_args
177 struct ps_compile_args super
;
179 WORD clip
; /* only a boolean, use a WORD for alignment */
180 unsigned char loop_ctrl
[MAX_CONST_I
][3];
183 struct stb_const_desc
185 unsigned char texunit
;
189 struct arb_ps_compiled_shader
191 struct arb_ps_compile_args args
;
192 struct arb_ps_np2fixup_info np2fixup_info
;
193 struct stb_const_desc bumpenvmatconst
[MAX_TEXTURES
];
194 struct stb_const_desc luminanceconst
[MAX_TEXTURES
];
195 UINT int_consts
[MAX_CONST_I
];
198 unsigned char numbumpenvmatconsts
;
202 struct arb_vs_compile_args
204 struct vs_compile_args super
;
210 unsigned char clip_texcoord
;
211 unsigned char clipplane_mask
;
213 DWORD boolclip_compare
;
218 unsigned char samplers
[4];
219 DWORD samplers_compare
;
221 unsigned char loop_ctrl
[MAX_CONST_I
][3];
224 struct arb_vs_compiled_shader
226 struct arb_vs_compile_args args
;
228 UINT int_consts
[MAX_CONST_I
];
230 char need_color_unclamp
;
234 struct recorded_instruction
236 struct wined3d_shader_instruction ins
;
240 struct shader_arb_ctx_priv
245 /* plain GL_ARB_vertex_program or GL_ARB_fragment_program */
247 /* GL_NV_vertex_progam2_option or GL_NV_fragment_program_option */
249 /* GL_NV_vertex_program3 or GL_NV_fragment_program2 */
253 const struct arb_vs_compile_args
*cur_vs_args
;
254 const struct arb_ps_compile_args
*cur_ps_args
;
255 const struct arb_ps_compiled_shader
*compiled_fprog
;
256 const struct arb_vs_compiled_shader
*compiled_vprog
;
257 struct arb_ps_np2fixup_info
*cur_np2fixup_info
;
258 struct list control_frames
;
262 unsigned int num_loops
, loop_depth
, num_ifcs
;
265 unsigned int vs_clipplanes
;
269 /* For 3.0 vertex shaders */
270 const char *vs_output
[MAX_REG_OUTPUT
];
271 /* For 2.x and earlier vertex shaders */
272 const char *texcrd_output
[8], *color_output
[2], *fog_output
;
274 /* 3.0 pshader input for compatibility with fixed function */
275 const char *ps_input
[MAX_REG_INPUT
];
280 struct wined3d_shader_signature_element
*sig
;
282 struct wine_rb_entry entry
;
285 struct arb_pshader_private
{
286 struct arb_ps_compiled_shader
*gl_shaders
;
287 UINT num_gl_shaders
, shader_array_size
;
288 BOOL has_signature_idx
;
289 DWORD input_signature_idx
;
290 DWORD clipplane_emulation
;
294 struct arb_vshader_private
{
295 struct arb_vs_compiled_shader
*gl_shaders
;
296 UINT num_gl_shaders
, shader_array_size
;
300 struct shader_arb_priv
302 GLuint current_vprogram_id
;
303 GLuint current_fprogram_id
;
304 const struct arb_ps_compiled_shader
*compiled_fprog
;
305 const struct arb_vs_compiled_shader
*compiled_vprog
;
306 GLuint depth_blt_vprogram_id
;
307 GLuint depth_blt_fprogram_id_full
[tex_type_count
];
308 GLuint depth_blt_fprogram_id_masked
[tex_type_count
];
309 BOOL use_arbfp_fixed_func
;
310 struct wine_rb_tree fragment_shaders
;
311 BOOL last_ps_const_clamped
;
312 BOOL last_vs_color_unclamp
;
314 struct wine_rb_tree signature_tree
;
318 /* GL locking for state handlers is done by the caller. */
319 static BOOL
need_rel_addr_const(const struct arb_vshader_private
*shader_data
,
320 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
)
322 if (shader_data
->rel_offset
) return TRUE
;
323 if (!reg_maps
->usesmova
) return FALSE
;
324 return !gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
];
327 /* Returns TRUE if result.clip from GL_NV_vertex_program2 should be used and FALSE otherwise */
328 static inline BOOL
use_nv_clip(const struct wined3d_gl_info
*gl_info
)
330 return gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]
331 && !(gl_info
->quirks
& WINED3D_QUIRK_NV_CLIP_BROKEN
);
334 static BOOL
need_helper_const(const struct arb_vshader_private
*shader_data
,
335 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
)
337 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
)) return TRUE
;
338 if (!gl_info
->supported
[NV_VERTEX_PROGRAM
]) return TRUE
; /* Need to init colors. */
339 if (gl_info
->quirks
& WINED3D_QUIRK_ARB_VS_OFFSET_LIMIT
) return TRUE
; /* Load the immval offset. */
340 if (gl_info
->quirks
& WINED3D_QUIRK_SET_TEXCOORD_W
) return TRUE
; /* Have to init texcoords. */
341 if (!use_nv_clip(gl_info
)) return TRUE
; /* Init the clip texcoord */
342 if (reg_maps
->usesnrm
) return TRUE
; /* 0.0 */
343 if (reg_maps
->usespow
) return TRUE
; /* EPS, 0.0 and 1.0 */
347 static unsigned int reserved_vs_const(const struct arb_vshader_private
*shader_data
,
348 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
)
350 unsigned int ret
= 1;
351 /* We use one PARAM for the pos fixup, and in some cases one to load
352 * some immediate values into the shader. */
353 if (need_helper_const(shader_data
, reg_maps
, gl_info
)) ++ret
;
354 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
)) ++ret
;
358 /* Loads floating point constants into the currently set ARB_vertex/fragment_program.
359 * When constant_list == NULL, it will load all the constants.
361 * @target_type should be either GL_VERTEX_PROGRAM_ARB (for vertex shaders)
362 * or GL_FRAGMENT_PROGRAM_ARB (for pixel shaders)
364 /* GL locking is done by the caller */
365 static unsigned int shader_arb_load_constantsF(struct wined3d_shader
*shader
, const struct wined3d_gl_info
*gl_info
,
366 GLuint target_type
, unsigned int max_constants
, const float *constants
, char *dirty_consts
)
368 local_constant
* lconst
;
372 if (TRACE_ON(d3d_constants
))
374 for(i
= 0; i
< max_constants
; i
++) {
375 if(!dirty_consts
[i
]) continue;
376 TRACE_(d3d_constants
)("Loading constants %i: %f, %f, %f, %f\n", i
,
377 constants
[i
* 4 + 0], constants
[i
* 4 + 1],
378 constants
[i
* 4 + 2], constants
[i
* 4 + 3]);
384 /* In 1.X pixel shaders constants are implicitly clamped in the range [-1;1] */
385 if (target_type
== GL_FRAGMENT_PROGRAM_ARB
&& shader
->reg_maps
.shader_version
.major
== 1)
388 /* ps 1.x supports only 8 constants, clamp only those. When switching between 1.x and higher
389 * shaders, the first 8 constants are marked dirty for reload
391 for(; i
< min(8, max_constants
); i
++) {
392 if(!dirty_consts
[i
]) continue;
396 if (constants
[j
+ 0] > 1.0f
) lcl_const
[0] = 1.0f
;
397 else if (constants
[j
+ 0] < -1.0f
) lcl_const
[0] = -1.0f
;
398 else lcl_const
[0] = constants
[j
+ 0];
400 if (constants
[j
+ 1] > 1.0f
) lcl_const
[1] = 1.0f
;
401 else if (constants
[j
+ 1] < -1.0f
) lcl_const
[1] = -1.0f
;
402 else lcl_const
[1] = constants
[j
+ 1];
404 if (constants
[j
+ 2] > 1.0f
) lcl_const
[2] = 1.0f
;
405 else if (constants
[j
+ 2] < -1.0f
) lcl_const
[2] = -1.0f
;
406 else lcl_const
[2] = constants
[j
+ 2];
408 if (constants
[j
+ 3] > 1.0f
) lcl_const
[3] = 1.0f
;
409 else if (constants
[j
+ 3] < -1.0f
) lcl_const
[3] = -1.0f
;
410 else lcl_const
[3] = constants
[j
+ 3];
412 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, i
, lcl_const
));
415 /* If further constants are dirty, reload them without clamping.
417 * The alternative is not to touch them, but then we cannot reset the dirty constant count
418 * to zero. That's bad for apps that only use PS 1.x shaders, because in that case the code
419 * above would always re-check the first 8 constants since max_constant remains at the init
424 if (gl_info
->supported
[EXT_GPU_PROGRAM_PARAMETERS
])
426 /* TODO: Benchmark if we're better of with finding the dirty constants ourselves,
427 * or just reloading *all* constants at once
429 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type, i, max_constants, constants + (i * 4)));
431 for(; i
< max_constants
; i
++) {
432 if(!dirty_consts
[i
]) continue;
434 /* Find the next block of dirty constants */
437 for(i
++; (i
< max_constants
) && dirty_consts
[i
]; i
++) {
441 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type
, j
, i
- j
, constants
+ (j
* 4)));
444 for(; i
< max_constants
; i
++) {
445 if(dirty_consts
[i
]) {
447 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, i
, constants
+ (i
* 4)));
451 checkGLcall("glProgramEnvParameter4fvARB()");
453 /* Load immediate constants */
454 if (shader
->load_local_constsF
)
456 if (TRACE_ON(d3d_shader
))
458 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, local_constant
, entry
)
460 GLfloat
* values
= (GLfloat
*)lconst
->value
;
461 TRACE_(d3d_constants
)("Loading local constants %i: %f, %f, %f, %f\n", lconst
->idx
,
462 values
[0], values
[1], values
[2], values
[3]);
465 /* Immediate constants are clamped for 1.X shaders at loading times */
467 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, local_constant
, entry
)
469 dirty_consts
[lconst
->idx
] = 1; /* Dirtify so the non-immediate constant overwrites it next time */
470 ret
= max(ret
, lconst
->idx
+ 1);
471 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, lconst
->idx
, (GLfloat
*)lconst
->value
));
473 checkGLcall("glProgramEnvParameter4fvARB()");
474 return ret
; /* The loaded immediate constants need reloading for the next shader */
476 return 0; /* No constants are dirty now */
481 * Loads the texture dimensions for NP2 fixup into the currently set ARB_[vertex/fragment]_programs.
483 static void shader_arb_load_np2fixup_constants(void *shader_priv
,
484 const struct wined3d_gl_info
*gl_info
, const struct wined3d_state
*state
)
486 const struct shader_arb_priv
* priv
= shader_priv
;
488 /* NP2 texcoord fixup is (currently) only done for pixelshaders. */
489 if (!use_ps(state
)) return;
491 if (priv
->compiled_fprog
&& priv
->compiled_fprog
->np2fixup_info
.super
.active
) {
492 const struct arb_ps_np2fixup_info
* const fixup
= &priv
->compiled_fprog
->np2fixup_info
;
494 WORD active
= fixup
->super
.active
;
495 GLfloat np2fixup_constants
[4 * MAX_FRAGMENT_SAMPLERS
];
497 for (i
= 0; active
; active
>>= 1, ++i
)
499 const struct wined3d_texture
*tex
= state
->textures
[i
];
500 const unsigned char idx
= fixup
->super
.idx
[i
];
501 GLfloat
*tex_dim
= &np2fixup_constants
[(idx
>> 1) * 4];
503 if (!(active
& 1)) continue;
506 FIXME("Nonexistent texture is flagged for NP2 texcoord fixup\n");
512 tex_dim
[2] = tex
->pow2_matrix
[0];
513 tex_dim
[3] = tex
->pow2_matrix
[5];
517 tex_dim
[0] = tex
->pow2_matrix
[0];
518 tex_dim
[1] = tex
->pow2_matrix
[5];
522 for (i
= 0; i
< fixup
->super
.num_consts
; ++i
) {
523 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
524 fixup
->offset
+ i
, &np2fixup_constants
[i
* 4]));
529 /* GL locking is done by the caller. */
530 static void shader_arb_ps_local_constants(const struct arb_ps_compiled_shader
*gl_shader
,
531 const struct wined3d_context
*context
, const struct wined3d_state
*state
, UINT rt_height
)
533 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
536 for(i
= 0; i
< gl_shader
->numbumpenvmatconsts
; i
++)
538 int texunit
= gl_shader
->bumpenvmatconst
[i
].texunit
;
540 /* The state manager takes care that this function is always called if the bump env matrix changes */
541 const float *data
= (const float *)&state
->texture_states
[texunit
][WINED3DTSS_BUMPENVMAT00
];
542 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
543 gl_shader
->bumpenvmatconst
[i
].const_num
, data
));
545 if (gl_shader
->luminanceconst
[i
].const_num
!= WINED3D_CONST_NUM_UNUSED
)
547 /* WINED3DTSS_BUMPENVLSCALE and WINED3DTSS_BUMPENVLOFFSET are next to each other.
548 * point gl to the scale, and load 4 floats. x = scale, y = offset, z and w are junk, we
549 * don't care about them. The pointers are valid for sure because the stateblock is bigger.
550 * (they're WINED3DTSS_TEXTURETRANSFORMFLAGS and WINED3DTSS_ADDRESSW, so most likely 0 or NaN
552 const float *scale
= (const float *)&state
->texture_states
[texunit
][WINED3DTSS_BUMPENVLSCALE
];
553 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
554 gl_shader
->luminanceconst
[i
].const_num
, scale
));
557 checkGLcall("Load bumpmap consts");
559 if(gl_shader
->ycorrection
!= WINED3D_CONST_NUM_UNUSED
)
561 /* ycorrection.x: Backbuffer height(onscreen) or 0(offscreen).
562 * ycorrection.y: -1.0(onscreen), 1.0(offscreen)
567 val
[0] = context
->render_offscreen
? 0.0f
: (float) rt_height
;
568 val
[1] = context
->render_offscreen
? 1.0f
: -1.0f
;
571 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, gl_shader
->ycorrection
, val
));
572 checkGLcall("y correction loading");
575 if (!gl_shader
->num_int_consts
) return;
577 for(i
= 0; i
< MAX_CONST_I
; i
++)
579 if(gl_shader
->int_consts
[i
] != WINED3D_CONST_NUM_UNUSED
)
582 val
[0] = (float)state
->ps_consts_i
[4 * i
];
583 val
[1] = (float)state
->ps_consts_i
[4 * i
+ 1];
584 val
[2] = (float)state
->ps_consts_i
[4 * i
+ 2];
587 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, gl_shader
->int_consts
[i
], val
));
590 checkGLcall("Load ps int consts");
593 /* GL locking is done by the caller. */
594 static void shader_arb_vs_local_constants(const struct arb_vs_compiled_shader
*gl_shader
,
595 const struct wined3d_context
*context
, const struct wined3d_state
*state
)
597 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
598 float position_fixup
[4];
601 /* Upload the position fixup */
602 shader_get_position_fixup(context
, state
, position_fixup
);
603 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, gl_shader
->pos_fixup
, position_fixup
));
605 if (!gl_shader
->num_int_consts
) return;
607 for(i
= 0; i
< MAX_CONST_I
; i
++)
609 if(gl_shader
->int_consts
[i
] != WINED3D_CONST_NUM_UNUSED
)
612 val
[0] = (float)state
->vs_consts_i
[4 * i
];
613 val
[1] = (float)state
->vs_consts_i
[4 * i
+ 1];
614 val
[2] = (float)state
->vs_consts_i
[4 * i
+ 2];
617 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, gl_shader
->int_consts
[i
], val
));
620 checkGLcall("Load vs int consts");
624 * Loads the app-supplied constants into the currently set ARB_[vertex/fragment]_programs.
626 * We only support float constants in ARB at the moment, so don't
627 * worry about the Integers or Booleans
629 /* GL locking is done by the caller (state handler) */
630 static void shader_arb_load_constants(const struct wined3d_context
*context
, char usePixelShader
, char useVertexShader
)
632 struct wined3d_device
*device
= context
->swapchain
->device
;
633 struct wined3d_stateblock
*stateBlock
= device
->stateBlock
;
634 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
635 struct shader_arb_priv
*priv
= device
->shader_priv
;
639 struct wined3d_shader
*vshader
= stateBlock
->state
.vertex_shader
;
640 const struct arb_vs_compiled_shader
*gl_shader
= priv
->compiled_vprog
;
642 /* Load DirectX 9 float constants for vertex shader */
643 device
->highest_dirty_vs_const
= shader_arb_load_constantsF(vshader
, gl_info
, GL_VERTEX_PROGRAM_ARB
,
644 device
->highest_dirty_vs_const
, stateBlock
->state
.vs_consts_f
, context
->vshader_const_dirty
);
645 shader_arb_vs_local_constants(gl_shader
, context
, &stateBlock
->state
);
650 struct wined3d_shader
*pshader
= stateBlock
->state
.pixel_shader
;
651 const struct arb_ps_compiled_shader
*gl_shader
= priv
->compiled_fprog
;
652 UINT rt_height
= device
->fb
.render_targets
[0]->resource
.height
;
654 /* Load DirectX 9 float constants for pixel shader */
655 device
->highest_dirty_ps_const
= shader_arb_load_constantsF(pshader
, gl_info
, GL_FRAGMENT_PROGRAM_ARB
,
656 device
->highest_dirty_ps_const
, stateBlock
->state
.ps_consts_f
, context
->pshader_const_dirty
);
657 shader_arb_ps_local_constants(gl_shader
, context
, &stateBlock
->state
, rt_height
);
661 static void shader_arb_update_float_vertex_constants(struct wined3d_device
*device
, UINT start
, UINT count
)
663 struct wined3d_context
*context
= context_get_current();
665 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
666 * context. On a context switch the old context will be fully dirtified */
667 if (!context
|| context
->swapchain
->device
!= device
) return;
669 memset(context
->vshader_const_dirty
+ start
, 1, sizeof(*context
->vshader_const_dirty
) * count
);
670 device
->highest_dirty_vs_const
= max(device
->highest_dirty_vs_const
, start
+ count
);
673 static void shader_arb_update_float_pixel_constants(struct wined3d_device
*device
, UINT start
, UINT count
)
675 struct wined3d_context
*context
= context_get_current();
677 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
678 * context. On a context switch the old context will be fully dirtified */
679 if (!context
|| context
->swapchain
->device
!= device
) return;
681 memset(context
->pshader_const_dirty
+ start
, 1, sizeof(*context
->pshader_const_dirty
) * count
);
682 device
->highest_dirty_ps_const
= max(device
->highest_dirty_ps_const
, start
+ count
);
685 static DWORD
*local_const_mapping(const struct wined3d_shader
*shader
)
689 const local_constant
*lconst
;
691 if (shader
->load_local_constsF
|| list_empty(&shader
->constantsF
))
694 ret
= HeapAlloc(GetProcessHeap(), 0, sizeof(DWORD
) * shader
->limits
.constant_float
);
697 ERR("Out of memory\n");
701 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, local_constant
, entry
)
703 ret
[lconst
->idx
] = idx
++;
708 /* Generate the variable & register declarations for the ARB_vertex_program output target */
709 static DWORD
shader_generate_arb_declarations(struct wined3d_shader
*shader
,
710 const struct wined3d_shader_reg_maps
*reg_maps
, struct wined3d_shader_buffer
*buffer
,
711 const struct wined3d_gl_info
*gl_info
, DWORD
*lconst_map
,
712 DWORD
*num_clipplanes
, struct shader_arb_ctx_priv
*ctx
)
714 DWORD i
, next_local
= 0;
715 char pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
716 unsigned max_constantsF
;
717 const local_constant
*lconst
;
720 /* In pixel shaders, all private constants are program local, we don't need anything
721 * from program.env. Thus we can advertise the full set of constants in pixel shaders.
722 * If we need a private constant the GL implementation will squeeze it in somewhere
724 * With vertex shaders we need the posFixup and on some GL implementations 4 helper
725 * immediate values. The posFixup is loaded using program.env for now, so always
726 * subtract one from the number of constants. If the shader uses indirect addressing,
727 * account for the helper const too because we have to declare all availabke d3d constants
728 * and don't know which are actually used.
732 max_constantsF
= gl_info
->limits
.arb_ps_native_constants
;
733 /* 24 is the minimum MAX_PROGRAM_ENV_PARAMETERS_ARB value. */
734 if (max_constantsF
< 24)
735 max_constantsF
= gl_info
->limits
.arb_ps_float_constants
;
739 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
740 max_constantsF
= gl_info
->limits
.arb_vs_native_constants
;
741 /* 96 is the minimum MAX_PROGRAM_ENV_PARAMETERS_ARB value.
742 * Also prevents max_constantsF from becoming less than 0 and
744 if (max_constantsF
< 96)
745 max_constantsF
= gl_info
->limits
.arb_vs_float_constants
;
747 if (reg_maps
->usesrelconstF
)
749 DWORD highest_constf
= 0, clip_limit
;
751 max_constantsF
-= reserved_vs_const(shader_data
, reg_maps
, gl_info
);
752 max_constantsF
-= count_bits(reg_maps
->integer_constants
);
754 for (i
= 0; i
< shader
->limits
.constant_float
; ++i
)
757 DWORD shift
= i
& 0x1f;
758 if(reg_maps
->constf
[idx
] & (1 << shift
)) highest_constf
= i
;
761 if(use_nv_clip(gl_info
) && ctx
->target_version
>= NV2
)
763 if(ctx
->cur_vs_args
->super
.clip_enabled
)
764 clip_limit
= gl_info
->limits
.clipplanes
;
770 unsigned int mask
= ctx
->cur_vs_args
->clip
.boolclip
.clipplane_mask
;
771 clip_limit
= min(count_bits(mask
), 4);
773 *num_clipplanes
= min(clip_limit
, max_constantsF
- highest_constf
- 1);
774 max_constantsF
-= *num_clipplanes
;
775 if(*num_clipplanes
< clip_limit
)
777 WARN("Only %u clipplanes out of %u enabled\n", *num_clipplanes
, gl_info
->limits
.clipplanes
);
782 if (ctx
->target_version
>= NV2
) *num_clipplanes
= gl_info
->limits
.clipplanes
;
783 else *num_clipplanes
= min(gl_info
->limits
.clipplanes
, 4);
787 for (i
= 0, map
= reg_maps
->temporary
; map
; map
>>= 1, ++i
)
789 if (map
& 1) shader_addline(buffer
, "TEMP R%u;\n", i
);
792 for (i
= 0, map
= reg_maps
->address
; map
; map
>>= 1, ++i
)
794 if (map
& 1) shader_addline(buffer
, "ADDRESS A%u;\n", i
);
797 if (pshader
&& reg_maps
->shader_version
.major
== 1 && reg_maps
->shader_version
.minor
<= 3)
799 for (i
= 0, map
= reg_maps
->texcoord
; map
; map
>>= 1, ++i
)
801 if (map
& 1) shader_addline(buffer
, "TEMP T%u;\n", i
);
805 /* Load local constants using the program-local space,
806 * this avoids reloading them each time the shader is used
810 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, local_constant
, entry
)
812 shader_addline(buffer
, "PARAM C%u = program.local[%u];\n", lconst
->idx
,
813 lconst_map
[lconst
->idx
]);
814 next_local
= max(next_local
, lconst_map
[lconst
->idx
] + 1);
818 /* After subtracting privately used constants from the hardware limit(they are loaded as
819 * local constants), make sure the shader doesn't violate the env constant limit
823 max_constantsF
= min(max_constantsF
, gl_info
->limits
.arb_ps_float_constants
);
827 max_constantsF
= min(max_constantsF
, gl_info
->limits
.arb_vs_float_constants
);
830 /* Avoid declaring more constants than needed */
831 max_constantsF
= min(max_constantsF
, shader
->limits
.constant_float
);
833 /* we use the array-based constants array if the local constants are marked for loading,
834 * because then we use indirect addressing, or when the local constant list is empty,
835 * because then we don't know if we're using indirect addressing or not. If we're hardcoding
836 * local constants do not declare the loaded constants as an array because ARB compilers usually
837 * do not optimize unused constants away
839 if (reg_maps
->usesrelconstF
)
841 /* Need to PARAM the environment parameters (constants) so we can use relative addressing */
842 shader_addline(buffer
, "PARAM C[%d] = { program.env[0..%d] };\n",
843 max_constantsF
, max_constantsF
- 1);
845 for(i
= 0; i
< max_constantsF
; i
++) {
848 mask
= 1 << (i
& 0x1f);
849 if (!shader_constant_is_local(shader
, i
) && (reg_maps
->constf
[idx
] & mask
))
851 shader_addline(buffer
, "PARAM C%d = program.env[%d];\n",i
, i
);
859 static const char * const shift_tab
[] = {
860 "dummy", /* 0 (none) */
861 "coefmul.x", /* 1 (x2) */
862 "coefmul.y", /* 2 (x4) */
863 "coefmul.z", /* 3 (x8) */
864 "coefmul.w", /* 4 (x16) */
865 "dummy", /* 5 (x32) */
866 "dummy", /* 6 (x64) */
867 "dummy", /* 7 (x128) */
868 "dummy", /* 8 (d256) */
869 "dummy", /* 9 (d128) */
870 "dummy", /* 10 (d64) */
871 "dummy", /* 11 (d32) */
872 "coefdiv.w", /* 12 (d16) */
873 "coefdiv.z", /* 13 (d8) */
874 "coefdiv.y", /* 14 (d4) */
875 "coefdiv.x" /* 15 (d2) */
878 static void shader_arb_get_write_mask(const struct wined3d_shader_instruction
*ins
,
879 const struct wined3d_shader_dst_param
*dst
, char *write_mask
)
881 char *ptr
= write_mask
;
883 if (dst
->write_mask
!= WINED3DSP_WRITEMASK_ALL
)
886 if (dst
->write_mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
887 if (dst
->write_mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
888 if (dst
->write_mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
889 if (dst
->write_mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
895 static void shader_arb_get_swizzle(const struct wined3d_shader_src_param
*param
, BOOL fixup
, char *swizzle_str
)
897 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
898 * but addressed as "rgba". To fix this we need to swap the register's x
899 * and z components. */
900 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
901 char *ptr
= swizzle_str
;
903 /* swizzle bits fields: wwzzyyxx */
904 DWORD swizzle
= param
->swizzle
;
905 DWORD swizzle_x
= swizzle
& 0x03;
906 DWORD swizzle_y
= (swizzle
>> 2) & 0x03;
907 DWORD swizzle_z
= (swizzle
>> 4) & 0x03;
908 DWORD swizzle_w
= (swizzle
>> 6) & 0x03;
910 /* If the swizzle is the default swizzle (ie, "xyzw"), we don't need to
911 * generate a swizzle string. Unless we need to our own swizzling. */
912 if (swizzle
!= WINED3DSP_NOSWIZZLE
|| fixup
)
915 if (swizzle_x
== swizzle_y
&& swizzle_x
== swizzle_z
&& swizzle_x
== swizzle_w
) {
916 *ptr
++ = swizzle_chars
[swizzle_x
];
918 *ptr
++ = swizzle_chars
[swizzle_x
];
919 *ptr
++ = swizzle_chars
[swizzle_y
];
920 *ptr
++ = swizzle_chars
[swizzle_z
];
921 *ptr
++ = swizzle_chars
[swizzle_w
];
928 static void shader_arb_request_a0(const struct wined3d_shader_instruction
*ins
, const char *src
)
930 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
931 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
933 if (!strcmp(priv
->addr_reg
, src
)) return;
935 strcpy(priv
->addr_reg
, src
);
936 shader_addline(buffer
, "ARL A0.x, %s;\n", src
);
939 static void shader_arb_get_src_param(const struct wined3d_shader_instruction
*ins
,
940 const struct wined3d_shader_src_param
*src
, unsigned int tmpreg
, char *outregstr
);
942 static void shader_arb_get_register_name(const struct wined3d_shader_instruction
*ins
,
943 const struct wined3d_shader_register
*reg
, char *register_name
, BOOL
*is_color
)
945 /* oPos, oFog and oPts in D3D */
946 static const char * const rastout_reg_names
[] = {"TMP_OUT", "result.fogcoord", "result.pointsize"};
947 struct wined3d_shader
*shader
= ins
->ctx
->shader
;
948 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
949 BOOL pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
950 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
956 case WINED3DSPR_TEMP
:
957 sprintf(register_name
, "R%u", reg
->idx
);
960 case WINED3DSPR_INPUT
:
963 if (reg_maps
->shader_version
.major
< 3)
965 if (!reg
->idx
) strcpy(register_name
, "fragment.color.primary");
966 else strcpy(register_name
, "fragment.color.secondary");
973 shader_arb_get_src_param(ins
, reg
->rel_addr
, 0, rel_reg
);
975 if (!strcmp(rel_reg
, "**aL_emul**"))
977 DWORD idx
= ctx
->aL
+ reg
->idx
;
978 if(idx
< MAX_REG_INPUT
)
980 strcpy(register_name
, ctx
->ps_input
[idx
]);
984 ERR("Pixel shader input register out of bounds: %u\n", idx
);
985 sprintf(register_name
, "out_of_bounds_%u", idx
);
988 else if (reg_maps
->input_registers
& 0x0300)
990 /* There are two ways basically:
992 * 1) Use the unrolling code that is used for loop emulation and unroll the loop.
993 * That means trouble if the loop also contains a breakc or if the control values
994 * aren't local constants.
995 * 2) Generate an if block that checks if aL.y < 8, == 8 or == 9 and selects the
996 * source dynamically. The trouble is that we cannot simply read aL.y because it
997 * is an ADDRESS register. We could however push it, load .zw with a value and use
998 * ADAC to load the condition code register and pop it again afterwards
1000 FIXME("Relative input register addressing with more than 8 registers\n");
1002 /* This is better than nothing for now */
1003 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
);
1005 else if(ctx
->cur_ps_args
->super
.vp_mode
!= vertexshader
)
1007 /* This is problematic because we'd have to consult the ctx->ps_input strings
1008 * for where to find the varying. Some may be "0.0", others can be texcoords or
1009 * colors. This needs either a pipeline replacement to make the vertex shader feed
1010 * proper varyings, or loop unrolling
1012 * For now use the texcoords and hope for the best
1014 FIXME("Non-vertex shader varying input with indirect addressing\n");
1015 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
);
1019 /* D3D supports indirect addressing only with aL in loop registers. The loop instruction
1020 * pulls GL_NV_fragment_program2 in
1022 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
);
1027 if(reg
->idx
< MAX_REG_INPUT
)
1029 strcpy(register_name
, ctx
->ps_input
[reg
->idx
]);
1033 ERR("Pixel shader input register out of bounds: %u\n", reg
->idx
);
1034 sprintf(register_name
, "out_of_bounds_%u", reg
->idx
);
1041 if (ctx
->cur_vs_args
->super
.swizzle_map
& (1 << reg
->idx
)) *is_color
= TRUE
;
1042 sprintf(register_name
, "vertex.attrib[%u]", reg
->idx
);
1046 case WINED3DSPR_CONST
:
1047 if (!pshader
&& reg
->rel_addr
)
1049 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1050 UINT rel_offset
= shader_data
->rel_offset
;
1053 if (reg_maps
->shader_version
.major
< 2)
1055 sprintf(rel_reg
, "A0.x");
1057 shader_arb_get_src_param(ins
, reg
->rel_addr
, 0, rel_reg
);
1058 if(ctx
->target_version
== ARB
) {
1059 if (!strcmp(rel_reg
, "**aL_emul**"))
1063 shader_arb_request_a0(ins
, rel_reg
);
1064 sprintf(rel_reg
, "A0.x");
1069 sprintf(register_name
, "C[%u]", ctx
->aL
+ reg
->idx
);
1070 else if (reg
->idx
>= rel_offset
)
1071 sprintf(register_name
, "C[%s + %u]", rel_reg
, reg
->idx
- rel_offset
);
1073 sprintf(register_name
, "C[%s - %u]", rel_reg
, rel_offset
- reg
->idx
);
1077 if (reg_maps
->usesrelconstF
)
1078 sprintf(register_name
, "C[%u]", reg
->idx
);
1080 sprintf(register_name
, "C%u", reg
->idx
);
1084 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
1087 if (reg_maps
->shader_version
.major
== 1
1088 && reg_maps
->shader_version
.minor
<= 3)
1090 /* In ps <= 1.3, Tx is a temporary register as destination to all instructions,
1091 * and as source to most instructions. For some instructions it is the texcoord
1092 * input. Those instructions know about the special use
1094 sprintf(register_name
, "T%u", reg
->idx
);
1096 /* in ps 1.4 and 2.x Tx is always a (read-only) varying */
1097 sprintf(register_name
, "fragment.texcoord[%u]", reg
->idx
);
1102 if (reg_maps
->shader_version
.major
== 1 || ctx
->target_version
>= NV2
)
1104 sprintf(register_name
, "A%u", reg
->idx
);
1108 sprintf(register_name
, "A%u_SHADOW", reg
->idx
);
1113 case WINED3DSPR_COLOROUT
:
1114 if (ctx
->cur_ps_args
->super
.srgb_correction
&& !reg
->idx
)
1116 strcpy(register_name
, "TMP_COLOR");
1120 if(ctx
->cur_ps_args
->super
.srgb_correction
) FIXME("sRGB correction on higher render targets\n");
1121 if (reg_maps
->rt_mask
> 1)
1123 sprintf(register_name
, "result.color[%u]", reg
->idx
);
1127 strcpy(register_name
, "result.color");
1132 case WINED3DSPR_RASTOUT
:
1133 if(reg
->idx
== 1) sprintf(register_name
, "%s", ctx
->fog_output
);
1134 else sprintf(register_name
, "%s", rastout_reg_names
[reg
->idx
]);
1137 case WINED3DSPR_DEPTHOUT
:
1138 strcpy(register_name
, "result.depth");
1141 case WINED3DSPR_ATTROUT
:
1142 /* case WINED3DSPR_OUTPUT: */
1143 if (pshader
) sprintf(register_name
, "oD[%u]", reg
->idx
);
1144 else strcpy(register_name
, ctx
->color_output
[reg
->idx
]);
1147 case WINED3DSPR_TEXCRDOUT
:
1150 sprintf(register_name
, "oT[%u]", reg
->idx
);
1154 if (reg_maps
->shader_version
.major
< 3)
1156 strcpy(register_name
, ctx
->texcrd_output
[reg
->idx
]);
1160 strcpy(register_name
, ctx
->vs_output
[reg
->idx
]);
1165 case WINED3DSPR_LOOP
:
1166 if(ctx
->target_version
>= NV2
)
1168 /* Pshader has an implicitly declared loop index counter A0.x that cannot be renamed */
1169 if(pshader
) sprintf(register_name
, "A0.x");
1170 else sprintf(register_name
, "aL.y");
1174 /* Unfortunately this code cannot return the value of ctx->aL here. An immediate value
1175 * would be valid, but if aL is used for indexing(its only use), there's likely an offset,
1176 * thus the result would be something like C[15 + 30], which is not valid in the ARB program
1177 * grammar. So return a marker for the emulated aL and intercept it in constant and varying
1180 sprintf(register_name
, "**aL_emul**");
1185 case WINED3DSPR_CONSTINT
:
1186 sprintf(register_name
, "I%u", reg
->idx
);
1189 case WINED3DSPR_MISCTYPE
:
1192 sprintf(register_name
, "vpos");
1194 else if(reg
->idx
== 1)
1196 sprintf(register_name
, "fragment.facing.x");
1200 FIXME("Unknown MISCTYPE register index %u\n", reg
->idx
);
1205 FIXME("Unhandled register type %#x[%u]\n", reg
->type
, reg
->idx
);
1206 sprintf(register_name
, "unrecognized_register[%u]", reg
->idx
);
1211 static void shader_arb_get_dst_param(const struct wined3d_shader_instruction
*ins
,
1212 const struct wined3d_shader_dst_param
*wined3d_dst
, char *str
)
1214 char register_name
[255];
1218 shader_arb_get_register_name(ins
, &wined3d_dst
->reg
, register_name
, &is_color
);
1219 strcpy(str
, register_name
);
1221 shader_arb_get_write_mask(ins
, wined3d_dst
, write_mask
);
1222 strcat(str
, write_mask
);
1225 static const char *shader_arb_get_fixup_swizzle(enum fixup_channel_source channel_source
)
1227 switch(channel_source
)
1229 case CHANNEL_SOURCE_ZERO
: return "0";
1230 case CHANNEL_SOURCE_ONE
: return "1";
1231 case CHANNEL_SOURCE_X
: return "x";
1232 case CHANNEL_SOURCE_Y
: return "y";
1233 case CHANNEL_SOURCE_Z
: return "z";
1234 case CHANNEL_SOURCE_W
: return "w";
1236 FIXME("Unhandled channel source %#x\n", channel_source
);
1241 static void gen_color_correction(struct wined3d_shader_buffer
*buffer
, const char *reg
,
1242 DWORD dst_mask
, const char *one
, const char *two
, struct color_fixup_desc fixup
)
1246 if (is_complex_fixup(fixup
))
1248 enum complex_fixup complex_fixup
= get_complex_fixup(fixup
);
1249 FIXME("Complex fixup (%#x) not supported\n", complex_fixup
);
1254 if (fixup
.x_source
!= CHANNEL_SOURCE_X
) mask
|= WINED3DSP_WRITEMASK_0
;
1255 if (fixup
.y_source
!= CHANNEL_SOURCE_Y
) mask
|= WINED3DSP_WRITEMASK_1
;
1256 if (fixup
.z_source
!= CHANNEL_SOURCE_Z
) mask
|= WINED3DSP_WRITEMASK_2
;
1257 if (fixup
.w_source
!= CHANNEL_SOURCE_W
) mask
|= WINED3DSP_WRITEMASK_3
;
1262 shader_addline(buffer
, "SWZ %s, %s, %s, %s, %s, %s;\n", reg
, reg
,
1263 shader_arb_get_fixup_swizzle(fixup
.x_source
), shader_arb_get_fixup_swizzle(fixup
.y_source
),
1264 shader_arb_get_fixup_swizzle(fixup
.z_source
), shader_arb_get_fixup_swizzle(fixup
.w_source
));
1268 if (fixup
.x_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_0
;
1269 if (fixup
.y_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_1
;
1270 if (fixup
.z_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_2
;
1271 if (fixup
.w_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_3
;
1277 char *ptr
= reg_mask
;
1279 if (mask
!= WINED3DSP_WRITEMASK_ALL
)
1282 if (mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
1283 if (mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
1284 if (mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
1285 if (mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
1289 shader_addline(buffer
, "MAD %s%s, %s, %s, -%s;\n", reg
, reg_mask
, reg
, two
, one
);
1293 static const char *shader_arb_get_modifier(const struct wined3d_shader_instruction
*ins
)
1296 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1297 if (!ins
->dst_count
) return "";
1299 mod
= ins
->dst
[0].modifiers
;
1301 /* Silently ignore PARTIALPRECISION if its not supported */
1302 if(priv
->target_version
== ARB
) mod
&= ~WINED3DSPDM_PARTIALPRECISION
;
1304 if(mod
& WINED3DSPDM_MSAMPCENTROID
)
1306 FIXME("Unhandled modifier WINED3DSPDM_MSAMPCENTROID\n");
1307 mod
&= ~WINED3DSPDM_MSAMPCENTROID
;
1312 case WINED3DSPDM_SATURATE
| WINED3DSPDM_PARTIALPRECISION
:
1315 case WINED3DSPDM_SATURATE
:
1318 case WINED3DSPDM_PARTIALPRECISION
:
1325 FIXME("Unknown modifiers 0x%08x\n", mod
);
1330 #define TEX_PROJ 0x1
1331 #define TEX_BIAS 0x2
1333 #define TEX_DERIV 0x10
1335 static void shader_hw_sample(const struct wined3d_shader_instruction
*ins
, DWORD sampler_idx
,
1336 const char *dst_str
, const char *coord_reg
, WORD flags
, const char *dsx
, const char *dsy
)
1338 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1339 DWORD sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[sampler_idx
];
1340 const struct wined3d_texture
*texture
;
1341 const char *tex_type
;
1342 BOOL np2_fixup
= FALSE
;
1343 struct wined3d_shader
*shader
= ins
->ctx
->shader
;
1344 struct wined3d_device
*device
= shader
->device
;
1345 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1347 BOOL pshader
= shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
1349 /* D3D vertex shader sampler IDs are vertex samplers(0-3), not global d3d samplers */
1350 if(!pshader
) sampler_idx
+= MAX_FRAGMENT_SAMPLERS
;
1352 switch(sampler_type
) {
1358 texture
= device
->stateBlock
->state
.textures
[sampler_idx
];
1359 if (texture
&& texture
->target
== GL_TEXTURE_RECTANGLE_ARB
)
1365 if (shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
))
1367 if (priv
->cur_np2fixup_info
->super
.active
& (1 << sampler_idx
))
1369 if (flags
) FIXME("Only ordinary sampling from NP2 textures is supported.\n");
1370 else np2_fixup
= TRUE
;
1375 case WINED3DSTT_VOLUME
:
1379 case WINED3DSTT_CUBE
:
1384 ERR("Unexpected texture type %d\n", sampler_type
);
1388 /* TEX, TXL, TXD and TXP do not support the "H" modifier,
1389 * so don't use shader_arb_get_modifier
1391 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
) mod
= "_SAT";
1394 /* Fragment samplers always have indentity mapping */
1395 if(sampler_idx
>= MAX_FRAGMENT_SAMPLERS
)
1397 sampler_idx
= priv
->cur_vs_args
->vertex
.samplers
[sampler_idx
- MAX_FRAGMENT_SAMPLERS
];
1400 if (flags
& TEX_DERIV
)
1402 if(flags
& TEX_PROJ
) FIXME("Projected texture sampling with custom derivatives\n");
1403 if(flags
& TEX_BIAS
) FIXME("Biased texture sampling with custom derivatives\n");
1404 shader_addline(buffer
, "TXD%s %s, %s, %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
,
1405 dsx
, dsy
,sampler_idx
, tex_type
);
1407 else if(flags
& TEX_LOD
)
1409 if(flags
& TEX_PROJ
) FIXME("Projected texture sampling with explicit lod\n");
1410 if(flags
& TEX_BIAS
) FIXME("Biased texture sampling with explicit lod\n");
1411 shader_addline(buffer
, "TXL%s %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
,
1412 sampler_idx
, tex_type
);
1414 else if (flags
& TEX_BIAS
)
1416 /* Shouldn't be possible, but let's check for it */
1417 if(flags
& TEX_PROJ
) FIXME("Biased and Projected texture sampling\n");
1418 /* TXB takes the 4th component of the source vector automatically, as d3d. Nothing more to do */
1419 shader_addline(buffer
, "TXB%s %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
, sampler_idx
, tex_type
);
1421 else if (flags
& TEX_PROJ
)
1423 shader_addline(buffer
, "TXP%s %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
, sampler_idx
, tex_type
);
1429 const unsigned char idx
= priv
->cur_np2fixup_info
->super
.idx
[sampler_idx
];
1430 shader_addline(buffer
, "MUL TA, np2fixup[%u].%s, %s;\n", idx
>> 1,
1431 (idx
% 2) ? "zwxy" : "xyzw", coord_reg
);
1433 shader_addline(buffer
, "TEX%s %s, TA, texture[%u], %s;\n", mod
, dst_str
, sampler_idx
, tex_type
);
1436 shader_addline(buffer
, "TEX%s %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
, sampler_idx
, tex_type
);
1441 gen_color_correction(buffer
, dst_str
, ins
->dst
[0].write_mask
,
1442 arb_get_helper_value(WINED3D_SHADER_TYPE_PIXEL
, ARB_ONE
),
1443 arb_get_helper_value(WINED3D_SHADER_TYPE_PIXEL
, ARB_TWO
),
1444 priv
->cur_ps_args
->super
.color_fixup
[sampler_idx
]);
1448 static void shader_arb_get_src_param(const struct wined3d_shader_instruction
*ins
,
1449 const struct wined3d_shader_src_param
*src
, unsigned int tmpreg
, char *outregstr
)
1451 /* Generate a line that does the input modifier computation and return the input register to use */
1452 BOOL is_color
= FALSE
;
1456 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1457 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1458 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
1459 const char *two
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_TWO
);
1461 /* Assume a new line will be added */
1464 /* Get register name */
1465 shader_arb_get_register_name(ins
, &src
->reg
, regstr
, &is_color
);
1466 shader_arb_get_swizzle(src
, is_color
, swzstr
);
1468 switch (src
->modifiers
)
1470 case WINED3DSPSM_NONE
:
1471 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
1474 case WINED3DSPSM_NEG
:
1475 sprintf(outregstr
, "-%s%s", regstr
, swzstr
);
1478 case WINED3DSPSM_BIAS
:
1479 shader_addline(buffer
, "ADD T%c, %s, -coefdiv.x;\n", 'A' + tmpreg
, regstr
);
1481 case WINED3DSPSM_BIASNEG
:
1482 shader_addline(buffer
, "ADD T%c, -%s, coefdiv.x;\n", 'A' + tmpreg
, regstr
);
1484 case WINED3DSPSM_SIGN
:
1485 shader_addline(buffer
, "MAD T%c, %s, %s, -%s;\n", 'A' + tmpreg
, regstr
, two
, one
);
1487 case WINED3DSPSM_SIGNNEG
:
1488 shader_addline(buffer
, "MAD T%c, %s, -%s, %s;\n", 'A' + tmpreg
, regstr
, two
, one
);
1490 case WINED3DSPSM_COMP
:
1491 shader_addline(buffer
, "SUB T%c, %s, %s;\n", 'A' + tmpreg
, one
, regstr
);
1493 case WINED3DSPSM_X2
:
1494 shader_addline(buffer
, "ADD T%c, %s, %s;\n", 'A' + tmpreg
, regstr
, regstr
);
1496 case WINED3DSPSM_X2NEG
:
1497 shader_addline(buffer
, "ADD T%c, -%s, -%s;\n", 'A' + tmpreg
, regstr
, regstr
);
1499 case WINED3DSPSM_DZ
:
1500 shader_addline(buffer
, "RCP T%c, %s.z;\n", 'A' + tmpreg
, regstr
);
1501 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
1503 case WINED3DSPSM_DW
:
1504 shader_addline(buffer
, "RCP T%c, %s.w;\n", 'A' + tmpreg
, regstr
);
1505 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
1507 case WINED3DSPSM_ABS
:
1508 if(ctx
->target_version
>= NV2
) {
1509 sprintf(outregstr
, "|%s%s|", regstr
, swzstr
);
1512 shader_addline(buffer
, "ABS T%c, %s;\n", 'A' + tmpreg
, regstr
);
1515 case WINED3DSPSM_ABSNEG
:
1516 if(ctx
->target_version
>= NV2
) {
1517 sprintf(outregstr
, "-|%s%s|", regstr
, swzstr
);
1519 shader_addline(buffer
, "ABS T%c, %s;\n", 'A' + tmpreg
, regstr
);
1520 sprintf(outregstr
, "-T%c%s", 'A' + tmpreg
, swzstr
);
1525 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
1529 /* Return modified or original register, with swizzle */
1531 sprintf(outregstr
, "T%c%s", 'A' + tmpreg
, swzstr
);
1534 static void pshader_hw_bem(const struct wined3d_shader_instruction
*ins
)
1536 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1537 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1539 char src_name
[2][50];
1540 DWORD sampler_code
= dst
->reg
.idx
;
1542 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1544 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed
1546 * Keep in mind that src_name[1] can be "TB" and src_name[0] can be "TA" because modifiers like _x2 are valid
1547 * with bem. So delay loading the first parameter until after the perturbation calculation which needs two
1550 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1551 shader_addline(buffer
, "SWZ TA, bumpenvmat%d, x, z, 0, 0;\n", sampler_code
);
1552 shader_addline(buffer
, "DP3 TC.r, TA, %s;\n", src_name
[1]);
1553 shader_addline(buffer
, "SWZ TA, bumpenvmat%d, y, w, 0, 0;\n", sampler_code
);
1554 shader_addline(buffer
, "DP3 TC.g, TA, %s;\n", src_name
[1]);
1556 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1557 shader_addline(buffer
, "ADD %s, %s, TC;\n", dst_name
, src_name
[0]);
1560 static DWORD
negate_modifiers(DWORD mod
, char *extra_char
)
1565 case WINED3DSPSM_NONE
: return WINED3DSPSM_NEG
;
1566 case WINED3DSPSM_NEG
: return WINED3DSPSM_NONE
;
1567 case WINED3DSPSM_BIAS
: return WINED3DSPSM_BIASNEG
;
1568 case WINED3DSPSM_BIASNEG
: return WINED3DSPSM_BIAS
;
1569 case WINED3DSPSM_SIGN
: return WINED3DSPSM_SIGNNEG
;
1570 case WINED3DSPSM_SIGNNEG
: return WINED3DSPSM_SIGN
;
1571 case WINED3DSPSM_COMP
: *extra_char
= '-'; return WINED3DSPSM_COMP
;
1572 case WINED3DSPSM_X2
: return WINED3DSPSM_X2NEG
;
1573 case WINED3DSPSM_X2NEG
: return WINED3DSPSM_X2
;
1574 case WINED3DSPSM_DZ
: *extra_char
= '-'; return WINED3DSPSM_DZ
;
1575 case WINED3DSPSM_DW
: *extra_char
= '-'; return WINED3DSPSM_DW
;
1576 case WINED3DSPSM_ABS
: return WINED3DSPSM_ABSNEG
;
1577 case WINED3DSPSM_ABSNEG
: return WINED3DSPSM_ABS
;
1579 FIXME("Unknown modifier %u\n", mod
);
1583 static void pshader_hw_cnd(const struct wined3d_shader_instruction
*ins
)
1585 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1586 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1588 char src_name
[3][50];
1589 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1590 ins
->ctx
->reg_maps
->shader_version
.minor
);
1593 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1594 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1596 /* The coissue flag changes the semantic of the cnd instruction in <= 1.3 shaders */
1597 if (shader_version
<= WINED3D_SHADER_VERSION(1, 3) && ins
->coissue
)
1599 shader_addline(buffer
, "MOV%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
[1]);
1601 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
1604 /* src0 may have a negate srcmod set, so we can't blindly add "-" to the name */
1605 src0_copy
.modifiers
= negate_modifiers(src0_copy
.modifiers
, &extra_neg
);
1607 shader_arb_get_src_param(ins
, &src0_copy
, 0, src_name
[0]);
1608 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1609 shader_addline(buffer
, "ADD TA, %c%s, coefdiv.x;\n", extra_neg
, src_name
[0]);
1610 /* No modifiers supported on CMP */
1611 shader_addline(buffer
, "CMP %s, TA, %s, %s;\n", dst_name
, src_name
[1], src_name
[2]);
1613 /* _SAT on CMP doesn't make much sense, but it is not a pure NOP */
1614 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
)
1616 shader_arb_get_register_name(ins
, &dst
->reg
, src_name
[0], &is_color
);
1617 shader_addline(buffer
, "MOV_SAT %s, %s;\n", dst_name
, dst_name
);
1622 static void pshader_hw_cmp(const struct wined3d_shader_instruction
*ins
)
1624 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1625 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1627 char src_name
[3][50];
1630 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1632 /* Generate input register names (with modifiers) */
1633 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1634 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1635 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1637 /* No modifiers are supported on CMP */
1638 shader_addline(buffer
, "CMP %s, %s, %s, %s;\n", dst_name
,
1639 src_name
[0], src_name
[2], src_name
[1]);
1641 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
)
1643 shader_arb_get_register_name(ins
, &dst
->reg
, src_name
[0], &is_color
);
1644 shader_addline(buffer
, "MOV_SAT %s, %s;\n", dst_name
, src_name
[0]);
1648 /** Process the WINED3DSIO_DP2ADD instruction in ARB.
1649 * dst = dot2(src0, src1) + src2 */
1650 static void pshader_hw_dp2add(const struct wined3d_shader_instruction
*ins
)
1652 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1653 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1655 char src_name
[3][50];
1656 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1658 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1659 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1660 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1662 if(ctx
->target_version
>= NV3
)
1664 /* GL_NV_fragment_program2 has a 1:1 matching instruction */
1665 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1666 shader_addline(buffer
, "DP2A%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1667 dst_name
, src_name
[0], src_name
[1], src_name
[2]);
1669 else if(ctx
->target_version
>= NV2
)
1671 /* dst.x = src2.?, src0.x, src1.x + src0.y * src1.y
1672 * dst.y = src2.?, src0.x, src1.z + src0.y * src1.w
1673 * dst.z = src2.?, src0.x, src1.x + src0.y * src1.y
1674 * dst.z = src2.?, src0.x, src1.z + src0.y * src1.w
1676 * Make sure that src1.zw = src1.xy, then we get a classic dp2add
1678 * .xyxy and other swizzles that we could get with this are not valid in
1679 * plain ARBfp, but luckily the NV extension grammar lifts this limitation.
1681 struct wined3d_shader_src_param tmp_param
= ins
->src
[1];
1682 DWORD swizzle
= tmp_param
.swizzle
& 0xf; /* Selects .xy */
1683 tmp_param
.swizzle
= swizzle
| (swizzle
<< 4); /* Creates .xyxy */
1685 shader_arb_get_src_param(ins
, &tmp_param
, 1, src_name
[1]);
1687 shader_addline(buffer
, "X2D%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1688 dst_name
, src_name
[2], src_name
[0], src_name
[1]);
1692 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1693 /* 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]
1694 * src_name[0] can be TA, but TA is a private temp for modifiers, so it is save to overwrite
1696 shader_addline(buffer
, "MOV TA, %s;\n", src_name
[0]);
1697 shader_addline(buffer
, "MOV TA.z, 0.0;\n");
1698 shader_addline(buffer
, "DP3 TA, TA, %s;\n", src_name
[1]);
1699 shader_addline(buffer
, "ADD%s %s, TA, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
[2]);
1703 /* Map the opcode 1-to-1 to the GL code */
1704 static void shader_hw_map2gl(const struct wined3d_shader_instruction
*ins
)
1706 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1707 const char *instruction
;
1708 char arguments
[256], dst_str
[50];
1710 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1712 switch (ins
->handler_idx
)
1714 case WINED3DSIH_ABS
: instruction
= "ABS"; break;
1715 case WINED3DSIH_ADD
: instruction
= "ADD"; break;
1716 case WINED3DSIH_CRS
: instruction
= "XPD"; break;
1717 case WINED3DSIH_DP3
: instruction
= "DP3"; break;
1718 case WINED3DSIH_DP4
: instruction
= "DP4"; break;
1719 case WINED3DSIH_DST
: instruction
= "DST"; break;
1720 case WINED3DSIH_FRC
: instruction
= "FRC"; break;
1721 case WINED3DSIH_LIT
: instruction
= "LIT"; break;
1722 case WINED3DSIH_LRP
: instruction
= "LRP"; break;
1723 case WINED3DSIH_MAD
: instruction
= "MAD"; break;
1724 case WINED3DSIH_MAX
: instruction
= "MAX"; break;
1725 case WINED3DSIH_MIN
: instruction
= "MIN"; break;
1726 case WINED3DSIH_MOV
: instruction
= "MOV"; break;
1727 case WINED3DSIH_MUL
: instruction
= "MUL"; break;
1728 case WINED3DSIH_SGE
: instruction
= "SGE"; break;
1729 case WINED3DSIH_SLT
: instruction
= "SLT"; break;
1730 case WINED3DSIH_SUB
: instruction
= "SUB"; break;
1731 case WINED3DSIH_MOVA
:instruction
= "ARR"; break;
1732 case WINED3DSIH_DSX
: instruction
= "DDX"; break;
1733 default: instruction
= "";
1734 FIXME("Unhandled opcode %#x\n", ins
->handler_idx
);
1738 /* Note that shader_arb_add_dst_param() adds spaces. */
1739 arguments
[0] = '\0';
1740 shader_arb_get_dst_param(ins
, dst
, dst_str
);
1741 for (i
= 0; i
< ins
->src_count
; ++i
)
1744 strcat(arguments
, ", ");
1745 shader_arb_get_src_param(ins
, &ins
->src
[i
], i
, operand
);
1746 strcat(arguments
, operand
);
1748 shader_addline(buffer
, "%s%s %s%s;\n", instruction
, shader_arb_get_modifier(ins
), dst_str
, arguments
);
1751 static void shader_hw_nop(const struct wined3d_shader_instruction
*ins
)
1753 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1754 shader_addline(buffer
, "NOP;\n");
1757 static void shader_hw_mov(const struct wined3d_shader_instruction
*ins
)
1759 struct wined3d_shader
*shader
= ins
->ctx
->shader
;
1760 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
1761 BOOL pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
1762 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1763 const char *zero
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ZERO
);
1764 const char *one
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ONE
);
1765 const char *two
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_TWO
);
1767 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1768 char src0_param
[256];
1770 if (ins
->handler_idx
== WINED3DSIH_MOVA
)
1772 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1774 const char *offset
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_VS_REL_OFFSET
);
1776 if(ctx
->target_version
>= NV2
) {
1777 shader_hw_map2gl(ins
);
1780 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_param
);
1781 shader_arb_get_write_mask(ins
, &ins
->dst
[0], write_mask
);
1783 /* This implements the mova formula used in GLSL. The first two instructions
1784 * prepare the sign() part. Note that it is fine to have my_sign(0.0) = 1.0
1788 * A0.x = arl(floor(abs(0.0) + 0.5) * 1.0) = floor(0.5) = 0.0 since arl does a floor
1790 * The ARL is performed when A0 is used - the requested component is read from A0_SHADOW into
1791 * A0.x. We can use the overwritten component of A0_shadow as temporary storage for the sign.
1793 shader_addline(buffer
, "SGE A0_SHADOW%s, %s, %s;\n", write_mask
, src0_param
, zero
);
1794 shader_addline(buffer
, "MAD A0_SHADOW%s, A0_SHADOW, %s, -%s;\n", write_mask
, two
, one
);
1796 shader_addline(buffer
, "ABS TA%s, %s;\n", write_mask
, src0_param
);
1797 shader_addline(buffer
, "ADD TA%s, TA, rel_addr_const.x;\n", write_mask
);
1798 shader_addline(buffer
, "FLR TA%s, TA;\n", write_mask
);
1799 if (shader_data
->rel_offset
)
1801 shader_addline(buffer
, "ADD TA%s, TA, %s;\n", write_mask
, offset
);
1803 shader_addline(buffer
, "MUL A0_SHADOW%s, TA, A0_SHADOW;\n", write_mask
);
1805 ((struct shader_arb_ctx_priv
*)ins
->ctx
->backend_data
)->addr_reg
[0] = '\0';
1807 else if (reg_maps
->shader_version
.major
== 1
1808 && !shader_is_pshader_version(reg_maps
->shader_version
.type
)
1809 && ins
->dst
[0].reg
.type
== WINED3DSPR_ADDR
)
1811 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1812 src0_param
[0] = '\0';
1814 if (shader_data
->rel_offset
)
1816 const char *offset
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_VS_REL_OFFSET
);
1817 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_param
);
1818 shader_addline(buffer
, "ADD TA.x, %s, %s;\n", src0_param
, offset
);
1819 shader_addline(buffer
, "ARL A0.x, TA.x;\n");
1823 /* Apple's ARB_vertex_program implementation does not accept an ARL source argument
1824 * with more than one component. Thus replicate the first source argument over all
1825 * 4 components. For example, .xyzw -> .x (or better: .xxxx), .zwxy -> .z, etc) */
1826 struct wined3d_shader_src_param tmp_src
= ins
->src
[0];
1827 tmp_src
.swizzle
= (tmp_src
.swizzle
& 0x3) * 0x55;
1828 shader_arb_get_src_param(ins
, &tmp_src
, 0, src0_param
);
1829 shader_addline(buffer
, "ARL A0.x, %s;\n", src0_param
);
1832 else if (ins
->dst
[0].reg
.type
== WINED3DSPR_COLOROUT
&& !ins
->dst
[0].reg
.idx
&& pshader
)
1834 if (ctx
->cur_ps_args
->super
.srgb_correction
&& shader
->u
.ps
.color0_mov
)
1836 shader_addline(buffer
, "#mov handled in srgb write code\n");
1839 shader_hw_map2gl(ins
);
1843 shader_hw_map2gl(ins
);
1847 static void pshader_hw_texkill(const struct wined3d_shader_instruction
*ins
)
1849 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1850 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1853 /* No swizzles are allowed in d3d's texkill. PS 1.x ignores the 4th component as documented,
1854 * but >= 2.0 honors it(undocumented, but tested by the d3d9 testsuit)
1856 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1858 if (ins
->ctx
->reg_maps
->shader_version
.major
>= 2)
1860 const char *kilsrc
= "TA";
1863 shader_arb_get_register_name(ins
, &dst
->reg
, reg_dest
, &is_color
);
1864 if(dst
->write_mask
== WINED3DSP_WRITEMASK_ALL
)
1870 /* Sigh. KIL doesn't support swizzles/writemasks. KIL passes a writemask, but ".xy" for example
1871 * is not valid as a swizzle in ARB (needs ".xyyy"). Use SWZ to load the register properly, and set
1872 * masked out components to 0(won't kill)
1874 char x
= '0', y
= '0', z
= '0', w
= '0';
1875 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
) x
= 'x';
1876 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
) y
= 'y';
1877 if(dst
->write_mask
& WINED3DSP_WRITEMASK_2
) z
= 'z';
1878 if(dst
->write_mask
& WINED3DSP_WRITEMASK_3
) w
= 'w';
1879 shader_addline(buffer
, "SWZ TA, %s, %c, %c, %c, %c;\n", reg_dest
, x
, y
, z
, w
);
1881 shader_addline(buffer
, "KIL %s;\n", kilsrc
);
1883 /* ARB fp doesn't like swizzles on the parameter of the KIL instruction. To mask the 4th component,
1884 * copy the register into our general purpose TMP variable, overwrite .w and pass TMP to KIL
1886 * ps_1_3 shaders use the texcoord incarnation of the Tx register. ps_1_4 shaders can use the same,
1887 * or pass in any temporary register(in shader phase 2)
1889 if(ins
->ctx
->reg_maps
->shader_version
.minor
<= 3) {
1890 sprintf(reg_dest
, "fragment.texcoord[%u]", dst
->reg
.idx
);
1892 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1894 shader_addline(buffer
, "SWZ TA, %s, x, y, z, 1;\n", reg_dest
);
1895 shader_addline(buffer
, "KIL TA;\n");
1899 static void pshader_hw_tex(const struct wined3d_shader_instruction
*ins
)
1901 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1902 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1903 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1904 ins
->ctx
->reg_maps
->shader_version
.minor
);
1905 struct wined3d_shader_src_param src
;
1909 DWORD reg_sampler_code
;
1912 /* All versions have a destination register */
1913 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1915 /* 1.0-1.4: Use destination register number as texture code.
1916 2.0+: Use provided sampler number as texure code. */
1917 if (shader_version
< WINED3D_SHADER_VERSION(2,0))
1918 reg_sampler_code
= dst
->reg
.idx
;
1920 reg_sampler_code
= ins
->src
[1].reg
.idx
;
1922 /* 1.0-1.3: Use the texcoord varying.
1923 1.4+: Use provided coordinate source register. */
1924 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
1925 sprintf(reg_coord
, "fragment.texcoord[%u]", reg_sampler_code
);
1927 /* TEX is the only instruction that can handle DW and DZ natively */
1929 if(src
.modifiers
== WINED3DSPSM_DW
) src
.modifiers
= WINED3DSPSM_NONE
;
1930 if(src
.modifiers
== WINED3DSPSM_DZ
) src
.modifiers
= WINED3DSPSM_NONE
;
1931 shader_arb_get_src_param(ins
, &src
, 0, reg_coord
);
1935 * 1.1, 1.2, 1.3: Use WINED3DTSS_TEXTURETRANSFORMFLAGS
1936 * 1.4: Use WINED3DSPSM_DZ or WINED3DSPSM_DW on src[0]
1937 * 2.0+: Use WINED3DSI_TEXLD_PROJECT on the opcode
1939 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
1942 if (reg_sampler_code
< MAX_TEXTURES
)
1943 flags
= priv
->cur_ps_args
->super
.tex_transform
>> reg_sampler_code
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
;
1944 if (flags
& WINED3D_PSARGS_PROJECTED
)
1945 myflags
|= TEX_PROJ
;
1947 else if (shader_version
< WINED3D_SHADER_VERSION(2,0))
1949 DWORD src_mod
= ins
->src
[0].modifiers
;
1950 if (src_mod
== WINED3DSPSM_DZ
) {
1951 /* TXP cannot handle DZ natively, so move the z coordinate to .w. reg_coord is a read-only
1952 * varying register, so we need a temp reg
1954 shader_addline(ins
->ctx
->buffer
, "SWZ TA, %s, x, y, z, z;\n", reg_coord
);
1955 strcpy(reg_coord
, "TA");
1956 myflags
|= TEX_PROJ
;
1957 } else if(src_mod
== WINED3DSPSM_DW
) {
1958 myflags
|= TEX_PROJ
;
1961 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) myflags
|= TEX_PROJ
;
1962 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) myflags
|= TEX_BIAS
;
1964 shader_hw_sample(ins
, reg_sampler_code
, reg_dest
, reg_coord
, myflags
, NULL
, NULL
);
1967 static void pshader_hw_texcoord(const struct wined3d_shader_instruction
*ins
)
1969 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1970 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1971 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1972 ins
->ctx
->reg_maps
->shader_version
.minor
);
1975 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
1977 DWORD reg
= dst
->reg
.idx
;
1979 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
1980 shader_addline(buffer
, "MOV_SAT %s, fragment.texcoord[%u];\n", dst_str
, reg
);
1984 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_src
);
1985 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
1986 shader_addline(buffer
, "MOV %s, %s;\n", dst_str
, reg_src
);
1990 static void pshader_hw_texreg2ar(const struct wined3d_shader_instruction
*ins
)
1992 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1995 DWORD reg1
= ins
->dst
[0].reg
.idx
;
1999 /* Note that texreg2ar treats Tx as a temporary register, not as a varying */
2000 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2001 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2002 /* Move .x first in case src_str is "TA" */
2003 shader_addline(buffer
, "MOV TA.y, %s.x;\n", src_str
);
2004 shader_addline(buffer
, "MOV TA.x, %s.w;\n", src_str
);
2005 if (reg1
< MAX_TEXTURES
)
2007 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2008 flags
= priv
->cur_ps_args
->super
.tex_transform
>> reg1
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
;
2010 shader_hw_sample(ins
, reg1
, dst_str
, "TA", flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2013 static void pshader_hw_texreg2gb(const struct wined3d_shader_instruction
*ins
)
2015 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2017 DWORD reg1
= ins
->dst
[0].reg
.idx
;
2021 /* Note that texreg2gb treats Tx as a temporary register, not as a varying */
2022 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2023 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2024 shader_addline(buffer
, "MOV TA.x, %s.y;\n", src_str
);
2025 shader_addline(buffer
, "MOV TA.y, %s.z;\n", src_str
);
2026 shader_hw_sample(ins
, reg1
, dst_str
, "TA", 0, NULL
, NULL
);
2029 static void pshader_hw_texreg2rgb(const struct wined3d_shader_instruction
*ins
)
2031 DWORD reg1
= ins
->dst
[0].reg
.idx
;
2035 /* Note that texreg2rg treats Tx as a temporary register, not as a varying */
2036 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2037 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2038 shader_hw_sample(ins
, reg1
, dst_str
, src_str
, 0, NULL
, NULL
);
2041 static void pshader_hw_texbem(const struct wined3d_shader_instruction
*ins
)
2043 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2044 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2045 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2046 char reg_coord
[40], dst_reg
[50], src_reg
[50];
2047 DWORD reg_dest_code
;
2049 /* All versions have a destination register. The Tx where the texture coordinates come
2050 * from is the varying incarnation of the texture register
2052 reg_dest_code
= dst
->reg
.idx
;
2053 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_reg
);
2054 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_reg
);
2055 sprintf(reg_coord
, "fragment.texcoord[%u]", reg_dest_code
);
2057 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed
2058 * The Tx in which the perturbation map is stored is the tempreg incarnation of the texture register
2060 * GL_NV_fragment_program_option could handle this in one instruction via X2D:
2061 * X2D TA.xy, fragment.texcoord, T%u, bumpenvmat%u.xzyw
2063 * However, the NV extensions are never enabled for <= 2.0 shaders because of the performance penalty that
2064 * comes with it, and texbem is an 1.x only instruction. No 1.x instruction forces us to enable the NV
2067 shader_addline(buffer
, "SWZ TB, bumpenvmat%d, x, z, 0, 0;\n", reg_dest_code
);
2068 shader_addline(buffer
, "DP3 TA.x, TB, %s;\n", src_reg
);
2069 shader_addline(buffer
, "SWZ TB, bumpenvmat%d, y, w, 0, 0;\n", reg_dest_code
);
2070 shader_addline(buffer
, "DP3 TA.y, TB, %s;\n", src_reg
);
2072 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2073 * so we can't let the GL handle this.
2075 if ((priv
->cur_ps_args
->super
.tex_transform
>> reg_dest_code
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
)
2076 & WINED3D_PSARGS_PROJECTED
)
2078 shader_addline(buffer
, "RCP TB.w, %s.w;\n", reg_coord
);
2079 shader_addline(buffer
, "MUL TB.xy, %s, TB.w;\n", reg_coord
);
2080 shader_addline(buffer
, "ADD TA.xy, TA, TB;\n");
2082 shader_addline(buffer
, "ADD TA.xy, TA, %s;\n", reg_coord
);
2085 shader_hw_sample(ins
, reg_dest_code
, dst_reg
, "TA", 0, NULL
, NULL
);
2087 if (ins
->handler_idx
== WINED3DSIH_TEXBEML
)
2089 /* No src swizzles are allowed, so this is ok */
2090 shader_addline(buffer
, "MAD TA, %s.z, luminance%d.x, luminance%d.y;\n",
2091 src_reg
, reg_dest_code
, reg_dest_code
);
2092 shader_addline(buffer
, "MUL %s, %s, TA;\n", dst_reg
, dst_reg
);
2096 static void pshader_hw_texm3x2pad(const struct wined3d_shader_instruction
*ins
)
2098 DWORD reg
= ins
->dst
[0].reg
.idx
;
2099 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2100 char src0_name
[50], dst_name
[50];
2102 struct wined3d_shader_register tmp_reg
= ins
->dst
[0].reg
;
2104 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2105 /* The next instruction will be a texm3x2tex or texm3x2depth that writes to the uninitialized
2106 * T<reg+1> register. Use this register to store the calculated vector
2108 tmp_reg
.idx
= reg
+ 1;
2109 shader_arb_get_register_name(ins
, &tmp_reg
, dst_name
, &is_color
);
2110 shader_addline(buffer
, "DP3 %s.x, fragment.texcoord[%u], %s;\n", dst_name
, reg
, src0_name
);
2113 static void pshader_hw_texm3x2tex(const struct wined3d_shader_instruction
*ins
)
2115 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2117 DWORD reg
= ins
->dst
[0].reg
.idx
;
2118 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2124 /* We know that we're writing to the uninitialized T<reg> register, so use it for temporary storage */
2125 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2127 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2128 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2129 shader_addline(buffer
, "DP3 %s.y, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2130 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2131 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3DTTFF_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2134 static void pshader_hw_texm3x3pad(const struct wined3d_shader_instruction
*ins
)
2136 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2137 DWORD reg
= ins
->dst
[0].reg
.idx
;
2138 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2139 char src0_name
[50], dst_name
[50];
2140 struct wined3d_shader_register tmp_reg
= ins
->dst
[0].reg
;
2143 /* There are always 2 texm3x3pad instructions followed by one texm3x3[tex,vspec, ...] instruction, with
2144 * incrementing ins->dst[0].register_idx numbers. So the pad instruction already knows the final destination
2145 * register, and this register is uninitialized(otherwise the assembler complains that it is 'redeclared')
2147 tmp_reg
.idx
= reg
+ 2 - tex_mx
->current_row
;
2148 shader_arb_get_register_name(ins
, &tmp_reg
, dst_name
, &is_color
);
2150 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2151 shader_addline(buffer
, "DP3 %s.%c, fragment.texcoord[%u], %s;\n",
2152 dst_name
, 'x' + tex_mx
->current_row
, reg
, src0_name
);
2153 tex_mx
->texcoord_w
[tex_mx
->current_row
++] = reg
;
2156 static void pshader_hw_texm3x3tex(const struct wined3d_shader_instruction
*ins
)
2158 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2159 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2161 DWORD reg
= ins
->dst
[0].reg
.idx
;
2162 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2164 char src0_name
[50], dst_name
[50];
2167 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2168 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2169 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_name
, reg
, src0_name
);
2171 /* Sample the texture using the calculated coordinates */
2172 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2173 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2174 shader_hw_sample(ins
, reg
, dst_str
, dst_name
, flags
& WINED3DTTFF_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2175 tex_mx
->current_row
= 0;
2178 static void pshader_hw_texm3x3vspec(const struct wined3d_shader_instruction
*ins
)
2180 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2181 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2183 DWORD reg
= ins
->dst
[0].reg
.idx
;
2184 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2190 /* Get the dst reg without writemask strings. We know this register is uninitialized, so we can use all
2191 * components for temporary data storage
2193 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2194 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2195 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2197 /* Construct the eye-ray vector from w coordinates */
2198 shader_addline(buffer
, "MOV TB.x, fragment.texcoord[%u].w;\n", tex_mx
->texcoord_w
[0]);
2199 shader_addline(buffer
, "MOV TB.y, fragment.texcoord[%u].w;\n", tex_mx
->texcoord_w
[1]);
2200 shader_addline(buffer
, "MOV TB.z, fragment.texcoord[%u].w;\n", reg
);
2202 /* Calculate reflection vector
2204 shader_addline(buffer
, "DP3 %s.w, %s, TB;\n", dst_reg
, dst_reg
);
2205 /* The .w is ignored when sampling, so I can use TB.w to calculate dot(N, N) */
2206 shader_addline(buffer
, "DP3 TB.w, %s, %s;\n", dst_reg
, dst_reg
);
2207 shader_addline(buffer
, "RCP TB.w, TB.w;\n");
2208 shader_addline(buffer
, "MUL %s.w, %s.w, TB.w;\n", dst_reg
, dst_reg
);
2209 shader_addline(buffer
, "MUL %s, %s.w, %s;\n", dst_reg
, dst_reg
, dst_reg
);
2210 shader_addline(buffer
, "MAD %s, coefmul.x, %s, -TB;\n", dst_reg
, dst_reg
);
2212 /* Sample the texture using the calculated coordinates */
2213 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2214 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2215 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3DTTFF_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2216 tex_mx
->current_row
= 0;
2219 static void pshader_hw_texm3x3spec(const struct wined3d_shader_instruction
*ins
)
2221 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2222 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2224 DWORD reg
= ins
->dst
[0].reg
.idx
;
2225 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2232 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2233 shader_arb_get_src_param(ins
, &ins
->src
[0], 1, src1_name
);
2234 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2235 /* Note: dst_reg.xy is input here, generated by two texm3x3pad instructions */
2236 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2238 /* Calculate reflection vector.
2241 * dst_reg.xyz = 2 * --------- * N - E
2244 * Which normalizes the normal vector
2246 shader_addline(buffer
, "DP3 %s.w, %s, %s;\n", dst_reg
, dst_reg
, src1_name
);
2247 shader_addline(buffer
, "DP3 TC.w, %s, %s;\n", dst_reg
, dst_reg
);
2248 shader_addline(buffer
, "RCP TC.w, TC.w;\n");
2249 shader_addline(buffer
, "MUL %s.w, %s.w, TC.w;\n", dst_reg
, dst_reg
);
2250 shader_addline(buffer
, "MUL %s, %s.w, %s;\n", dst_reg
, dst_reg
, dst_reg
);
2251 shader_addline(buffer
, "MAD %s, coefmul.x, %s, -%s;\n", dst_reg
, dst_reg
, src1_name
);
2253 /* Sample the texture using the calculated coordinates */
2254 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2255 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2256 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3DTTFF_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2257 tex_mx
->current_row
= 0;
2260 static void pshader_hw_texdepth(const struct wined3d_shader_instruction
*ins
)
2262 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2263 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2265 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2266 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2268 /* texdepth has an implicit destination, the fragment depth value. It's only parameter,
2269 * which is essentially an input, is the destination register because it is the first
2270 * parameter. According to the msdn, this must be register r5, but let's keep it more flexible
2271 * here(writemasks/swizzles are not valid on texdepth)
2273 shader_arb_get_dst_param(ins
, dst
, dst_name
);
2275 /* According to the msdn, the source register(must be r5) is unusable after
2276 * the texdepth instruction, so we're free to modify it
2278 shader_addline(buffer
, "MIN %s.y, %s.y, %s;\n", dst_name
, dst_name
, one
);
2280 /* How to deal with the special case dst_name.g == 0? if r != 0, then
2281 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
2282 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
2284 shader_addline(buffer
, "RCP %s.y, %s.y;\n", dst_name
, dst_name
);
2285 shader_addline(buffer
, "MUL TA.x, %s.x, %s.y;\n", dst_name
, dst_name
);
2286 shader_addline(buffer
, "MIN TA.x, TA.x, %s;\n", one
);
2287 shader_addline(buffer
, "MAX result.depth, TA.x, %s;\n", zero
);
2290 /** Process the WINED3DSIO_TEXDP3TEX instruction in ARB:
2291 * Take a 3-component dot product of the TexCoord[dstreg] and src,
2292 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
2293 static void pshader_hw_texdp3tex(const struct wined3d_shader_instruction
*ins
)
2295 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2296 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
;
2300 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2301 shader_addline(buffer
, "MOV TB, 0.0;\n");
2302 shader_addline(buffer
, "DP3 TB.x, fragment.texcoord[%u], %s;\n", sampler_idx
, src0
);
2304 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2305 shader_hw_sample(ins
, sampler_idx
, dst_str
, "TB", 0 /* Only one coord, can't be projected */, NULL
, NULL
);
2308 /** Process the WINED3DSIO_TEXDP3 instruction in ARB:
2309 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
2310 static void pshader_hw_texdp3(const struct wined3d_shader_instruction
*ins
)
2312 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2315 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2317 /* Handle output register */
2318 shader_arb_get_dst_param(ins
, dst
, dst_str
);
2319 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2320 shader_addline(buffer
, "DP3 %s, fragment.texcoord[%u], %s;\n", dst_str
, dst
->reg
.idx
, src0
);
2323 /** Process the WINED3DSIO_TEXM3X3 instruction in ARB
2324 * Perform the 3rd row of a 3x3 matrix multiply */
2325 static void pshader_hw_texm3x3(const struct wined3d_shader_instruction
*ins
)
2327 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2328 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2329 char dst_str
[50], dst_name
[50];
2333 shader_arb_get_dst_param(ins
, dst
, dst_str
);
2334 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2335 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2336 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_name
, dst
->reg
.idx
, src0
);
2337 shader_addline(buffer
, "MOV %s, %s;\n", dst_str
, dst_name
);
2340 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in ARB:
2341 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
2342 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
2343 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
2345 static void pshader_hw_texm3x2depth(const struct wined3d_shader_instruction
*ins
)
2347 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2348 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2349 char src0
[50], dst_name
[50];
2351 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2352 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2354 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2355 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2356 shader_addline(buffer
, "DP3 %s.y, fragment.texcoord[%u], %s;\n", dst_name
, dst
->reg
.idx
, src0
);
2358 /* How to deal with the special case dst_name.g == 0? if r != 0, then
2359 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
2360 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
2362 shader_addline(buffer
, "RCP %s.y, %s.y;\n", dst_name
, dst_name
);
2363 shader_addline(buffer
, "MUL %s.x, %s.x, %s.y;\n", dst_name
, dst_name
, dst_name
);
2364 shader_addline(buffer
, "MIN %s.x, %s.x, %s;\n", dst_name
, dst_name
, one
);
2365 shader_addline(buffer
, "MAX result.depth, %s.x, %s;\n", dst_name
, zero
);
2368 /** Handles transforming all WINED3DSIO_M?x? opcodes for
2369 Vertex/Pixel shaders to ARB_vertex_program codes */
2370 static void shader_hw_mnxn(const struct wined3d_shader_instruction
*ins
)
2373 int nComponents
= 0;
2374 struct wined3d_shader_dst_param tmp_dst
= {{0}};
2375 struct wined3d_shader_src_param tmp_src
[2] = {{{0}}};
2376 struct wined3d_shader_instruction tmp_ins
;
2378 memset(&tmp_ins
, 0, sizeof(tmp_ins
));
2380 /* Set constants for the temporary argument */
2381 tmp_ins
.ctx
= ins
->ctx
;
2382 tmp_ins
.dst_count
= 1;
2383 tmp_ins
.dst
= &tmp_dst
;
2384 tmp_ins
.src_count
= 2;
2385 tmp_ins
.src
= tmp_src
;
2387 switch(ins
->handler_idx
)
2389 case WINED3DSIH_M4x4
:
2391 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2393 case WINED3DSIH_M4x3
:
2395 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2397 case WINED3DSIH_M3x4
:
2399 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2401 case WINED3DSIH_M3x3
:
2403 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2405 case WINED3DSIH_M3x2
:
2407 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2410 FIXME("Unhandled opcode %#x\n", ins
->handler_idx
);
2414 tmp_dst
= ins
->dst
[0];
2415 tmp_src
[0] = ins
->src
[0];
2416 tmp_src
[1] = ins
->src
[1];
2417 for (i
= 0; i
< nComponents
; i
++) {
2418 tmp_dst
.write_mask
= WINED3DSP_WRITEMASK_0
<< i
;
2419 shader_hw_map2gl(&tmp_ins
);
2420 ++tmp_src
[1].reg
.idx
;
2424 static void shader_hw_rcp(const struct wined3d_shader_instruction
*ins
)
2426 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2431 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
); /* Destination */
2432 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src
);
2433 if (ins
->src
[0].swizzle
== WINED3DSP_NOSWIZZLE
)
2435 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
2441 shader_addline(buffer
, "RCP%s %s, %s;\n", shader_arb_get_modifier(ins
), dst
, src
);
2444 static void shader_hw_scalar_op(const struct wined3d_shader_instruction
*ins
)
2446 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2447 const char *instruction
;
2452 switch(ins
->handler_idx
)
2454 case WINED3DSIH_RSQ
: instruction
= "RSQ"; break;
2455 case WINED3DSIH_RCP
: instruction
= "RCP"; break;
2456 case WINED3DSIH_EXP
: instruction
= "EX2"; break;
2457 case WINED3DSIH_EXPP
: instruction
= "EXP"; break;
2458 default: instruction
= "";
2459 FIXME("Unhandled opcode %#x\n", ins
->handler_idx
);
2463 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
); /* Destination */
2464 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src
);
2465 if (ins
->src
[0].swizzle
== WINED3DSP_NOSWIZZLE
)
2467 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
2473 shader_addline(buffer
, "%s%s %s, %s;\n", instruction
, shader_arb_get_modifier(ins
), dst
, src
);
2476 static void shader_hw_nrm(const struct wined3d_shader_instruction
*ins
)
2478 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2481 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2482 BOOL pshader
= shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2483 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2485 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2486 shader_arb_get_src_param(ins
, &ins
->src
[0], 1 /* Use TB */, src_name
);
2488 /* In D3D, NRM of a vector with length zero returns zero. Catch this situation, as
2489 * otherwise NRM or RSQ would return NaN */
2490 if(pshader
&& priv
->target_version
>= NV3
)
2492 /* GL_NV_fragment_program2's NRM needs protection against length zero vectors too
2494 * TODO: Find out if DP3+NRM+MOV is really faster than DP3+RSQ+MUL
2496 shader_addline(buffer
, "DP3C TA, %s, %s;\n", src_name
, src_name
);
2497 shader_addline(buffer
, "NRM%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
);
2498 shader_addline(buffer
, "MOV %s (EQ), %s;\n", dst_name
, zero
);
2500 else if(priv
->target_version
>= NV2
)
2502 shader_addline(buffer
, "DP3C TA.x, %s, %s;\n", src_name
, src_name
);
2503 shader_addline(buffer
, "RSQ TA.x (NE), TA.x;\n");
2504 shader_addline(buffer
, "MUL%s %s, %s, TA.x;\n", shader_arb_get_modifier(ins
), dst_name
,
2509 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2511 shader_addline(buffer
, "DP3 TA.x, %s, %s;\n", src_name
, src_name
);
2512 /* Pass any non-zero value to RSQ if the input vector has a length of zero. The
2513 * RSQ result doesn't matter, as long as multiplying it by 0 returns 0.
2515 shader_addline(buffer
, "SGE TA.y, -TA.x, %s;\n", zero
);
2516 shader_addline(buffer
, "MAD TA.x, %s, TA.y, TA.x;\n", one
);
2518 shader_addline(buffer
, "RSQ TA.x, TA.x;\n");
2519 /* dst.w = src[0].w * 1 / (src.x^2 + src.y^2 + src.z^2)^(1/2) according to msdn*/
2520 shader_addline(buffer
, "MUL%s %s, %s, TA.x;\n", shader_arb_get_modifier(ins
), dst_name
,
2525 static void shader_hw_lrp(const struct wined3d_shader_instruction
*ins
)
2527 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2529 char src_name
[3][50];
2531 /* ARB_fragment_program has a convenient LRP instruction */
2532 if(shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)) {
2533 shader_hw_map2gl(ins
);
2537 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2538 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
2539 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
2540 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
2542 shader_addline(buffer
, "SUB TA, %s, %s;\n", src_name
[1], src_name
[2]);
2543 shader_addline(buffer
, "MAD%s %s, %s, TA, %s;\n", shader_arb_get_modifier(ins
),
2544 dst_name
, src_name
[0], src_name
[2]);
2547 static void shader_hw_sincos(const struct wined3d_shader_instruction
*ins
)
2549 /* This instruction exists in ARB, but the d3d instruction takes two extra parameters which
2550 * must contain fixed constants. So we need a separate function to filter those constants and
2553 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2554 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2555 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2557 char src_name0
[50], src_name1
[50], src_name2
[50];
2560 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
2561 if(shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)) {
2562 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2563 /* No modifiers are supported on SCS */
2564 shader_addline(buffer
, "SCS %s, %s;\n", dst_name
, src_name0
);
2566 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
)
2568 shader_arb_get_register_name(ins
, &dst
->reg
, src_name0
, &is_color
);
2569 shader_addline(buffer
, "MOV_SAT %s, %s;\n", dst_name
, src_name0
);
2571 } else if(priv
->target_version
>= NV2
) {
2572 shader_arb_get_register_name(ins
, &dst
->reg
, dst_name
, &is_color
);
2574 /* Sincos writemask must be .x, .y or .xy */
2575 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
)
2576 shader_addline(buffer
, "COS%s %s.x, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name0
);
2577 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
)
2578 shader_addline(buffer
, "SIN%s %s.y, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name0
);
2580 /* Approximate sine and cosine with a taylor series, as per math textbook. The application passes 8
2581 * helper constants(D3DSINCOSCONST1 and D3DSINCOSCONST2) in src1 and src2.
2583 * sin(x) = x - x^3/3! + x^5/5! - x^7/7! + ...
2584 * cos(x) = 1 - x^2/2! + x^4/4! - x^6/6! + ...
2586 * The constants we get are:
2588 * +1 +1, -1 -1 +1 +1 -1 -1
2589 * ---- , ---- , ---- , ----- , ----- , ----- , ------
2590 * 1!*2 2!*4 3!*8 4!*16 5!*32 6!*64 7!*128
2592 * If used with x^2, x^3, x^4 etc they calculate sin(x/2) and cos(x/2):
2596 * (x/2)^4 = x^4 / 16
2597 * (x/2)^5 = x^5 / 32
2600 * To get the final result:
2601 * sin(x) = 2 * sin(x/2) * cos(x/2)
2602 * cos(x) = cos(x/2)^2 - sin(x/2)^2
2603 * (from sin(x+y) and cos(x+y) rules)
2605 * As per MSDN, dst.z is undefined after the operation, and so is
2606 * dst.x and dst.y if they're masked out by the writemask. Ie
2607 * sincos dst.y, src1, c0, c1
2608 * returns the sine in dst.y. dst.x and dst.z are undefined, dst.w is not touched. The assembler
2609 * vsa.exe also stops with an error if the dest register is the same register as the source
2610 * register. This means we can use dest.xyz as temporary storage. The assembler vsa.exe output also
2611 * indicates that sincos consumes 8 instruction slots in vs_2_0(and, strangely, in vs_3_0).
2613 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
2614 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name2
);
2615 shader_arb_get_register_name(ins
, &dst
->reg
, dst_name
, &is_color
);
2617 shader_addline(buffer
, "MUL %s.x, %s, %s;\n", dst_name
, src_name0
, src_name0
); /* x ^ 2 */
2618 shader_addline(buffer
, "MUL TA.y, %s.x, %s;\n", dst_name
, src_name0
); /* x ^ 3 */
2619 shader_addline(buffer
, "MUL %s.y, TA.y, %s;\n", dst_name
, src_name0
); /* x ^ 4 */
2620 shader_addline(buffer
, "MUL TA.z, %s.y, %s;\n", dst_name
, src_name0
); /* x ^ 5 */
2621 shader_addline(buffer
, "MUL %s.z, TA.z, %s;\n", dst_name
, src_name0
); /* x ^ 6 */
2622 shader_addline(buffer
, "MUL TA.w, %s.z, %s;\n", dst_name
, src_name0
); /* x ^ 7 */
2626 * Unfortunately we don't get the constants in a DP4-capable form. Is there a way to
2627 * properly merge that with MULs in the code above?
2628 * The swizzles .yz and xw however fit into the .yzxw swizzle added to ps_2_0. Maybe
2629 * we can merge the sine and cosine MAD rows to calculate them together.
2631 shader_addline(buffer
, "MUL TA.x, %s, %s.w;\n", src_name0
, src_name2
); /* x^1, +1/(1!*2) */
2632 shader_addline(buffer
, "MAD TA.x, TA.y, %s.x, TA.x;\n", src_name2
); /* -1/(3!*8) */
2633 shader_addline(buffer
, "MAD TA.x, TA.z, %s.w, TA.x;\n", src_name1
); /* +1/(5!*32) */
2634 shader_addline(buffer
, "MAD TA.x, TA.w, %s.x, TA.x;\n", src_name1
); /* -1/(7!*128) */
2637 shader_addline(buffer
, "MAD TA.y, %s.x, %s.y, %s.z;\n", dst_name
, src_name2
, src_name2
); /* -1/(2!*4), +1.0 */
2638 shader_addline(buffer
, "MAD TA.y, %s.y, %s.z, TA.y;\n", dst_name
, src_name1
); /* +1/(4!*16) */
2639 shader_addline(buffer
, "MAD TA.y, %s.z, %s.y, TA.y;\n", dst_name
, src_name1
); /* -1/(6!*64) */
2641 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
) {
2643 shader_addline(buffer
, "MUL TA.z, TA.y, TA.y;\n");
2644 shader_addline(buffer
, "MAD %s.x, -TA.x, TA.x, TA.z;\n", dst_name
);
2646 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
) {
2648 shader_addline(buffer
, "MUL %s.y, TA.x, TA.y;\n", dst_name
);
2649 shader_addline(buffer
, "ADD %s.y, %s.y, %s.y;\n", dst_name
, dst_name
, dst_name
);
2654 static void shader_hw_sgn(const struct wined3d_shader_instruction
*ins
)
2656 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2659 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
2661 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2662 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
);
2664 /* SGN is only valid in vertex shaders */
2665 if(ctx
->target_version
>= NV2
) {
2666 shader_addline(buffer
, "SSG%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
);
2670 /* If SRC > 0.0, -SRC < SRC = TRUE, otherwise false.
2671 * if SRC < 0.0, SRC < -SRC = TRUE. If neither is true, src = 0.0
2673 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
) {
2674 shader_addline(buffer
, "SLT %s, -%s, %s;\n", dst_name
, src_name
, src_name
);
2676 /* src contains TA? Write to the dest first. This won't overwrite our destination.
2677 * Then use TA, and calculate the final result
2679 * Not reading from TA? Store the first result in TA to avoid overwriting the
2680 * destination if src reg = dst reg
2682 if(strstr(src_name
, "TA"))
2684 shader_addline(buffer
, "SLT %s, %s, -%s;\n", dst_name
, src_name
, src_name
);
2685 shader_addline(buffer
, "SLT TA, -%s, %s;\n", src_name
, src_name
);
2686 shader_addline(buffer
, "ADD %s, %s, -TA;\n", dst_name
, dst_name
);
2690 shader_addline(buffer
, "SLT TA, -%s, %s;\n", src_name
, src_name
);
2691 shader_addline(buffer
, "SLT %s, %s, -%s;\n", dst_name
, src_name
, src_name
);
2692 shader_addline(buffer
, "ADD %s, TA, -%s;\n", dst_name
, dst_name
);
2697 static void shader_hw_dsy(const struct wined3d_shader_instruction
*ins
)
2699 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2705 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
);
2706 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src
);
2707 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2709 shader_addline(buffer
, "DDY %s, %s;\n", dst
, src
);
2710 shader_addline(buffer
, "MUL%s %s, %s, ycorrection.y;\n", shader_arb_get_modifier(ins
), dst
, dst_name
);
2713 static DWORD
abs_modifier(DWORD mod
, BOOL
*need_abs
)
2719 case WINED3DSPSM_NONE
: return WINED3DSPSM_ABS
;
2720 case WINED3DSPSM_NEG
: return WINED3DSPSM_ABS
;
2721 case WINED3DSPSM_BIAS
: *need_abs
= TRUE
; return WINED3DSPSM_BIAS
;
2722 case WINED3DSPSM_BIASNEG
: *need_abs
= TRUE
; return WINED3DSPSM_BIASNEG
;
2723 case WINED3DSPSM_SIGN
: *need_abs
= TRUE
; return WINED3DSPSM_SIGN
;
2724 case WINED3DSPSM_SIGNNEG
: *need_abs
= TRUE
; return WINED3DSPSM_SIGNNEG
;
2725 case WINED3DSPSM_COMP
: *need_abs
= TRUE
; return WINED3DSPSM_COMP
;
2726 case WINED3DSPSM_X2
: *need_abs
= TRUE
; return WINED3DSPSM_X2
;
2727 case WINED3DSPSM_X2NEG
: *need_abs
= TRUE
; return WINED3DSPSM_X2NEG
;
2728 case WINED3DSPSM_DZ
: *need_abs
= TRUE
; return WINED3DSPSM_DZ
;
2729 case WINED3DSPSM_DW
: *need_abs
= TRUE
; return WINED3DSPSM_DW
;
2730 case WINED3DSPSM_ABS
: return WINED3DSPSM_ABS
;
2731 case WINED3DSPSM_ABSNEG
: return WINED3DSPSM_ABS
;
2733 FIXME("Unknown modifier %u\n", mod
);
2737 static void shader_hw_log(const struct wined3d_shader_instruction
*ins
)
2739 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2740 char src0
[50], dst
[50];
2741 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
2742 BOOL need_abs
= FALSE
;
2745 switch(ins
->handler_idx
)
2747 case WINED3DSIH_LOG
: instr
= "LG2"; break;
2748 case WINED3DSIH_LOGP
: instr
= "LOG"; break;
2750 ERR("Unexpected instruction %d\n", ins
->handler_idx
);
2754 /* LOG and LOGP operate on the absolute value of the input */
2755 src0_copy
.modifiers
= abs_modifier(src0_copy
.modifiers
, &need_abs
);
2757 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
);
2758 shader_arb_get_src_param(ins
, &src0_copy
, 0, src0
);
2762 shader_addline(buffer
, "ABS TA, %s;\n", src0
);
2763 shader_addline(buffer
, "%s%s %s, TA;\n", instr
, shader_arb_get_modifier(ins
), dst
);
2767 shader_addline(buffer
, "%s%s %s, %s;\n", instr
, shader_arb_get_modifier(ins
), dst
, src0
);
2771 static void shader_hw_pow(const struct wined3d_shader_instruction
*ins
)
2773 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2774 char src0
[50], src1
[50], dst
[50];
2775 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
2776 BOOL need_abs
= FALSE
;
2777 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2778 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2780 /* POW operates on the absolute value of the input */
2781 src0_copy
.modifiers
= abs_modifier(src0_copy
.modifiers
, &need_abs
);
2783 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
);
2784 shader_arb_get_src_param(ins
, &src0_copy
, 0, src0
);
2785 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src1
);
2788 shader_addline(buffer
, "ABS TA.x, %s;\n", src0
);
2790 shader_addline(buffer
, "MOV TA.x, %s;\n", src0
);
2792 if (priv
->target_version
>= NV2
)
2794 shader_addline(buffer
, "MOVC TA.y, %s;\n", src1
);
2795 shader_addline(buffer
, "POW%s %s, TA.x, TA.y;\n", shader_arb_get_modifier(ins
), dst
);
2796 shader_addline(buffer
, "MOV %s (EQ.y), %s;\n", dst
, one
);
2800 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2801 const char *flt_eps
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_EPS
);
2803 shader_addline(buffer
, "ABS TA.y, %s;\n", src1
);
2804 shader_addline(buffer
, "SGE TA.y, -TA.y, %s;\n", zero
);
2805 /* Possibly add flt_eps to avoid getting float special values */
2806 shader_addline(buffer
, "MAD TA.z, TA.y, %s, %s;\n", flt_eps
, src1
);
2807 shader_addline(buffer
, "POW%s TA.x, TA.x, TA.z;\n", shader_arb_get_modifier(ins
));
2808 shader_addline(buffer
, "MAD TA.x, -TA.x, TA.y, TA.x;\n");
2809 shader_addline(buffer
, "MAD %s, TA.y, %s, TA.x;\n", dst
, one
);
2813 static void shader_hw_loop(const struct wined3d_shader_instruction
*ins
)
2815 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2817 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2820 shader_arb_get_src_param(ins
, &ins
->src
[1], 0, src_name
);
2824 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2825 struct list
*e
= list_head(&priv
->control_frames
);
2826 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2828 if(priv
->loop_depth
> 1) shader_addline(buffer
, "PUSHA aL;\n");
2829 /* The constant loader makes sure to load -1 into iX.w */
2830 shader_addline(buffer
, "ARLC aL, %s.xywz;\n", src_name
);
2831 shader_addline(buffer
, "BRA loop_%u_end (LE.x);\n", control_frame
->no
.loop
);
2832 shader_addline(buffer
, "loop_%u_start:\n", control_frame
->no
.loop
);
2836 shader_addline(buffer
, "LOOP %s;\n", src_name
);
2840 static void shader_hw_rep(const struct wined3d_shader_instruction
*ins
)
2842 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2844 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2846 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
);
2848 /* The constant loader makes sure to load -1 into iX.w */
2851 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2852 struct list
*e
= list_head(&priv
->control_frames
);
2853 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2855 if(priv
->loop_depth
> 1) shader_addline(buffer
, "PUSHA aL;\n");
2857 shader_addline(buffer
, "ARLC aL, %s.xywz;\n", src_name
);
2858 shader_addline(buffer
, "BRA loop_%u_end (LE.x);\n", control_frame
->no
.loop
);
2859 shader_addline(buffer
, "loop_%u_start:\n", control_frame
->no
.loop
);
2863 shader_addline(buffer
, "REP %s;\n", src_name
);
2867 static void shader_hw_endloop(const struct wined3d_shader_instruction
*ins
)
2869 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2870 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2874 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2875 struct list
*e
= list_head(&priv
->control_frames
);
2876 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2878 shader_addline(buffer
, "ARAC aL.xy, aL;\n");
2879 shader_addline(buffer
, "BRA loop_%u_start (GT.x);\n", control_frame
->no
.loop
);
2880 shader_addline(buffer
, "loop_%u_end:\n", control_frame
->no
.loop
);
2882 if(priv
->loop_depth
> 1) shader_addline(buffer
, "POPA aL;\n");
2886 shader_addline(buffer
, "ENDLOOP;\n");
2890 static void shader_hw_endrep(const struct wined3d_shader_instruction
*ins
)
2892 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2893 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2897 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2898 struct list
*e
= list_head(&priv
->control_frames
);
2899 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2901 shader_addline(buffer
, "ARAC aL.xy, aL;\n");
2902 shader_addline(buffer
, "BRA loop_%u_start (GT.x);\n", control_frame
->no
.loop
);
2903 shader_addline(buffer
, "loop_%u_end:\n", control_frame
->no
.loop
);
2905 if(priv
->loop_depth
> 1) shader_addline(buffer
, "POPA aL;\n");
2909 shader_addline(buffer
, "ENDREP;\n");
2913 static const struct control_frame
*find_last_loop(const struct shader_arb_ctx_priv
*priv
)
2915 struct control_frame
*control_frame
;
2917 LIST_FOR_EACH_ENTRY(control_frame
, &priv
->control_frames
, struct control_frame
, entry
)
2919 if(control_frame
->type
== LOOP
|| control_frame
->type
== REP
) return control_frame
;
2921 ERR("Could not find loop for break\n");
2925 static void shader_hw_break(const struct wined3d_shader_instruction
*ins
)
2927 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2928 const struct control_frame
*control_frame
= find_last_loop(ins
->ctx
->backend_data
);
2929 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2933 shader_addline(buffer
, "BRA loop_%u_end;\n", control_frame
->no
.loop
);
2937 shader_addline(buffer
, "BRK;\n");
2941 static const char *get_compare(COMPARISON_TYPE flags
)
2945 case COMPARISON_GT
: return "GT";
2946 case COMPARISON_EQ
: return "EQ";
2947 case COMPARISON_GE
: return "GE";
2948 case COMPARISON_LT
: return "LT";
2949 case COMPARISON_NE
: return "NE";
2950 case COMPARISON_LE
: return "LE";
2952 FIXME("Unrecognized comparison value: %u\n", flags
);
2957 static COMPARISON_TYPE
invert_compare(COMPARISON_TYPE flags
)
2961 case COMPARISON_GT
: return COMPARISON_LE
;
2962 case COMPARISON_EQ
: return COMPARISON_NE
;
2963 case COMPARISON_GE
: return COMPARISON_LT
;
2964 case COMPARISON_LT
: return COMPARISON_GE
;
2965 case COMPARISON_NE
: return COMPARISON_EQ
;
2966 case COMPARISON_LE
: return COMPARISON_GT
;
2968 FIXME("Unrecognized comparison value: %u\n", flags
);
2973 static void shader_hw_breakc(const struct wined3d_shader_instruction
*ins
)
2975 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2976 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2977 const struct control_frame
*control_frame
= find_last_loop(ins
->ctx
->backend_data
);
2980 const char *comp
= get_compare(ins
->flags
);
2982 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
2983 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
2987 /* SUBC CC, src0, src1" works only in pixel shaders, so use TA to throw
2988 * away the subtraction result
2990 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
2991 shader_addline(buffer
, "BRA loop_%u_end (%s.x);\n", control_frame
->no
.loop
, comp
);
2995 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
2996 shader_addline(buffer
, "BRK (%s.x);\n", comp
);
3000 static void shader_hw_ifc(const struct wined3d_shader_instruction
*ins
)
3002 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3003 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3004 struct list
*e
= list_head(&priv
->control_frames
);
3005 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3009 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3011 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
3012 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
3016 /* Invert the flag. We jump to the else label if the condition is NOT true */
3017 comp
= get_compare(invert_compare(ins
->flags
));
3018 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3019 shader_addline(buffer
, "BRA ifc_%u_else (%s.x);\n", control_frame
->no
.ifc
, comp
);
3023 comp
= get_compare(ins
->flags
);
3024 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3025 shader_addline(buffer
, "IF %s.x;\n", comp
);
3029 static void shader_hw_else(const struct wined3d_shader_instruction
*ins
)
3031 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3032 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3033 struct list
*e
= list_head(&priv
->control_frames
);
3034 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3035 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3039 shader_addline(buffer
, "BRA ifc_%u_endif;\n", control_frame
->no
.ifc
);
3040 shader_addline(buffer
, "ifc_%u_else:\n", control_frame
->no
.ifc
);
3041 control_frame
->had_else
= TRUE
;
3045 shader_addline(buffer
, "ELSE;\n");
3049 static void shader_hw_endif(const struct wined3d_shader_instruction
*ins
)
3051 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3052 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3053 struct list
*e
= list_head(&priv
->control_frames
);
3054 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3055 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3059 if(control_frame
->had_else
)
3061 shader_addline(buffer
, "ifc_%u_endif:\n", control_frame
->no
.ifc
);
3065 shader_addline(buffer
, "#No else branch. else is endif\n");
3066 shader_addline(buffer
, "ifc_%u_else:\n", control_frame
->no
.ifc
);
3071 shader_addline(buffer
, "ENDIF;\n");
3075 static void shader_hw_texldd(const struct wined3d_shader_instruction
*ins
)
3077 DWORD sampler_idx
= ins
->src
[1].reg
.idx
;
3079 char reg_src
[3][40];
3080 WORD flags
= TEX_DERIV
;
3082 shader_arb_get_dst_param(ins
, &ins
->dst
[0], reg_dest
);
3083 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_src
[0]);
3084 shader_arb_get_src_param(ins
, &ins
->src
[2], 1, reg_src
[1]);
3085 shader_arb_get_src_param(ins
, &ins
->src
[3], 2, reg_src
[2]);
3087 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) flags
|= TEX_PROJ
;
3088 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) flags
|= TEX_BIAS
;
3090 shader_hw_sample(ins
, sampler_idx
, reg_dest
, reg_src
[0], flags
, reg_src
[1], reg_src
[2]);
3093 static void shader_hw_texldl(const struct wined3d_shader_instruction
*ins
)
3095 DWORD sampler_idx
= ins
->src
[1].reg
.idx
;
3098 WORD flags
= TEX_LOD
;
3100 shader_arb_get_dst_param(ins
, &ins
->dst
[0], reg_dest
);
3101 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_coord
);
3103 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) flags
|= TEX_PROJ
;
3104 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) flags
|= TEX_BIAS
;
3106 shader_hw_sample(ins
, sampler_idx
, reg_dest
, reg_coord
, flags
, NULL
, NULL
);
3109 static void shader_hw_label(const struct wined3d_shader_instruction
*ins
)
3111 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3112 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3114 priv
->in_main_func
= FALSE
;
3115 /* Call instructions activate the NV extensions, not labels and rets. If there is an uncalled
3116 * subroutine, don't generate a label that will make GL complain
3118 if(priv
->target_version
== ARB
) return;
3120 shader_addline(buffer
, "l%u:\n", ins
->src
[0].reg
.idx
);
3123 static void vshader_add_footer(struct shader_arb_ctx_priv
*priv_ctx
,
3124 const struct arb_vshader_private
*shader_data
, const struct arb_vs_compile_args
*args
,
3125 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
,
3126 struct wined3d_shader_buffer
*buffer
)
3130 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
3131 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
3132 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
3133 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
3135 if (args
->super
.fog_src
== VS_FOG_Z
)
3136 shader_addline(buffer
, "MOV result.fogcoord, TMP_OUT.z;\n");
3137 else if (!reg_maps
->fog
)
3138 /* posFixup.x is always 1.0, so we can safely use it */
3139 shader_addline(buffer
, "ADD result.fogcoord, posFixup.x, -posFixup.x;\n");
3141 /* Clipplanes are always stored without y inversion */
3142 if (use_nv_clip(gl_info
) && priv_ctx
->target_version
>= NV2
)
3144 if (args
->super
.clip_enabled
)
3146 for (i
= 0; i
< priv_ctx
->vs_clipplanes
; i
++)
3148 shader_addline(buffer
, "DP4 result.clip[%u].x, TMP_OUT, state.clip[%u].plane;\n", i
, i
);
3152 else if (args
->clip
.boolclip
.clip_texcoord
)
3154 unsigned int cur_clip
= 0;
3155 char component
[4] = {'x', 'y', 'z', 'w'};
3156 const char *zero
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_ZERO
);
3158 for (i
= 0; i
< gl_info
->limits
.clipplanes
; ++i
)
3160 if (args
->clip
.boolclip
.clipplane_mask
& (1 << i
))
3162 shader_addline(buffer
, "DP4 TA.%c, TMP_OUT, state.clip[%u].plane;\n",
3163 component
[cur_clip
++], i
);
3169 shader_addline(buffer
, "MOV TA, %s;\n", zero
);
3172 shader_addline(buffer
, "MOV TA.yzw, %s;\n", zero
);
3175 shader_addline(buffer
, "MOV TA.zw, %s;\n", zero
);
3178 shader_addline(buffer
, "MOV TA.w, %s;\n", zero
);
3181 shader_addline(buffer
, "MOV result.texcoord[%u], TA;\n",
3182 args
->clip
.boolclip
.clip_texcoord
- 1);
3185 /* Write the final position.
3187 * OpenGL coordinates specify the center of the pixel while d3d coords specify
3188 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
3189 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
3190 * contains 1.0 to allow a mad, but arb vs swizzles are too restricted for that.
3192 shader_addline(buffer
, "MUL TA, posFixup, TMP_OUT.w;\n");
3193 shader_addline(buffer
, "ADD TMP_OUT.x, TMP_OUT.x, TA.z;\n");
3194 shader_addline(buffer
, "MAD TMP_OUT.y, TMP_OUT.y, posFixup.y, TA.w;\n");
3196 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
3197 * and the glsl equivalent
3199 if (need_helper_const(shader_data
, reg_maps
, gl_info
))
3201 const char *two
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_TWO
);
3202 shader_addline(buffer
, "MAD TMP_OUT.z, TMP_OUT.z, %s, -TMP_OUT.w;\n", two
);
3206 shader_addline(buffer
, "ADD TMP_OUT.z, TMP_OUT.z, TMP_OUT.z;\n");
3207 shader_addline(buffer
, "ADD TMP_OUT.z, TMP_OUT.z, -TMP_OUT.w;\n");
3210 shader_addline(buffer
, "MOV result.position, TMP_OUT;\n");
3212 priv_ctx
->footer_written
= TRUE
;
3215 static void shader_hw_ret(const struct wined3d_shader_instruction
*ins
)
3217 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3218 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3219 struct wined3d_shader
*shader
= ins
->ctx
->shader
;
3220 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3222 if(priv
->target_version
== ARB
) return;
3226 if (priv
->in_main_func
) vshader_add_footer(priv
, shader
->backend_data
,
3227 priv
->cur_vs_args
, ins
->ctx
->reg_maps
, ins
->ctx
->gl_info
, buffer
);
3230 shader_addline(buffer
, "RET;\n");
3233 static void shader_hw_call(const struct wined3d_shader_instruction
*ins
)
3235 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3236 shader_addline(buffer
, "CAL l%u;\n", ins
->src
[0].reg
.idx
);
3239 /* GL locking is done by the caller */
3240 static GLuint
create_arb_blt_vertex_program(const struct wined3d_gl_info
*gl_info
)
3242 GLuint program_id
= 0;
3245 const char *blt_vprogram
=
3247 "PARAM c[1] = { { 1, 0.5 } };\n"
3248 "MOV result.position, vertex.position;\n"
3249 "MOV result.color, c[0].x;\n"
3250 "MOV result.texcoord[0], vertex.texcoord[0];\n"
3253 GL_EXTCALL(glGenProgramsARB(1, &program_id
));
3254 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, program_id
));
3255 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
3256 strlen(blt_vprogram
), blt_vprogram
));
3257 checkGLcall("glProgramStringARB()");
3259 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
3262 FIXME("Vertex program error at position %d: %s\n\n", pos
,
3263 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
3264 shader_arb_dump_program_source(blt_vprogram
);
3270 GL_EXTCALL(glGetProgramivARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
3271 checkGLcall("glGetProgramivARB()");
3272 if (!native
) WARN("Program exceeds native resource limits.\n");
3278 /* GL locking is done by the caller */
3279 static GLuint
create_arb_blt_fragment_program(const struct wined3d_gl_info
*gl_info
,
3280 enum tex_types tex_type
, BOOL masked
)
3282 GLuint program_id
= 0;
3283 const char *fprogram
;
3286 static const char * const blt_fprograms_full
[tex_type_count
] =
3293 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
3294 "MOV result.depth.z, R0.x;\n"
3301 "TEX R0.x, fragment.texcoord[0], texture[0], CUBE;\n"
3302 "MOV result.depth.z, R0.x;\n"
3307 "TEX R0.x, fragment.texcoord[0], texture[0], RECT;\n"
3308 "MOV result.depth.z, R0.x;\n"
3312 static const char * const blt_fprograms_masked
[tex_type_count
] =
3318 "PARAM mask = program.local[0];\n"
3320 "SLT R0.xy, fragment.position, mask.zwzw;\n"
3321 "MUL R0.x, R0.x, R0.y;\n"
3323 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
3324 "MOV result.depth.z, R0.x;\n"
3330 "PARAM mask = program.local[0];\n"
3332 "SLT R0.xy, fragment.position, mask.zwzw;\n"
3333 "MUL R0.x, R0.x, R0.y;\n"
3335 "TEX R0.x, fragment.texcoord[0], texture[0], CUBE;\n"
3336 "MOV result.depth.z, R0.x;\n"
3340 "PARAM mask = program.local[0];\n"
3342 "SLT R0.xy, fragment.position, mask.zwzw;\n"
3343 "MUL R0.x, R0.x, R0.y;\n"
3345 "TEX R0.x, fragment.texcoord[0], texture[0], RECT;\n"
3346 "MOV result.depth.z, R0.x;\n"
3350 fprogram
= masked
? blt_fprograms_masked
[tex_type
] : blt_fprograms_full
[tex_type
];
3353 FIXME("tex_type %#x not supported, falling back to tex_2d\n", tex_type
);
3355 fprogram
= masked
? blt_fprograms_masked
[tex_type
] : blt_fprograms_full
[tex_type
];
3358 GL_EXTCALL(glGenProgramsARB(1, &program_id
));
3359 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, program_id
));
3360 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
, strlen(fprogram
), fprogram
));
3361 checkGLcall("glProgramStringARB()");
3363 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
3366 FIXME("Fragment program error at position %d: %s\n\n", pos
,
3367 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
3368 shader_arb_dump_program_source(fprogram
);
3374 GL_EXTCALL(glGetProgramivARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
3375 checkGLcall("glGetProgramivARB()");
3376 if (!native
) WARN("Program exceeds native resource limits.\n");
3382 static void arbfp_add_sRGB_correction(struct wined3d_shader_buffer
*buffer
, const char *fragcolor
,
3383 const char *tmp1
, const char *tmp2
, const char *tmp3
, const char *tmp4
, BOOL condcode
)
3385 /* Perform sRGB write correction. See GLX_EXT_framebuffer_sRGB */
3389 /* Sigh. MOVC CC doesn't work, so use one of the temps as dummy dest */
3390 shader_addline(buffer
, "SUBC %s, %s.x, srgb_consts1.y;\n", tmp1
, fragcolor
);
3391 /* Calculate the > 0.0031308 case */
3392 shader_addline(buffer
, "POW %s.x (GE), %s.x, srgb_consts1.z;\n", fragcolor
, fragcolor
);
3393 shader_addline(buffer
, "POW %s.y (GE), %s.y, srgb_consts1.z;\n", fragcolor
, fragcolor
);
3394 shader_addline(buffer
, "POW %s.z (GE), %s.z, srgb_consts1.z;\n", fragcolor
, fragcolor
);
3395 shader_addline(buffer
, "MUL %s.xyz (GE), %s, srgb_consts1.w;\n", fragcolor
, fragcolor
);
3396 shader_addline(buffer
, "SUB %s.xyz (GE), %s, srgb_consts2.x;\n", fragcolor
, fragcolor
);
3397 /* Calculate the < case */
3398 shader_addline(buffer
, "MUL %s.xyz (LT), srgb_consts1.x, %s;\n", fragcolor
, fragcolor
);
3402 /* Calculate the > 0.0031308 case */
3403 shader_addline(buffer
, "POW %s.x, %s.x, srgb_consts1.z;\n", tmp1
, fragcolor
);
3404 shader_addline(buffer
, "POW %s.y, %s.y, srgb_consts1.z;\n", tmp1
, fragcolor
);
3405 shader_addline(buffer
, "POW %s.z, %s.z, srgb_consts1.z;\n", tmp1
, fragcolor
);
3406 shader_addline(buffer
, "MUL %s, %s, srgb_consts1.w;\n", tmp1
, tmp1
);
3407 shader_addline(buffer
, "SUB %s, %s, srgb_consts2.x;\n", tmp1
, tmp1
);
3408 /* Calculate the < case */
3409 shader_addline(buffer
, "MUL %s, srgb_consts1.x, %s;\n", tmp2
, fragcolor
);
3410 /* Get 1.0 / 0.0 masks for > 0.0031308 and < 0.0031308 */
3411 shader_addline(buffer
, "SLT %s, srgb_consts1.y, %s;\n", tmp3
, fragcolor
);
3412 shader_addline(buffer
, "SGE %s, srgb_consts1.y, %s;\n", tmp4
, fragcolor
);
3413 /* Store the components > 0.0031308 in the destination */
3414 shader_addline(buffer
, "MUL %s.xyz, %s, %s;\n", fragcolor
, tmp1
, tmp3
);
3415 /* Add the components that are < 0.0031308 */
3416 shader_addline(buffer
, "MAD %s.xyz, %s, %s, %s;\n", fragcolor
, tmp2
, tmp4
, fragcolor
);
3417 /* Move everything into result.color at once. Nvidia hardware cannot handle partial
3418 * result.color writes(.rgb first, then .a), or handle overwriting already written
3419 * components. The assembler uses a temporary register in this case, which is usually
3420 * not allocated from one of our registers that were used earlier.
3423 /* [0.0;1.0] clamping. Not needed, this is done implicitly */
3426 static const DWORD
*find_loop_control_values(const struct wined3d_shader
*shader
, DWORD idx
)
3428 const local_constant
*constant
;
3430 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsI
, local_constant
, entry
)
3432 if (constant
->idx
== idx
)
3434 return constant
->value
;
3440 static void init_ps_input(const struct wined3d_shader
*shader
,
3441 const struct arb_ps_compile_args
*args
, struct shader_arb_ctx_priv
*priv
)
3443 static const char * const texcoords
[8] =
3445 "fragment.texcoord[0]", "fragment.texcoord[1]", "fragment.texcoord[2]", "fragment.texcoord[3]",
3446 "fragment.texcoord[4]", "fragment.texcoord[5]", "fragment.texcoord[6]", "fragment.texcoord[7]"
3449 const struct wined3d_shader_signature_element
*sig
= shader
->input_signature
;
3450 const char *semantic_name
;
3453 switch(args
->super
.vp_mode
)
3455 case pretransformed
:
3457 /* The pixelshader has to collect the varyings on its own. In any case properly load
3458 * color0 and color1. In the case of pretransformed vertices also load texcoords. Set
3459 * other attribs to 0.0.
3461 * For fixedfunction this behavior is correct, according to the tests. For pretransformed
3462 * we'd either need a replacement shader that can load other attribs like BINORMAL, or
3463 * load the texcoord attrib pointers to match the pixel shader signature
3465 for(i
= 0; i
< MAX_REG_INPUT
; i
++)
3467 semantic_name
= sig
[i
].semantic_name
;
3468 semantic_idx
= sig
[i
].semantic_idx
;
3469 if (!semantic_name
) continue;
3471 if(shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_COLOR
))
3473 if (!semantic_idx
) priv
->ps_input
[i
] = "fragment.color.primary";
3474 else if(semantic_idx
== 1) priv
->ps_input
[i
] = "fragment.color.secondary";
3475 else priv
->ps_input
[i
] = "0.0";
3477 else if(args
->super
.vp_mode
== fixedfunction
)
3479 priv
->ps_input
[i
] = "0.0";
3481 else if(shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_TEXCOORD
))
3483 if(semantic_idx
< 8) priv
->ps_input
[i
] = texcoords
[semantic_idx
];
3484 else priv
->ps_input
[i
] = "0.0";
3486 else if(shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_FOG
))
3488 if (!semantic_idx
) priv
->ps_input
[i
] = "fragment.fogcoord";
3489 else priv
->ps_input
[i
] = "0.0";
3493 priv
->ps_input
[i
] = "0.0";
3496 TRACE("v%u, semantic %s%u is %s\n", i
, semantic_name
, semantic_idx
, priv
->ps_input
[i
]);
3501 /* That one is easy. The vertex shaders provide v0-v7 in fragment.texcoord and v8 and v9 in
3504 for(i
= 0; i
< 8; i
++)
3506 priv
->ps_input
[i
] = texcoords
[i
];
3508 priv
->ps_input
[8] = "fragment.color.primary";
3509 priv
->ps_input
[9] = "fragment.color.secondary";
3514 /* GL locking is done by the caller */
3515 static GLuint
shader_arb_generate_pshader(struct wined3d_shader
*shader
,
3516 const struct wined3d_gl_info
*gl_info
, struct wined3d_shader_buffer
*buffer
,
3517 const struct arb_ps_compile_args
*args
, struct arb_ps_compiled_shader
*compiled
)
3519 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
3520 const DWORD
*function
= shader
->function
;
3521 const local_constant
*lconst
;
3524 DWORD
*lconst_map
= local_const_mapping(shader
), next_local
;
3525 struct shader_arb_ctx_priv priv_ctx
;
3526 BOOL dcl_td
= FALSE
;
3527 BOOL want_nv_prog
= FALSE
;
3528 struct arb_pshader_private
*shader_priv
= shader
->backend_data
;
3533 unsigned int i
, found
= 0;
3535 for (i
= 0, map
= reg_maps
->temporary
; map
; map
>>= 1, ++i
)
3538 || (shader
->u
.ps
.color0_mov
&& i
== shader
->u
.ps
.color0_reg
)
3539 || (reg_maps
->shader_version
.major
< 2 && !i
))
3542 sprintf(srgbtmp
[found
], "R%u", i
);
3544 if (found
== 4) break;
3549 sprintf(srgbtmp
[0], "TA");
3550 sprintf(srgbtmp
[1], "TB");
3551 sprintf(srgbtmp
[2], "TC");
3552 sprintf(srgbtmp
[3], "TD");
3556 sprintf(srgbtmp
[1], "TA");
3557 sprintf(srgbtmp
[2], "TB");
3558 sprintf(srgbtmp
[3], "TC");
3561 sprintf(srgbtmp
[2], "TA");
3562 sprintf(srgbtmp
[3], "TB");
3565 sprintf(srgbtmp
[3], "TA");
3571 /* Create the hw ARB shader */
3572 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
3573 priv_ctx
.cur_ps_args
= args
;
3574 priv_ctx
.compiled_fprog
= compiled
;
3575 priv_ctx
.cur_np2fixup_info
= &compiled
->np2fixup_info
;
3576 init_ps_input(shader
, args
, &priv_ctx
);
3577 list_init(&priv_ctx
.control_frames
);
3579 /* Avoid enabling NV_fragment_program* if we do not need it.
3581 * Enabling GL_NV_fragment_program_option causes the driver to occupy a temporary register,
3582 * and it slows down the shader execution noticeably(about 5%). Usually our instruction emulation
3583 * is faster than what we gain from using higher native instructions. There are some things though
3584 * that cannot be emulated. In that case enable the extensions.
3585 * If the extension is enabled, instruction handlers that support both ways will use it.
3587 * Testing shows no performance difference between OPTION NV_fragment_program2 and NV_fragment_program.
3588 * So enable the best we can get.
3590 if(reg_maps
->usesdsx
|| reg_maps
->usesdsy
|| reg_maps
->loop_depth
> 0 || reg_maps
->usestexldd
||
3591 reg_maps
->usestexldl
|| reg_maps
->usesfacing
|| reg_maps
->usesifc
|| reg_maps
->usescall
)
3593 want_nv_prog
= TRUE
;
3596 shader_addline(buffer
, "!!ARBfp1.0\n");
3597 if (want_nv_prog
&& gl_info
->supported
[NV_FRAGMENT_PROGRAM2
])
3599 shader_addline(buffer
, "OPTION NV_fragment_program2;\n");
3600 priv_ctx
.target_version
= NV3
;
3602 else if (want_nv_prog
&& gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
])
3604 shader_addline(buffer
, "OPTION NV_fragment_program;\n");
3605 priv_ctx
.target_version
= NV2
;
3609 /* This is an error - either we're advertising the wrong shader version, or aren't enforcing some
3612 ERR("The shader requires instructions that are not available in plain GL_ARB_fragment_program\n");
3615 priv_ctx
.target_version
= ARB
;
3618 if (reg_maps
->rt_mask
> 1)
3620 shader_addline(buffer
, "OPTION ARB_draw_buffers;\n");
3623 if (reg_maps
->shader_version
.major
< 3)
3625 switch(args
->super
.fog
) {
3629 shader_addline(buffer
, "OPTION ARB_fog_linear;\n");
3632 shader_addline(buffer
, "OPTION ARB_fog_exp;\n");
3635 shader_addline(buffer
, "OPTION ARB_fog_exp2;\n");
3640 /* For now always declare the temps. At least the Nvidia assembler optimizes completely
3641 * unused temps away(but occupies them for the whole shader if they're used once). Always
3642 * declaring them avoids tricky bookkeeping work
3644 shader_addline(buffer
, "TEMP TA;\n"); /* Used for modifiers */
3645 shader_addline(buffer
, "TEMP TB;\n"); /* Used for modifiers */
3646 shader_addline(buffer
, "TEMP TC;\n"); /* Used for modifiers */
3647 if(dcl_td
) shader_addline(buffer
, "TEMP TD;\n"); /* Used for sRGB writing */
3648 shader_addline(buffer
, "PARAM coefdiv = { 0.5, 0.25, 0.125, 0.0625 };\n");
3649 shader_addline(buffer
, "PARAM coefmul = { 2, 4, 8, 16 };\n");
3650 shader_addline(buffer
, "PARAM ps_helper_const = { 0.0, 1.0, %1.10f, 0.0 };\n", eps
);
3652 if (reg_maps
->shader_version
.major
< 2)
3654 strcpy(fragcolor
, "R0");
3658 if (args
->super
.srgb_correction
)
3660 if (shader
->u
.ps
.color0_mov
)
3662 sprintf(fragcolor
, "R%u", shader
->u
.ps
.color0_reg
);
3666 shader_addline(buffer
, "TEMP TMP_COLOR;\n");
3667 strcpy(fragcolor
, "TMP_COLOR");
3670 strcpy(fragcolor
, "result.color");
3674 if(args
->super
.srgb_correction
) {
3675 shader_addline(buffer
, "PARAM srgb_consts1 = {%f, %f, %f, %f};\n",
3676 srgb_mul_low
, srgb_cmp
, srgb_pow
, srgb_mul_high
);
3677 shader_addline(buffer
, "PARAM srgb_consts2 = {%f, %f, %f, %f};\n",
3678 srgb_sub_high
, 0.0, 0.0, 0.0);
3681 /* Base Declarations */
3682 next_local
= shader_generate_arb_declarations(shader
, reg_maps
,
3683 buffer
, gl_info
, lconst_map
, NULL
, &priv_ctx
);
3685 for (i
= 0, map
= reg_maps
->bumpmat
; map
; map
>>= 1, ++i
)
3687 unsigned char bump_const
;
3689 if (!(map
& 1)) continue;
3691 bump_const
= compiled
->numbumpenvmatconsts
;
3692 compiled
->bumpenvmatconst
[bump_const
].const_num
= WINED3D_CONST_NUM_UNUSED
;
3693 compiled
->bumpenvmatconst
[bump_const
].texunit
= i
;
3694 compiled
->luminanceconst
[bump_const
].const_num
= WINED3D_CONST_NUM_UNUSED
;
3695 compiled
->luminanceconst
[bump_const
].texunit
= i
;
3697 /* We can fit the constants into the constant limit for sure because texbem, texbeml, bem and beml are only supported
3698 * in 1.x shaders, and GL_ARB_fragment_program has a constant limit of 24 constants. So in the worst case we're loading
3699 * 8 shader constants, 8 bump matrices and 8 luminance parameters and are perfectly fine. (No NP2 fixup on bumpmapped
3700 * textures due to conditional NP2 restrictions)
3702 * Use local constants to load the bump env parameters, not program.env. This avoids collisions with d3d constants of
3703 * shaders in newer shader models. Since the bump env parameters have to share their space with NP2 fixup constants,
3704 * their location is shader dependent anyway and they cannot be loaded globally.
3706 compiled
->bumpenvmatconst
[bump_const
].const_num
= next_local
++;
3707 shader_addline(buffer
, "PARAM bumpenvmat%d = program.local[%d];\n",
3708 i
, compiled
->bumpenvmatconst
[bump_const
].const_num
);
3709 compiled
->numbumpenvmatconsts
= bump_const
+ 1;
3711 if (!(reg_maps
->luminanceparams
& (1 << i
))) continue;
3713 compiled
->luminanceconst
[bump_const
].const_num
= next_local
++;
3714 shader_addline(buffer
, "PARAM luminance%d = program.local[%d];\n",
3715 i
, compiled
->luminanceconst
[bump_const
].const_num
);
3718 for(i
= 0; i
< MAX_CONST_I
; i
++)
3720 compiled
->int_consts
[i
] = WINED3D_CONST_NUM_UNUSED
;
3721 if (reg_maps
->integer_constants
& (1 << i
) && priv_ctx
.target_version
>= NV2
)
3723 const DWORD
*control_values
= find_loop_control_values(shader
, i
);
3727 shader_addline(buffer
, "PARAM I%u = {%u, %u, %u, -1};\n", i
,
3728 control_values
[0], control_values
[1], control_values
[2]);
3732 compiled
->int_consts
[i
] = next_local
;
3733 compiled
->num_int_consts
++;
3734 shader_addline(buffer
, "PARAM I%u = program.local[%u];\n", i
, next_local
++);
3739 if(reg_maps
->vpos
|| reg_maps
->usesdsy
)
3741 compiled
->ycorrection
= next_local
;
3742 shader_addline(buffer
, "PARAM ycorrection = program.local[%u];\n", next_local
++);
3746 shader_addline(buffer
, "TEMP vpos;\n");
3747 /* ycorrection.x: Backbuffer height(onscreen) or 0(offscreen).
3748 * ycorrection.y: -1.0(onscreen), 1.0(offscreen)
3749 * ycorrection.z: 1.0
3750 * ycorrection.w: 0.0
3752 shader_addline(buffer
, "MAD vpos, fragment.position, ycorrection.zyww, ycorrection.wxww;\n");
3753 shader_addline(buffer
, "FLR vpos.xy, vpos;\n");
3758 compiled
->ycorrection
= WINED3D_CONST_NUM_UNUSED
;
3761 /* Load constants to fixup NP2 texcoords if there are still free constants left:
3762 * Constants (texture dimensions) for the NP2 fixup are loaded as local program parameters. This will consume
3763 * at most 8 (MAX_FRAGMENT_SAMPLERS / 2) parameters, which is highly unlikely, since the application had to
3764 * use 16 NP2 textures at the same time. In case that we run out of constants the fixup is simply not
3765 * applied / activated. This will probably result in wrong rendering of the texture, but will save us from
3766 * shader compilation errors and the subsequent errors when drawing with this shader. */
3767 if (priv_ctx
.cur_ps_args
->super
.np2_fixup
) {
3768 unsigned char cur_fixup_sampler
= 0;
3770 struct arb_ps_np2fixup_info
* const fixup
= priv_ctx
.cur_np2fixup_info
;
3771 const WORD map
= priv_ctx
.cur_ps_args
->super
.np2_fixup
;
3772 const UINT max_lconsts
= gl_info
->limits
.arb_ps_local_constants
;
3774 fixup
->offset
= next_local
;
3775 fixup
->super
.active
= 0;
3777 for (i
= 0; i
< MAX_FRAGMENT_SAMPLERS
; ++i
) {
3778 if (!(map
& (1 << i
))) continue;
3780 if (fixup
->offset
+ (cur_fixup_sampler
>> 1) < max_lconsts
) {
3781 fixup
->super
.active
|= (1 << i
);
3782 fixup
->super
.idx
[i
] = cur_fixup_sampler
++;
3784 FIXME("No free constant found to load NP2 fixup data into shader. "
3785 "Sampling from this texture will probably look wrong.\n");
3790 fixup
->super
.num_consts
= (cur_fixup_sampler
+ 1) >> 1;
3791 if (fixup
->super
.num_consts
) {
3792 shader_addline(buffer
, "PARAM np2fixup[%u] = { program.env[%u..%u] };\n",
3793 fixup
->super
.num_consts
, fixup
->offset
, fixup
->super
.num_consts
+ fixup
->offset
- 1);
3796 next_local
+= fixup
->super
.num_consts
;
3799 if (shader_priv
->clipplane_emulation
!= ~0U && args
->clip
)
3801 shader_addline(buffer
, "KIL fragment.texcoord[%u];\n", shader_priv
->clipplane_emulation
);
3804 /* Base Shader Body */
3805 shader_generate_main(shader
, buffer
, reg_maps
, function
, &priv_ctx
);
3807 if(args
->super
.srgb_correction
) {
3808 arbfp_add_sRGB_correction(buffer
, fragcolor
, srgbtmp
[0], srgbtmp
[1], srgbtmp
[2], srgbtmp
[3],
3809 priv_ctx
.target_version
>= NV2
);
3812 if(strcmp(fragcolor
, "result.color")) {
3813 shader_addline(buffer
, "MOV result.color, %s;\n", fragcolor
);
3815 shader_addline(buffer
, "END\n");
3817 /* TODO: change to resource.glObjectHandle or something like that */
3818 GL_EXTCALL(glGenProgramsARB(1, &retval
));
3820 TRACE("Creating a hw pixel shader, prg=%d\n", retval
);
3821 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, retval
));
3823 TRACE("Created hw pixel shader, prg=%d\n", retval
);
3824 /* Create the program and check for errors */
3825 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
3826 buffer
->bsize
, buffer
->buffer
));
3827 checkGLcall("glProgramStringARB()");
3829 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &errPos
);
3832 FIXME("HW PixelShader Error at position %d: %s\n\n",
3833 errPos
, debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
3834 shader_arb_dump_program_source(buffer
->buffer
);
3841 GL_EXTCALL(glGetProgramivARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
3842 checkGLcall("glGetProgramivARB()");
3843 if (!native
) WARN("Program exceeds native resource limits.\n");
3846 /* Load immediate constants */
3849 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, local_constant
, entry
)
3851 const float *value
= (const float *)lconst
->value
;
3852 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, lconst_map
[lconst
->idx
], value
));
3853 checkGLcall("glProgramLocalParameter4fvARB");
3855 HeapFree(GetProcessHeap(), 0, lconst_map
);
3861 static int compare_sig(const struct wined3d_shader_signature_element
*sig1
, const struct wined3d_shader_signature_element
*sig2
)
3866 for(i
= 0; i
< MAX_REG_INPUT
; i
++)
3868 if (!sig1
[i
].semantic_name
|| !sig2
[i
].semantic_name
)
3870 /* Compare pointers, not contents. One string is NULL(element does not exist), the other one is not NULL */
3871 if(sig1
[i
].semantic_name
!= sig2
[i
].semantic_name
) return sig1
[i
].semantic_name
< sig2
[i
].semantic_name
? -1 : 1;
3875 if ((ret
= strcmp(sig1
[i
].semantic_name
, sig2
[i
].semantic_name
))) return ret
;
3876 if(sig1
[i
].semantic_idx
!= sig2
[i
].semantic_idx
) return sig1
[i
].semantic_idx
< sig2
[i
].semantic_idx
? -1 : 1;
3877 if(sig1
[i
].sysval_semantic
!= sig2
[i
].sysval_semantic
) return sig1
[i
].sysval_semantic
< sig2
[i
].sysval_semantic
? -1 : 1;
3878 if(sig1
[i
].component_type
!= sig2
[i
].component_type
) return sig1
[i
].component_type
< sig2
[i
].component_type
? -1 : 1;
3879 if(sig1
[i
].register_idx
!= sig2
[i
].register_idx
) return sig1
[i
].register_idx
< sig2
[i
].register_idx
? -1 : 1;
3880 if(sig1
[i
].mask
!= sig2
->mask
) return sig1
[i
].mask
< sig2
[i
].mask
? -1 : 1;
3885 static struct wined3d_shader_signature_element
*clone_sig(const struct wined3d_shader_signature_element
*sig
)
3887 struct wined3d_shader_signature_element
*new;
3891 new = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*new) * MAX_REG_INPUT
);
3892 for(i
= 0; i
< MAX_REG_INPUT
; i
++)
3894 if (!sig
[i
].semantic_name
) continue;
3897 /* Clone the semantic string */
3898 name
= HeapAlloc(GetProcessHeap(), 0, strlen(sig
[i
].semantic_name
) + 1);
3899 strcpy(name
, sig
[i
].semantic_name
);
3900 new[i
].semantic_name
= name
;
3905 static DWORD
find_input_signature(struct shader_arb_priv
*priv
, const struct wined3d_shader_signature_element
*sig
)
3907 struct wine_rb_entry
*entry
= wine_rb_get(&priv
->signature_tree
, sig
);
3908 struct ps_signature
*found_sig
;
3912 found_sig
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
3913 TRACE("Found existing signature %u\n", found_sig
->idx
);
3914 return found_sig
->idx
;
3916 found_sig
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*sig
));
3917 found_sig
->sig
= clone_sig(sig
);
3918 found_sig
->idx
= priv
->ps_sig_number
++;
3919 TRACE("New signature stored and assigned number %u\n", found_sig
->idx
);
3920 if(wine_rb_put(&priv
->signature_tree
, sig
, &found_sig
->entry
) == -1)
3922 ERR("Failed to insert program entry.\n");
3924 return found_sig
->idx
;
3927 static void init_output_registers(struct wined3d_shader
*shader
, DWORD sig_num
,
3928 struct shader_arb_ctx_priv
*priv_ctx
, struct arb_vs_compiled_shader
*compiled
)
3931 static const char * const texcoords
[8] =
3933 "result.texcoord[0]", "result.texcoord[1]", "result.texcoord[2]", "result.texcoord[3]",
3934 "result.texcoord[4]", "result.texcoord[5]", "result.texcoord[6]", "result.texcoord[7]"
3936 struct wined3d_device
*device
= shader
->device
;
3937 const struct wined3d_shader_signature_element
*sig
;
3938 const char *semantic_name
;
3939 DWORD semantic_idx
, reg_idx
;
3941 /* Write generic input varyings 0 to 7 to result.texcoord[], varying 8 to result.color.primary
3942 * and varying 9 to result.color.secondary
3944 static const char * const decl_idx_to_string
[MAX_REG_INPUT
] =
3946 "result.texcoord[0]", "result.texcoord[1]", "result.texcoord[2]", "result.texcoord[3]",
3947 "result.texcoord[4]", "result.texcoord[5]", "result.texcoord[6]", "result.texcoord[7]",
3948 "result.color.primary", "result.color.secondary"
3953 TRACE("Pixel shader uses builtin varyings\n");
3954 /* Map builtins to builtins */
3955 for(i
= 0; i
< 8; i
++)
3957 priv_ctx
->texcrd_output
[i
] = texcoords
[i
];
3959 priv_ctx
->color_output
[0] = "result.color.primary";
3960 priv_ctx
->color_output
[1] = "result.color.secondary";
3961 priv_ctx
->fog_output
= "result.fogcoord";
3963 /* Map declared regs to builtins. Use "TA" to /dev/null unread output */
3964 for (i
= 0; i
< (sizeof(shader
->output_signature
) / sizeof(*shader
->output_signature
)); ++i
)
3966 semantic_name
= shader
->output_signature
[i
].semantic_name
;
3967 if (!semantic_name
) continue;
3969 if(shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_POSITION
))
3971 TRACE("o%u is TMP_OUT\n", i
);
3972 if (!shader
->output_signature
[i
].semantic_idx
) priv_ctx
->vs_output
[i
] = "TMP_OUT";
3973 else priv_ctx
->vs_output
[i
] = "TA";
3975 else if(shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_PSIZE
))
3977 TRACE("o%u is result.pointsize\n", i
);
3978 if (!shader
->output_signature
[i
].semantic_idx
) priv_ctx
->vs_output
[i
] = "result.pointsize";
3979 else priv_ctx
->vs_output
[i
] = "TA";
3981 else if(shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_COLOR
))
3983 TRACE("o%u is result.color.?, idx %u\n", i
, shader
->output_signature
[i
].semantic_idx
);
3984 if (!shader
->output_signature
[i
].semantic_idx
)
3985 priv_ctx
->vs_output
[i
] = "result.color.primary";
3986 else if (shader
->output_signature
[i
].semantic_idx
== 1)
3987 priv_ctx
->vs_output
[i
] = "result.color.secondary";
3988 else priv_ctx
->vs_output
[i
] = "TA";
3990 else if(shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_TEXCOORD
))
3992 TRACE("o%u is %s\n", i
, texcoords
[shader
->output_signature
[i
].semantic_idx
]);
3993 if (shader
->output_signature
[i
].semantic_idx
>= 8) priv_ctx
->vs_output
[i
] = "TA";
3994 else priv_ctx
->vs_output
[i
] = texcoords
[shader
->output_signature
[i
].semantic_idx
];
3996 else if(shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_FOG
))
3998 TRACE("o%u is result.fogcoord\n", i
);
3999 if (shader
->output_signature
[i
].semantic_idx
> 0) priv_ctx
->vs_output
[i
] = "TA";
4000 else priv_ctx
->vs_output
[i
] = "result.fogcoord";
4004 priv_ctx
->vs_output
[i
] = "TA";
4010 /* Instead of searching for the signature in the signature list, read the one from the current pixel shader.
4011 * Its maybe not the shader where the signature came from, but it is the same signature and faster to find
4013 sig
= device
->stateBlock
->state
.pixel_shader
->input_signature
;
4014 TRACE("Pixel shader uses declared varyings\n");
4016 /* Map builtin to declared. /dev/null the results by default to the TA temp reg */
4017 for(i
= 0; i
< 8; i
++)
4019 priv_ctx
->texcrd_output
[i
] = "TA";
4021 priv_ctx
->color_output
[0] = "TA";
4022 priv_ctx
->color_output
[1] = "TA";
4023 priv_ctx
->fog_output
= "TA";
4025 for(i
= 0; i
< MAX_REG_INPUT
; i
++)
4027 semantic_name
= sig
[i
].semantic_name
;
4028 semantic_idx
= sig
[i
].semantic_idx
;
4029 reg_idx
= sig
[i
].register_idx
;
4030 if (!semantic_name
) continue;
4032 /* If a declared input register is not written by builtin arguments, don't write to it.
4033 * GL_NV_vertex_program makes sure the input defaults to 0.0, which is correct with D3D
4035 * Don't care about POSITION and PSIZE here - this is a builtin vertex shader, position goes
4036 * to TMP_OUT in any case
4038 if(shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_TEXCOORD
))
4040 if(semantic_idx
< 8) priv_ctx
->texcrd_output
[semantic_idx
] = decl_idx_to_string
[reg_idx
];
4042 else if(shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_COLOR
))
4044 if(semantic_idx
< 2) priv_ctx
->color_output
[semantic_idx
] = decl_idx_to_string
[reg_idx
];
4046 else if(shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_FOG
))
4048 if (!semantic_idx
) priv_ctx
->fog_output
= decl_idx_to_string
[reg_idx
];
4055 if (!strcmp(decl_idx_to_string
[reg_idx
], "result.color.primary")
4056 || !strcmp(decl_idx_to_string
[reg_idx
], "result.color.secondary"))
4058 compiled
->need_color_unclamp
= TRUE
;
4062 /* Map declared to declared */
4063 for (i
= 0; i
< (sizeof(shader
->output_signature
) / sizeof(*shader
->output_signature
)); ++i
)
4065 /* Write unread output to TA to throw them away */
4066 priv_ctx
->vs_output
[i
] = "TA";
4067 semantic_name
= shader
->output_signature
[i
].semantic_name
;
4068 if (!semantic_name
) continue;
4070 if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_POSITION
)
4071 && !shader
->output_signature
[i
].semantic_idx
)
4073 priv_ctx
->vs_output
[i
] = "TMP_OUT";
4076 else if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_PSIZE
)
4077 && !shader
->output_signature
[i
].semantic_idx
)
4079 priv_ctx
->vs_output
[i
] = "result.pointsize";
4083 for(j
= 0; j
< MAX_REG_INPUT
; j
++)
4085 if (!sig
[j
].semantic_name
) continue;
4087 if (!strcmp(sig
[j
].semantic_name
, semantic_name
)
4088 && sig
[j
].semantic_idx
== shader
->output_signature
[i
].semantic_idx
)
4090 priv_ctx
->vs_output
[i
] = decl_idx_to_string
[sig
[j
].register_idx
];
4092 if (!strcmp(priv_ctx
->vs_output
[i
], "result.color.primary")
4093 || !strcmp(priv_ctx
->vs_output
[i
], "result.color.secondary"))
4095 compiled
->need_color_unclamp
= TRUE
;
4102 /* GL locking is done by the caller */
4103 static GLuint
shader_arb_generate_vshader(struct wined3d_shader
*shader
,
4104 const struct wined3d_gl_info
*gl_info
, struct wined3d_shader_buffer
*buffer
,
4105 const struct arb_vs_compile_args
*args
, struct arb_vs_compiled_shader
*compiled
)
4107 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
4108 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
4109 const DWORD
*function
= shader
->function
;
4110 const local_constant
*lconst
;
4112 DWORD next_local
, *lconst_map
= local_const_mapping(shader
);
4113 struct shader_arb_ctx_priv priv_ctx
;
4117 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
4118 priv_ctx
.cur_vs_args
= args
;
4119 list_init(&priv_ctx
.control_frames
);
4120 init_output_registers(shader
, args
->ps_signature
, &priv_ctx
, compiled
);
4122 /* Create the hw ARB shader */
4123 shader_addline(buffer
, "!!ARBvp1.0\n");
4125 /* Always enable the NV extension if available. Unlike fragment shaders, there is no
4126 * mesurable performance penalty, and we can always make use of it for clipplanes.
4128 if (gl_info
->supported
[NV_VERTEX_PROGRAM3
])
4130 shader_addline(buffer
, "OPTION NV_vertex_program3;\n");
4131 priv_ctx
.target_version
= NV3
;
4132 shader_addline(buffer
, "ADDRESS aL;\n");
4134 else if (gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
])
4136 shader_addline(buffer
, "OPTION NV_vertex_program2;\n");
4137 priv_ctx
.target_version
= NV2
;
4138 shader_addline(buffer
, "ADDRESS aL;\n");
4140 priv_ctx
.target_version
= ARB
;
4143 shader_addline(buffer
, "TEMP TMP_OUT;\n");
4144 if (need_helper_const(shader_data
, reg_maps
, gl_info
))
4146 shader_addline(buffer
, "PARAM helper_const = { 0.0, 1.0, 2.0, %1.10f};\n", eps
);
4148 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
))
4150 shader_addline(buffer
, "PARAM rel_addr_const = { 0.5, %d.0, 0.0, 0.0 };\n", shader_data
->rel_offset
);
4151 shader_addline(buffer
, "TEMP A0_SHADOW;\n");
4154 shader_addline(buffer
, "TEMP TA;\n");
4155 shader_addline(buffer
, "TEMP TB;\n");
4157 /* Base Declarations */
4158 next_local
= shader_generate_arb_declarations(shader
, reg_maps
, buffer
,
4159 gl_info
, lconst_map
, &priv_ctx
.vs_clipplanes
, &priv_ctx
);
4161 for(i
= 0; i
< MAX_CONST_I
; i
++)
4163 compiled
->int_consts
[i
] = WINED3D_CONST_NUM_UNUSED
;
4164 if(reg_maps
->integer_constants
& (1 << i
) && priv_ctx
.target_version
>= NV2
)
4166 const DWORD
*control_values
= find_loop_control_values(shader
, i
);
4170 shader_addline(buffer
, "PARAM I%u = {%u, %u, %u, -1};\n", i
,
4171 control_values
[0], control_values
[1], control_values
[2]);
4175 compiled
->int_consts
[i
] = next_local
;
4176 compiled
->num_int_consts
++;
4177 shader_addline(buffer
, "PARAM I%u = program.local[%u];\n", i
, next_local
++);
4182 /* We need a constant to fixup the final position */
4183 shader_addline(buffer
, "PARAM posFixup = program.local[%u];\n", next_local
);
4184 compiled
->pos_fixup
= next_local
++;
4186 /* Initialize output parameters. GL_ARB_vertex_program does not require special initialization values
4187 * for output parameters. D3D in theory does not do that either, but some applications depend on a
4188 * proper initialization of the secondary color, and programs using the fixed function pipeline without
4189 * a replacement shader depend on the texcoord.w being set properly.
4191 * GL_NV_vertex_program defines that all output values are initialized to {0.0, 0.0, 0.0, 1.0}. This
4192 * assertion is in effect even when using GL_ARB_vertex_program without any NV specific additions. So
4193 * skip this if NV_vertex_program is supported. Otherwise, initialize the secondary color. For the tex-
4194 * coords, we have a flag in the opengl caps. Many cards do not require the texcoord being set, and
4195 * this can eat a number of instructions, so skip it unless this cap is set as well
4197 if (!gl_info
->supported
[NV_VERTEX_PROGRAM
])
4199 struct wined3d_device
*device
= shader
->device
;
4200 const char *color_init
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_0001
);
4201 shader_addline(buffer
, "MOV result.color.secondary, %s;\n", color_init
);
4203 if (gl_info
->quirks
& WINED3D_QUIRK_SET_TEXCOORD_W
&& !device
->frag_pipe
->ffp_proj_control
)
4206 const char *one
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_ONE
);
4207 for(i
= 0; i
< min(8, MAX_REG_TEXCRD
); i
++)
4209 if (reg_maps
->texcoord_mask
[i
] && reg_maps
->texcoord_mask
[i
] != WINED3DSP_WRITEMASK_ALL
)
4210 shader_addline(buffer
, "MOV result.texcoord[%u].w, %s\n", i
, one
);
4215 /* The shader starts with the main function */
4216 priv_ctx
.in_main_func
= TRUE
;
4217 /* Base Shader Body */
4218 shader_generate_main(shader
, buffer
, reg_maps
, function
, &priv_ctx
);
4220 if (!priv_ctx
.footer_written
) vshader_add_footer(&priv_ctx
,
4221 shader_data
, args
, reg_maps
, gl_info
, buffer
);
4223 shader_addline(buffer
, "END\n");
4225 /* TODO: change to resource.glObjectHandle or something like that */
4226 GL_EXTCALL(glGenProgramsARB(1, &ret
));
4228 TRACE("Creating a hw vertex shader, prg=%d\n", ret
);
4229 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, ret
));
4231 TRACE("Created hw vertex shader, prg=%d\n", ret
);
4232 /* Create the program and check for errors */
4233 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
4234 buffer
->bsize
, buffer
->buffer
));
4235 checkGLcall("glProgramStringARB()");
4237 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &errPos
);
4240 FIXME("HW VertexShader Error at position %d: %s\n\n",
4241 errPos
, debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
4242 shader_arb_dump_program_source(buffer
->buffer
);
4249 GL_EXTCALL(glGetProgramivARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
4250 checkGLcall("glGetProgramivARB()");
4251 if (!native
) WARN("Program exceeds native resource limits.\n");
4253 /* Load immediate constants */
4256 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, local_constant
, entry
)
4258 const float *value
= (const float *)lconst
->value
;
4259 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, lconst_map
[lconst
->idx
], value
));
4263 HeapFree(GetProcessHeap(), 0, lconst_map
);
4268 /* GL locking is done by the caller */
4269 static struct arb_ps_compiled_shader
*find_arb_pshader(struct wined3d_shader
*shader
,
4270 const struct arb_ps_compile_args
*args
)
4272 struct wined3d_device
*device
= shader
->device
;
4273 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
4276 struct arb_ps_compiled_shader
*new_array
;
4277 struct wined3d_shader_buffer buffer
;
4278 struct arb_pshader_private
*shader_data
;
4281 if (!shader
->backend_data
)
4283 struct shader_arb_priv
*priv
= device
->shader_priv
;
4285 shader
->backend_data
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
4286 shader_data
= shader
->backend_data
;
4287 shader_data
->clamp_consts
= shader
->reg_maps
.shader_version
.major
== 1;
4289 if (shader
->reg_maps
.shader_version
.major
< 3)
4290 shader_data
->input_signature_idx
= ~0;
4292 shader_data
->input_signature_idx
= find_input_signature(priv
, shader
->input_signature
);
4294 shader_data
->has_signature_idx
= TRUE
;
4295 TRACE("Shader got assigned input signature index %u\n", shader_data
->input_signature_idx
);
4297 if (!device
->vs_clipping
)
4298 shader_data
->clipplane_emulation
= shader_find_free_input_register(&shader
->reg_maps
,
4299 gl_info
->limits
.texture_stages
- 1);
4301 shader_data
->clipplane_emulation
= ~0U;
4303 shader_data
= shader
->backend_data
;
4305 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4306 * so a linear search is more performant than a hashmap or a binary search
4307 * (cache coherency etc)
4309 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4311 if (!memcmp(&shader_data
->gl_shaders
[i
].args
, args
, sizeof(*args
)))
4312 return &shader_data
->gl_shaders
[i
];
4315 TRACE("No matching GL shader found, compiling a new shader\n");
4316 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4317 if (shader_data
->num_gl_shaders
)
4319 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4320 new_array
= HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, shader_data
->gl_shaders
,
4321 new_size
* sizeof(*shader_data
->gl_shaders
));
4323 new_array
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
->gl_shaders
));
4328 ERR("Out of memory\n");
4331 shader_data
->gl_shaders
= new_array
;
4332 shader_data
->shader_array_size
= new_size
;
4335 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4337 pixelshader_update_samplers(&shader
->reg_maps
, device
->stateBlock
->state
.textures
);
4339 if (!shader_buffer_init(&buffer
))
4341 ERR("Failed to initialize shader buffer.\n");
4345 ret
= shader_arb_generate_pshader(shader
, gl_info
, &buffer
, args
,
4346 &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
]);
4347 shader_buffer_free(&buffer
);
4348 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].prgId
= ret
;
4350 return &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++];
4353 static inline BOOL
vs_args_equal(const struct arb_vs_compile_args
*stored
, const struct arb_vs_compile_args
*new,
4354 const DWORD use_map
, BOOL skip_int
) {
4355 if((stored
->super
.swizzle_map
& use_map
) != new->super
.swizzle_map
) return FALSE
;
4356 if(stored
->super
.clip_enabled
!= new->super
.clip_enabled
) return FALSE
;
4357 if(stored
->super
.fog_src
!= new->super
.fog_src
) return FALSE
;
4358 if(stored
->clip
.boolclip_compare
!= new->clip
.boolclip_compare
) return FALSE
;
4359 if(stored
->ps_signature
!= new->ps_signature
) return FALSE
;
4360 if(stored
->vertex
.samplers_compare
!= new->vertex
.samplers_compare
) return FALSE
;
4361 if(skip_int
) return TRUE
;
4363 return !memcmp(stored
->loop_ctrl
, new->loop_ctrl
, sizeof(stored
->loop_ctrl
));
4366 static struct arb_vs_compiled_shader
*find_arb_vshader(struct wined3d_shader
*shader
,
4367 const struct arb_vs_compile_args
*args
)
4369 struct wined3d_device
*device
= shader
->device
;
4370 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
4371 DWORD use_map
= device
->strided_streams
.use_map
;
4374 struct arb_vs_compiled_shader
*new_array
;
4375 struct wined3d_shader_buffer buffer
;
4376 struct arb_vshader_private
*shader_data
;
4379 if (!shader
->backend_data
)
4381 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
4383 shader
->backend_data
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
4384 shader_data
= shader
->backend_data
;
4386 if ((gl_info
->quirks
& WINED3D_QUIRK_ARB_VS_OFFSET_LIMIT
)
4387 && reg_maps
->min_rel_offset
<= reg_maps
->max_rel_offset
)
4389 if (reg_maps
->max_rel_offset
- reg_maps
->min_rel_offset
> 127)
4391 FIXME("The difference between the minimum and maximum relative offset is > 127.\n");
4392 FIXME("Which this OpenGL implementation does not support. Try using GLSL.\n");
4393 FIXME("Min: %u, Max: %u.\n", reg_maps
->min_rel_offset
, reg_maps
->max_rel_offset
);
4395 else if (reg_maps
->max_rel_offset
- reg_maps
->min_rel_offset
> 63)
4396 shader_data
->rel_offset
= reg_maps
->min_rel_offset
+ 63;
4397 else if (reg_maps
->max_rel_offset
> 63)
4398 shader_data
->rel_offset
= reg_maps
->min_rel_offset
;
4401 shader_data
= shader
->backend_data
;
4403 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4404 * so a linear search is more performant than a hashmap or a binary search
4405 * (cache coherency etc)
4407 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
4408 if (vs_args_equal(&shader_data
->gl_shaders
[i
].args
, args
,
4409 use_map
, gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]))
4411 return &shader_data
->gl_shaders
[i
];
4415 TRACE("No matching GL shader found, compiling a new shader\n");
4417 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4418 if (shader_data
->num_gl_shaders
)
4420 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4421 new_array
= HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, shader_data
->gl_shaders
,
4422 new_size
* sizeof(*shader_data
->gl_shaders
));
4424 new_array
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
->gl_shaders
));
4429 ERR("Out of memory\n");
4432 shader_data
->gl_shaders
= new_array
;
4433 shader_data
->shader_array_size
= new_size
;
4436 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4438 if (!shader_buffer_init(&buffer
))
4440 ERR("Failed to initialize shader buffer.\n");
4444 ret
= shader_arb_generate_vshader(shader
, gl_info
, &buffer
, args
,
4445 &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
]);
4446 shader_buffer_free(&buffer
);
4447 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].prgId
= ret
;
4449 return &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++];
4452 static void find_arb_ps_compile_args(const struct wined3d_state
*state
,
4453 const struct wined3d_shader
*shader
, struct arb_ps_compile_args
*args
)
4455 struct wined3d_device
*device
= shader
->device
;
4456 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
4460 find_ps_compile_args(state
, shader
, &args
->super
);
4462 /* This forces all local boolean constants to 1 to make them stateblock independent */
4463 args
->bools
= shader
->reg_maps
.local_bool_consts
;
4465 for(i
= 0; i
< MAX_CONST_B
; i
++)
4467 if (state
->ps_consts_b
[i
])
4468 args
->bools
|= ( 1 << i
);
4471 /* Only enable the clip plane emulation KIL if at least one clipplane is enabled. The KIL instruction
4472 * is quite expensive because it forces the driver to disable early Z discards. It is cheaper to
4473 * duplicate the shader than have a no-op KIL instruction in every shader
4475 if (!device
->vs_clipping
&& use_vs(state
)
4476 && state
->render_states
[WINED3DRS_CLIPPING
]
4477 && state
->render_states
[WINED3DRS_CLIPPLANEENABLE
])
4482 /* Skip if unused or local, or supported natively */
4483 int_skip
= ~shader
->reg_maps
.integer_constants
| shader
->reg_maps
.local_int_consts
;
4484 if (int_skip
== 0xffff || gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
])
4486 memset(&args
->loop_ctrl
, 0, sizeof(args
->loop_ctrl
));
4490 for(i
= 0; i
< MAX_CONST_I
; i
++)
4492 if(int_skip
& (1 << i
))
4494 args
->loop_ctrl
[i
][0] = 0;
4495 args
->loop_ctrl
[i
][1] = 0;
4496 args
->loop_ctrl
[i
][2] = 0;
4500 args
->loop_ctrl
[i
][0] = state
->ps_consts_i
[i
* 4];
4501 args
->loop_ctrl
[i
][1] = state
->ps_consts_i
[i
* 4 + 1];
4502 args
->loop_ctrl
[i
][2] = state
->ps_consts_i
[i
* 4 + 2];
4507 static void find_arb_vs_compile_args(const struct wined3d_state
*state
,
4508 const struct wined3d_shader
*shader
, struct arb_vs_compile_args
*args
)
4510 struct wined3d_device
*device
= shader
->device
;
4511 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
4515 find_vs_compile_args(state
, shader
, &args
->super
);
4517 args
->clip
.boolclip_compare
= 0;
4520 const struct wined3d_shader
*ps
= state
->pixel_shader
;
4521 const struct arb_pshader_private
*shader_priv
= ps
->backend_data
;
4522 args
->ps_signature
= shader_priv
->input_signature_idx
;
4524 args
->clip
.boolclip
.clip_texcoord
= shader_priv
->clipplane_emulation
+ 1;
4528 args
->ps_signature
= ~0;
4529 if (!device
->vs_clipping
)
4531 args
->clip
.boolclip
.clip_texcoord
= ffp_clip_emul(state
) ? gl_info
->limits
.texture_stages
: 0;
4533 /* Otherwise: Setting boolclip_compare set clip_texcoord to 0 */
4536 if (args
->clip
.boolclip
.clip_texcoord
)
4538 if (state
->render_states
[WINED3DRS_CLIPPING
])
4539 args
->clip
.boolclip
.clipplane_mask
= (unsigned char)state
->render_states
[WINED3DRS_CLIPPLANEENABLE
];
4540 /* clipplane_mask was set to 0 by setting boolclip_compare to 0 */
4543 /* This forces all local boolean constants to 1 to make them stateblock independent */
4544 args
->clip
.boolclip
.bools
= shader
->reg_maps
.local_bool_consts
;
4545 /* TODO: Figure out if it would be better to store bool constants as bitmasks in the stateblock */
4546 for(i
= 0; i
< MAX_CONST_B
; i
++)
4548 if (state
->vs_consts_b
[i
])
4549 args
->clip
.boolclip
.bools
|= ( 1 << i
);
4552 args
->vertex
.samplers
[0] = device
->texUnitMap
[MAX_FRAGMENT_SAMPLERS
+ 0];
4553 args
->vertex
.samplers
[1] = device
->texUnitMap
[MAX_FRAGMENT_SAMPLERS
+ 1];
4554 args
->vertex
.samplers
[2] = device
->texUnitMap
[MAX_FRAGMENT_SAMPLERS
+ 2];
4555 args
->vertex
.samplers
[3] = 0;
4557 /* Skip if unused or local */
4558 int_skip
= ~shader
->reg_maps
.integer_constants
| shader
->reg_maps
.local_int_consts
;
4559 /* This is about flow control, not clipping. */
4560 if (int_skip
== 0xffff || gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
])
4562 memset(&args
->loop_ctrl
, 0, sizeof(args
->loop_ctrl
));
4566 for(i
= 0; i
< MAX_CONST_I
; i
++)
4568 if(int_skip
& (1 << i
))
4570 args
->loop_ctrl
[i
][0] = 0;
4571 args
->loop_ctrl
[i
][1] = 0;
4572 args
->loop_ctrl
[i
][2] = 0;
4576 args
->loop_ctrl
[i
][0] = state
->vs_consts_i
[i
* 4];
4577 args
->loop_ctrl
[i
][1] = state
->vs_consts_i
[i
* 4 + 1];
4578 args
->loop_ctrl
[i
][2] = state
->vs_consts_i
[i
* 4 + 2];
4583 /* GL locking is done by the caller */
4584 static void shader_arb_select(const struct wined3d_context
*context
, BOOL usePS
, BOOL useVS
)
4586 struct wined3d_device
*device
= context
->swapchain
->device
;
4587 struct shader_arb_priv
*priv
= device
->shader_priv
;
4588 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4589 const struct wined3d_state
*state
= &device
->stateBlock
->state
;
4592 /* Deal with pixel shaders first so the vertex shader arg function has the input signature ready */
4595 struct wined3d_shader
*ps
= state
->pixel_shader
;
4596 struct arb_ps_compile_args compile_args
;
4597 struct arb_ps_compiled_shader
*compiled
;
4599 TRACE("Using pixel shader %p.\n", ps
);
4600 find_arb_ps_compile_args(state
, ps
, &compile_args
);
4601 compiled
= find_arb_pshader(ps
, &compile_args
);
4602 priv
->current_fprogram_id
= compiled
->prgId
;
4603 priv
->compiled_fprog
= compiled
;
4605 /* Bind the fragment program */
4606 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, priv
->current_fprogram_id
));
4607 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->current_fprogram_id);");
4609 if(!priv
->use_arbfp_fixed_func
) {
4610 /* Enable OpenGL fragment programs */
4611 glEnable(GL_FRAGMENT_PROGRAM_ARB
);
4612 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
4614 TRACE("(%p) : Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n",
4615 device
, priv
->current_fprogram_id
);
4617 /* Pixel Shader 1.x constants are clamped to [-1;1], Pixel Shader 2.0 constants are not. If switching between
4618 * a 1.x and newer shader, reload the first 8 constants
4620 if (priv
->last_ps_const_clamped
!= ((struct arb_pshader_private
*)ps
->backend_data
)->clamp_consts
)
4622 priv
->last_ps_const_clamped
= ((struct arb_pshader_private
*)ps
->backend_data
)->clamp_consts
;
4623 device
->highest_dirty_ps_const
= max(device
->highest_dirty_ps_const
, 8);
4624 for(i
= 0; i
< 8; i
++)
4626 context
->pshader_const_dirty
[i
] = 1;
4628 /* Also takes care of loading local constants */
4629 shader_arb_load_constants(context
, TRUE
, FALSE
);
4633 UINT rt_height
= device
->fb
.render_targets
[0]->resource
.height
;
4634 shader_arb_ps_local_constants(compiled
, context
, state
, rt_height
);
4637 /* Force constant reloading for the NP2 fixup (see comment in shader_glsl_select for more info) */
4638 if (compiled
->np2fixup_info
.super
.active
)
4639 shader_arb_load_np2fixup_constants(priv
, gl_info
, state
);
4641 else if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
] && !priv
->use_arbfp_fixed_func
)
4643 /* Disable only if we're not using arbfp fixed function fragment processing. If this is used,
4644 * keep GL_FRAGMENT_PROGRAM_ARB enabled, and the fixed function pipeline will bind the fixed function
4645 * replacement shader
4647 glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4648 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4649 priv
->current_fprogram_id
= 0;
4654 struct wined3d_shader
*vs
= state
->vertex_shader
;
4655 struct arb_vs_compile_args compile_args
;
4656 struct arb_vs_compiled_shader
*compiled
;
4658 TRACE("Using vertex shader %p\n", vs
);
4659 find_arb_vs_compile_args(state
, vs
, &compile_args
);
4660 compiled
= find_arb_vshader(vs
, &compile_args
);
4661 priv
->current_vprogram_id
= compiled
->prgId
;
4662 priv
->compiled_vprog
= compiled
;
4664 /* Bind the vertex program */
4665 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->current_vprogram_id
));
4666 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->current_vprogram_id);");
4668 /* Enable OpenGL vertex programs */
4669 glEnable(GL_VERTEX_PROGRAM_ARB
);
4670 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
4671 TRACE("(%p) : Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", device
, priv
->current_vprogram_id
);
4672 shader_arb_vs_local_constants(compiled
, context
, state
);
4674 if(priv
->last_vs_color_unclamp
!= compiled
->need_color_unclamp
) {
4675 priv
->last_vs_color_unclamp
= compiled
->need_color_unclamp
;
4677 if (gl_info
->supported
[ARB_COLOR_BUFFER_FLOAT
])
4679 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, !compiled
->need_color_unclamp
));
4680 checkGLcall("glClampColorARB");
4682 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
4686 else if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
4688 priv
->current_vprogram_id
= 0;
4689 glDisable(GL_VERTEX_PROGRAM_ARB
);
4690 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4694 /* GL locking is done by the caller */
4695 static void shader_arb_select_depth_blt(void *shader_priv
, const struct wined3d_gl_info
*gl_info
,
4696 enum tex_types tex_type
, const SIZE
*ds_mask_size
)
4698 const float mask
[] = {0.0f
, 0.0f
, (float)ds_mask_size
->cx
, (float)ds_mask_size
->cy
};
4699 BOOL masked
= ds_mask_size
->cx
&& ds_mask_size
->cy
;
4700 struct shader_arb_priv
*priv
= shader_priv
;
4701 GLuint
*blt_fprogram
;
4703 if (!priv
->depth_blt_vprogram_id
) priv
->depth_blt_vprogram_id
= create_arb_blt_vertex_program(gl_info
);
4704 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->depth_blt_vprogram_id
));
4705 glEnable(GL_VERTEX_PROGRAM_ARB
);
4707 blt_fprogram
= masked
? &priv
->depth_blt_fprogram_id_masked
[tex_type
] : &priv
->depth_blt_fprogram_id_full
[tex_type
];
4708 if (!*blt_fprogram
) *blt_fprogram
= create_arb_blt_fragment_program(gl_info
, tex_type
, masked
);
4709 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, *blt_fprogram
));
4710 if (masked
) GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, 0, mask
));
4711 glEnable(GL_FRAGMENT_PROGRAM_ARB
);
4714 /* GL locking is done by the caller */
4715 static void shader_arb_deselect_depth_blt(void *shader_priv
, const struct wined3d_gl_info
*gl_info
)
4717 struct shader_arb_priv
*priv
= shader_priv
;
4719 if (priv
->current_vprogram_id
) {
4720 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->current_vprogram_id
));
4721 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
4723 TRACE("Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB.\n", priv
->current_vprogram_id
);
4727 glDisable(GL_VERTEX_PROGRAM_ARB
);
4728 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4731 if (priv
->current_fprogram_id
) {
4732 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, priv
->current_fprogram_id
));
4733 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
4735 TRACE("Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB.\n", priv
->current_fprogram_id
);
4737 else if(!priv
->use_arbfp_fixed_func
)
4739 glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4740 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4744 static void shader_arb_destroy(struct wined3d_shader
*shader
)
4746 struct wined3d_device
*device
= shader
->device
;
4747 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
4749 if (shader_is_pshader_version(shader
->reg_maps
.shader_version
.type
))
4751 struct arb_pshader_private
*shader_data
= shader
->backend_data
;
4754 if(!shader_data
) return; /* This can happen if a shader was never compiled */
4756 if (shader_data
->num_gl_shaders
)
4758 struct wined3d_context
*context
= context_acquire(device
, NULL
);
4761 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4763 GL_EXTCALL(glDeleteProgramsARB(1, &shader_data
->gl_shaders
[i
].prgId
));
4764 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader_data->gl_shaders[i].prgId))");
4768 context_release(context
);
4771 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
);
4772 HeapFree(GetProcessHeap(), 0, shader_data
);
4773 shader
->backend_data
= NULL
;
4777 struct arb_vshader_private
*shader_data
= shader
->backend_data
;
4780 if(!shader_data
) return; /* This can happen if a shader was never compiled */
4782 if (shader_data
->num_gl_shaders
)
4784 struct wined3d_context
*context
= context_acquire(device
, NULL
);
4787 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4789 GL_EXTCALL(glDeleteProgramsARB(1, &shader_data
->gl_shaders
[i
].prgId
));
4790 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader_data->gl_shaders[i].prgId))");
4794 context_release(context
);
4797 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
);
4798 HeapFree(GetProcessHeap(), 0, shader_data
);
4799 shader
->backend_data
= NULL
;
4803 static int sig_tree_compare(const void *key
, const struct wine_rb_entry
*entry
)
4805 struct ps_signature
*e
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
4806 return compare_sig(key
, e
->sig
);
4809 static const struct wine_rb_functions sig_tree_functions
=
4817 static HRESULT
shader_arb_alloc(struct wined3d_device
*device
)
4819 struct shader_arb_priv
*priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*priv
));
4820 if(wine_rb_init(&priv
->signature_tree
, &sig_tree_functions
) == -1)
4822 ERR("RB tree init failed\n");
4823 HeapFree(GetProcessHeap(), 0, priv
);
4824 return E_OUTOFMEMORY
;
4826 device
->shader_priv
= priv
;
4830 static void release_signature(struct wine_rb_entry
*entry
, void *context
)
4832 struct ps_signature
*sig
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
4834 for(i
= 0; i
< MAX_REG_INPUT
; i
++)
4836 HeapFree(GetProcessHeap(), 0, (char *) sig
->sig
[i
].semantic_name
);
4838 HeapFree(GetProcessHeap(), 0, sig
->sig
);
4839 HeapFree(GetProcessHeap(), 0, sig
);
4842 /* Context activation is done by the caller. */
4843 static void shader_arb_free(struct wined3d_device
*device
)
4845 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
4846 struct shader_arb_priv
*priv
= device
->shader_priv
;
4850 if(priv
->depth_blt_vprogram_id
) {
4851 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_vprogram_id
));
4853 for (i
= 0; i
< tex_type_count
; ++i
)
4855 if (priv
->depth_blt_fprogram_id_full
[i
])
4857 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_fprogram_id_full
[i
]));
4859 if (priv
->depth_blt_fprogram_id_masked
[i
])
4861 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_fprogram_id_masked
[i
]));
4866 wine_rb_destroy(&priv
->signature_tree
, release_signature
, NULL
);
4867 HeapFree(GetProcessHeap(), 0, device
->shader_priv
);
4870 static BOOL
shader_arb_dirty_const(void)
4875 static void shader_arb_get_caps(const struct wined3d_gl_info
*gl_info
, struct shader_caps
*caps
)
4877 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
4881 /* 96 is the minimum allowed value of MAX_PROGRAM_ENV_PARAMETERS_ARB
4882 * for vertex programs. If the native limit is less than that it's
4883 * not very useful, and e.g. Mesa swrast returns 0, probably to
4884 * indicate it's a software implementation. */
4885 if (gl_info
->limits
.arb_vs_native_constants
< 96)
4886 vs_consts
= gl_info
->limits
.arb_vs_float_constants
;
4888 vs_consts
= min(gl_info
->limits
.arb_vs_float_constants
, gl_info
->limits
.arb_vs_native_constants
);
4890 if (gl_info
->supported
[NV_VERTEX_PROGRAM3
])
4892 caps
->VertexShaderVersion
= WINED3DVS_VERSION(3,0);
4893 TRACE_(d3d_caps
)("Hardware vertex shader version 3.0 enabled (NV_VERTEX_PROGRAM3)\n");
4895 else if (vs_consts
>= 256)
4897 /* Shader Model 2.0 requires at least 256 vertex shader constants */
4898 caps
->VertexShaderVersion
= WINED3DVS_VERSION(2,0);
4899 TRACE_(d3d_caps
)("Hardware vertex shader version 2.0 enabled (ARB_PROGRAM)\n");
4903 caps
->VertexShaderVersion
= WINED3DVS_VERSION(1,1);
4904 TRACE_(d3d_caps
)("Hardware vertex shader version 1.1 enabled (ARB_PROGRAM)\n");
4906 caps
->MaxVertexShaderConst
= vs_consts
;
4910 caps
->VertexShaderVersion
= 0;
4911 caps
->MaxVertexShaderConst
= 0;
4914 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
4918 /* Similar as above for vertex programs, but the minimum for fragment
4919 * programs is 24. */
4920 if (gl_info
->limits
.arb_ps_native_constants
< 24)
4921 ps_consts
= gl_info
->limits
.arb_ps_float_constants
;
4923 ps_consts
= min(gl_info
->limits
.arb_ps_float_constants
, gl_info
->limits
.arb_ps_native_constants
);
4925 if (gl_info
->supported
[NV_FRAGMENT_PROGRAM2
])
4927 caps
->PixelShaderVersion
= WINED3DPS_VERSION(3,0);
4928 TRACE_(d3d_caps
)("Hardware pixel shader version 3.0 enabled (NV_FRAGMENT_PROGRAM2)\n");
4930 else if (ps_consts
>= 32)
4932 /* Shader Model 2.0 requires at least 32 pixel shader constants */
4933 caps
->PixelShaderVersion
= WINED3DPS_VERSION(2,0);
4934 TRACE_(d3d_caps
)("Hardware pixel shader version 2.0 enabled (ARB_PROGRAM)\n");
4938 caps
->PixelShaderVersion
= WINED3DPS_VERSION(1,4);
4939 TRACE_(d3d_caps
)("Hardware pixel shader version 1.4 enabled (ARB_PROGRAM)\n");
4941 caps
->PixelShader1xMaxValue
= 8.0f
;
4942 caps
->MaxPixelShaderConst
= ps_consts
;
4946 caps
->PixelShaderVersion
= 0;
4947 caps
->PixelShader1xMaxValue
= 0.0f
;
4948 caps
->MaxPixelShaderConst
= 0;
4951 caps
->VSClipping
= use_nv_clip(gl_info
);
4954 static BOOL
shader_arb_color_fixup_supported(struct color_fixup_desc fixup
)
4956 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
4958 TRACE("Checking support for color_fixup:\n");
4959 dump_color_fixup_desc(fixup
);
4962 /* We support everything except complex conversions. */
4963 if (!is_complex_fixup(fixup
))
4969 TRACE("[FAILED]\n");
4973 static void shader_arb_add_instruction_modifiers(const struct wined3d_shader_instruction
*ins
) {
4975 char write_mask
[20], regstr
[50];
4976 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
4977 BOOL is_color
= FALSE
;
4978 const struct wined3d_shader_dst_param
*dst
;
4980 if (!ins
->dst_count
) return;
4984 if (!shift
) return; /* Saturate alone is handled by the instructions */
4986 shader_arb_get_write_mask(ins
, dst
, write_mask
);
4987 shader_arb_get_register_name(ins
, &dst
->reg
, regstr
, &is_color
);
4989 /* Generate a line that does the output modifier computation
4990 * FIXME: _SAT vs shift? _SAT alone is already handled in the instructions, if this
4991 * maps problems in e.g. _d4_sat modify shader_arb_get_modifier
4993 shader_addline(buffer
, "MUL%s %s%s, %s, %s;\n", shader_arb_get_modifier(ins
),
4994 regstr
, write_mask
, regstr
, shift_tab
[shift
]);
4997 static const SHADER_HANDLER shader_arb_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
4999 /* WINED3DSIH_ABS */ shader_hw_map2gl
,
5000 /* WINED3DSIH_ADD */ shader_hw_map2gl
,
5001 /* WINED3DSIH_AND */ NULL
,
5002 /* WINED3DSIH_BEM */ pshader_hw_bem
,
5003 /* WINED3DSIH_BREAK */ shader_hw_break
,
5004 /* WINED3DSIH_BREAKC */ shader_hw_breakc
,
5005 /* WINED3DSIH_BREAKP */ NULL
,
5006 /* WINED3DSIH_CALL */ shader_hw_call
,
5007 /* WINED3DSIH_CALLNZ */ NULL
,
5008 /* WINED3DSIH_CMP */ pshader_hw_cmp
,
5009 /* WINED3DSIH_CND */ pshader_hw_cnd
,
5010 /* WINED3DSIH_CRS */ shader_hw_map2gl
,
5011 /* WINED3DSIH_CUT */ NULL
,
5012 /* WINED3DSIH_DCL */ NULL
,
5013 /* WINED3DSIH_DEF */ NULL
,
5014 /* WINED3DSIH_DEFB */ NULL
,
5015 /* WINED3DSIH_DEFI */ NULL
,
5016 /* WINED3DSIH_DIV */ NULL
,
5017 /* WINED3DSIH_DP2ADD */ pshader_hw_dp2add
,
5018 /* WINED3DSIH_DP3 */ shader_hw_map2gl
,
5019 /* WINED3DSIH_DP4 */ shader_hw_map2gl
,
5020 /* WINED3DSIH_DST */ shader_hw_map2gl
,
5021 /* WINED3DSIH_DSX */ shader_hw_map2gl
,
5022 /* WINED3DSIH_DSY */ shader_hw_dsy
,
5023 /* WINED3DSIH_ELSE */ shader_hw_else
,
5024 /* WINED3DSIH_EMIT */ NULL
,
5025 /* WINED3DSIH_ENDIF */ shader_hw_endif
,
5026 /* WINED3DSIH_ENDLOOP */ shader_hw_endloop
,
5027 /* WINED3DSIH_ENDREP */ shader_hw_endrep
,
5028 /* WINED3DSIH_EXP */ shader_hw_scalar_op
,
5029 /* WINED3DSIH_EXPP */ shader_hw_scalar_op
,
5030 /* WINED3DSIH_FRC */ shader_hw_map2gl
,
5031 /* WINED3DSIH_FTOI */ NULL
,
5032 /* WINED3DSIH_IADD */ NULL
,
5033 /* WINED3DSIH_IEQ */ NULL
,
5034 /* WINED3DSIH_IF */ NULL
/* Hardcoded into the shader */,
5035 /* WINED3DSIH_IFC */ shader_hw_ifc
,
5036 /* WINED3DSIH_IGE */ NULL
,
5037 /* WINED3DSIH_IMUL */ NULL
,
5038 /* WINED3DSIH_ITOF */ NULL
,
5039 /* WINED3DSIH_LABEL */ shader_hw_label
,
5040 /* WINED3DSIH_LD */ NULL
,
5041 /* WINED3DSIH_LIT */ shader_hw_map2gl
,
5042 /* WINED3DSIH_LOG */ shader_hw_log
,
5043 /* WINED3DSIH_LOGP */ shader_hw_log
,
5044 /* WINED3DSIH_LOOP */ shader_hw_loop
,
5045 /* WINED3DSIH_LRP */ shader_hw_lrp
,
5046 /* WINED3DSIH_LT */ NULL
,
5047 /* WINED3DSIH_M3x2 */ shader_hw_mnxn
,
5048 /* WINED3DSIH_M3x3 */ shader_hw_mnxn
,
5049 /* WINED3DSIH_M3x4 */ shader_hw_mnxn
,
5050 /* WINED3DSIH_M4x3 */ shader_hw_mnxn
,
5051 /* WINED3DSIH_M4x4 */ shader_hw_mnxn
,
5052 /* WINED3DSIH_MAD */ shader_hw_map2gl
,
5053 /* WINED3DSIH_MAX */ shader_hw_map2gl
,
5054 /* WINED3DSIH_MIN */ shader_hw_map2gl
,
5055 /* WINED3DSIH_MOV */ shader_hw_mov
,
5056 /* WINED3DSIH_MOVA */ shader_hw_mov
,
5057 /* WINED3DSIH_MOVC */ NULL
,
5058 /* WINED3DSIH_MUL */ shader_hw_map2gl
,
5059 /* WINED3DSIH_NOP */ shader_hw_nop
,
5060 /* WINED3DSIH_NRM */ shader_hw_nrm
,
5061 /* WINED3DSIH_PHASE */ NULL
,
5062 /* WINED3DSIH_POW */ shader_hw_pow
,
5063 /* WINED3DSIH_RCP */ shader_hw_rcp
,
5064 /* WINED3DSIH_REP */ shader_hw_rep
,
5065 /* WINED3DSIH_RET */ shader_hw_ret
,
5066 /* WINED3DSIH_RSQ */ shader_hw_scalar_op
,
5067 /* WINED3DSIH_SAMPLE */ NULL
,
5068 /* WINED3DSIH_SAMPLE_GRAD */ NULL
,
5069 /* WINED3DSIH_SAMPLE_LOD */ NULL
,
5070 /* WINED3DSIH_SETP */ NULL
,
5071 /* WINED3DSIH_SGE */ shader_hw_map2gl
,
5072 /* WINED3DSIH_SGN */ shader_hw_sgn
,
5073 /* WINED3DSIH_SINCOS */ shader_hw_sincos
,
5074 /* WINED3DSIH_SLT */ shader_hw_map2gl
,
5075 /* WINED3DSIH_SQRT */ NULL
,
5076 /* WINED3DSIH_SUB */ shader_hw_map2gl
,
5077 /* WINED3DSIH_TEX */ pshader_hw_tex
,
5078 /* WINED3DSIH_TEXBEM */ pshader_hw_texbem
,
5079 /* WINED3DSIH_TEXBEML */ pshader_hw_texbem
,
5080 /* WINED3DSIH_TEXCOORD */ pshader_hw_texcoord
,
5081 /* WINED3DSIH_TEXDEPTH */ pshader_hw_texdepth
,
5082 /* WINED3DSIH_TEXDP3 */ pshader_hw_texdp3
,
5083 /* WINED3DSIH_TEXDP3TEX */ pshader_hw_texdp3tex
,
5084 /* WINED3DSIH_TEXKILL */ pshader_hw_texkill
,
5085 /* WINED3DSIH_TEXLDD */ shader_hw_texldd
,
5086 /* WINED3DSIH_TEXLDL */ shader_hw_texldl
,
5087 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_hw_texm3x2depth
,
5088 /* WINED3DSIH_TEXM3x2PAD */ pshader_hw_texm3x2pad
,
5089 /* WINED3DSIH_TEXM3x2TEX */ pshader_hw_texm3x2tex
,
5090 /* WINED3DSIH_TEXM3x3 */ pshader_hw_texm3x3
,
5091 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
5092 /* WINED3DSIH_TEXM3x3PAD */ pshader_hw_texm3x3pad
,
5093 /* WINED3DSIH_TEXM3x3SPEC */ pshader_hw_texm3x3spec
,
5094 /* WINED3DSIH_TEXM3x3TEX */ pshader_hw_texm3x3tex
,
5095 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_hw_texm3x3vspec
,
5096 /* WINED3DSIH_TEXREG2AR */ pshader_hw_texreg2ar
,
5097 /* WINED3DSIH_TEXREG2GB */ pshader_hw_texreg2gb
,
5098 /* WINED3DSIH_TEXREG2RGB */ pshader_hw_texreg2rgb
,
5099 /* WINED3DSIH_UTOF */ NULL
,
5102 static BOOL
get_bool_const(const struct wined3d_shader_instruction
*ins
,
5103 const struct wined3d_shader
*shader
, DWORD idx
)
5105 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
5106 BOOL vshader
= shader_is_vshader_version(reg_maps
->shader_version
.type
);
5108 WORD flag
= (1 << idx
);
5109 const local_constant
*constant
;
5110 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5112 if (reg_maps
->local_bool_consts
& flag
)
5114 /* What good is a if(bool) with a hardcoded local constant? I don't know, but handle it */
5115 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsB
, local_constant
, entry
)
5117 if (constant
->idx
== idx
)
5119 return constant
->value
[0];
5122 ERR("Local constant not found\n");
5127 if(vshader
) bools
= priv
->cur_vs_args
->clip
.boolclip
.bools
;
5128 else bools
= priv
->cur_ps_args
->bools
;
5129 return bools
& flag
;
5133 static void get_loop_control_const(const struct wined3d_shader_instruction
*ins
,
5134 const struct wined3d_shader
*shader
, UINT idx
, struct wined3d_shader_loop_control
*loop_control
)
5136 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
5137 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5139 /* Integer constants can either be a local constant, or they can be stored in the shader
5140 * type specific compile args. */
5141 if (reg_maps
->local_int_consts
& (1 << idx
))
5143 const local_constant
*constant
;
5145 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsI
, local_constant
, entry
)
5147 if (constant
->idx
== idx
)
5149 loop_control
->count
= constant
->value
[0];
5150 loop_control
->start
= constant
->value
[1];
5151 /* Step is signed. */
5152 loop_control
->step
= (int)constant
->value
[2];
5156 /* If this happens the flag was set incorrectly */
5157 ERR("Local constant not found\n");
5158 loop_control
->count
= 0;
5159 loop_control
->start
= 0;
5160 loop_control
->step
= 0;
5164 switch (reg_maps
->shader_version
.type
)
5166 case WINED3D_SHADER_TYPE_VERTEX
:
5167 /* Count and aL start value are unsigned */
5168 loop_control
->count
= priv
->cur_vs_args
->loop_ctrl
[idx
][0];
5169 loop_control
->start
= priv
->cur_vs_args
->loop_ctrl
[idx
][1];
5170 /* Step is signed. */
5171 loop_control
->step
= ((char)priv
->cur_vs_args
->loop_ctrl
[idx
][2]);
5174 case WINED3D_SHADER_TYPE_PIXEL
:
5175 loop_control
->count
= priv
->cur_ps_args
->loop_ctrl
[idx
][0];
5176 loop_control
->start
= priv
->cur_ps_args
->loop_ctrl
[idx
][1];
5177 loop_control
->step
= ((char)priv
->cur_ps_args
->loop_ctrl
[idx
][2]);
5181 FIXME("Unhandled shader type %#x.\n", reg_maps
->shader_version
.type
);
5186 static void record_instruction(struct list
*list
, const struct wined3d_shader_instruction
*ins
)
5189 struct wined3d_shader_dst_param
*dst_param
= NULL
;
5190 struct wined3d_shader_src_param
*src_param
= NULL
, *rel_addr
= NULL
;
5191 struct recorded_instruction
*rec
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*rec
));
5194 ERR("Out of memory\n");
5199 dst_param
= HeapAlloc(GetProcessHeap(), 0, sizeof(*dst_param
));
5200 if(!dst_param
) goto free
;
5201 *dst_param
= *ins
->dst
;
5202 if(ins
->dst
->reg
.rel_addr
)
5204 rel_addr
= HeapAlloc(GetProcessHeap(), 0, sizeof(*dst_param
->reg
.rel_addr
));
5205 if(!rel_addr
) goto free
;
5206 *rel_addr
= *ins
->dst
->reg
.rel_addr
;
5207 dst_param
->reg
.rel_addr
= rel_addr
;
5209 rec
->ins
.dst
= dst_param
;
5211 src_param
= HeapAlloc(GetProcessHeap(), 0, sizeof(*src_param
) * ins
->src_count
);
5212 if(!src_param
) goto free
;
5213 for(i
= 0; i
< ins
->src_count
; i
++)
5215 src_param
[i
] = ins
->src
[i
];
5216 if(ins
->src
[i
].reg
.rel_addr
)
5218 rel_addr
= HeapAlloc(GetProcessHeap(), 0, sizeof(*rel_addr
));
5219 if(!rel_addr
) goto free
;
5220 *rel_addr
= *ins
->src
[i
].reg
.rel_addr
;
5221 src_param
[i
].reg
.rel_addr
= rel_addr
;
5224 rec
->ins
.src
= src_param
;
5225 list_add_tail(list
, &rec
->entry
);
5229 ERR("Out of memory\n");
5232 HeapFree(GetProcessHeap(), 0, (void *) dst_param
->reg
.rel_addr
);
5233 HeapFree(GetProcessHeap(), 0, dst_param
);
5237 for(i
= 0; i
< ins
->src_count
; i
++)
5239 HeapFree(GetProcessHeap(), 0, (void *) src_param
[i
].reg
.rel_addr
);
5241 HeapFree(GetProcessHeap(), 0, src_param
);
5243 HeapFree(GetProcessHeap(), 0, rec
);
5246 static void free_recorded_instruction(struct list
*list
)
5248 struct recorded_instruction
*rec_ins
, *entry2
;
5251 LIST_FOR_EACH_ENTRY_SAFE(rec_ins
, entry2
, list
, struct recorded_instruction
, entry
)
5253 list_remove(&rec_ins
->entry
);
5254 if(rec_ins
->ins
.dst
)
5256 HeapFree(GetProcessHeap(), 0, (void *) rec_ins
->ins
.dst
->reg
.rel_addr
);
5257 HeapFree(GetProcessHeap(), 0, (void *) rec_ins
->ins
.dst
);
5259 if(rec_ins
->ins
.src
)
5261 for(i
= 0; i
< rec_ins
->ins
.src_count
; i
++)
5263 HeapFree(GetProcessHeap(), 0, (void *) rec_ins
->ins
.src
[i
].reg
.rel_addr
);
5265 HeapFree(GetProcessHeap(), 0, (void *) rec_ins
->ins
.src
);
5267 HeapFree(GetProcessHeap(), 0, rec_ins
);
5271 static void shader_arb_handle_instruction(const struct wined3d_shader_instruction
*ins
) {
5272 SHADER_HANDLER hw_fct
;
5273 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5274 struct wined3d_shader
*shader
= ins
->ctx
->shader
;
5275 struct control_frame
*control_frame
;
5276 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
5279 if(ins
->handler_idx
== WINED3DSIH_LOOP
|| ins
->handler_idx
== WINED3DSIH_REP
)
5281 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
5282 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5284 if(ins
->handler_idx
== WINED3DSIH_LOOP
) control_frame
->type
= LOOP
;
5285 if(ins
->handler_idx
== WINED3DSIH_REP
) control_frame
->type
= REP
;
5287 if(priv
->target_version
>= NV2
)
5289 control_frame
->no
.loop
= priv
->num_loops
++;
5294 /* Don't bother recording when we're in a not used if branch */
5300 if(!priv
->recording
)
5302 list_init(&priv
->record
);
5303 priv
->recording
= TRUE
;
5304 control_frame
->outer_loop
= TRUE
;
5305 get_loop_control_const(ins
, shader
, ins
->src
[0].reg
.idx
, &control_frame
->loop_control
);
5306 return; /* Instruction is handled */
5308 /* Record this loop in the outer loop's recording */
5311 else if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
|| ins
->handler_idx
== WINED3DSIH_ENDREP
)
5313 if(priv
->target_version
>= NV2
)
5315 /* Nothing to do. The control frame is popped after the HW instr handler */
5319 struct list
*e
= list_head(&priv
->control_frames
);
5320 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5321 list_remove(&control_frame
->entry
);
5323 if(control_frame
->outer_loop
)
5325 unsigned int iteration
;
5329 /* Turn off recording before playback */
5330 priv
->recording
= FALSE
;
5332 /* Move the recorded instructions to a separate list and get them out of the private data
5333 * structure. If there are nested loops, the shader_arb_handle_instruction below will
5334 * be recorded again, thus priv->record might be overwritten
5337 list_move_tail(©
, &priv
->record
);
5338 list_init(&priv
->record
);
5340 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5342 shader_addline(buffer
, "#unrolling loop: %u iterations, aL=%u, inc %d\n",
5343 control_frame
->loop_control
.count
, control_frame
->loop_control
.start
,
5344 control_frame
->loop_control
.step
);
5345 aL
= control_frame
->loop_control
.start
;
5349 shader_addline(buffer
, "#unrolling rep: %u iterations\n", control_frame
->loop_control
.count
);
5352 for (iteration
= 0; iteration
< control_frame
->loop_control
.count
; ++iteration
)
5354 struct recorded_instruction
*rec_ins
;
5355 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5358 shader_addline(buffer
, "#Iteration %u, aL=%d\n", iteration
, aL
);
5362 shader_addline(buffer
, "#Iteration %u\n", iteration
);
5365 LIST_FOR_EACH_ENTRY(rec_ins
, ©
, struct recorded_instruction
, entry
)
5367 shader_arb_handle_instruction(&rec_ins
->ins
);
5370 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5372 aL
+= control_frame
->loop_control
.step
;
5375 shader_addline(buffer
, "#end loop/rep\n");
5377 free_recorded_instruction(©
);
5378 HeapFree(GetProcessHeap(), 0, control_frame
);
5379 return; /* Instruction is handled */
5383 /* This is a nested loop. Proceed to the normal recording function */
5384 HeapFree(GetProcessHeap(), 0, control_frame
);
5391 record_instruction(&priv
->record
, ins
);
5396 if(ins
->handler_idx
== WINED3DSIH_IF
)
5398 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
5399 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5400 control_frame
->type
= IF
;
5402 bool_const
= get_bool_const(ins
, shader
, ins
->src
[0].reg
.idx
);
5403 if(ins
->src
[0].modifiers
== WINED3DSPSM_NOT
) bool_const
= !bool_const
;
5404 if (!priv
->muted
&& !bool_const
)
5406 shader_addline(buffer
, "#if(FALSE){\n");
5408 control_frame
->muting
= TRUE
;
5410 else shader_addline(buffer
, "#if(TRUE) {\n");
5412 return; /* Instruction is handled */
5414 else if(ins
->handler_idx
== WINED3DSIH_IFC
)
5416 /* IF(bool) and if_cond(a, b) use the same ELSE and ENDIF tokens */
5417 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
5418 control_frame
->type
= IFC
;
5419 control_frame
->no
.ifc
= priv
->num_ifcs
++;
5420 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5422 else if(ins
->handler_idx
== WINED3DSIH_ELSE
)
5424 struct list
*e
= list_head(&priv
->control_frames
);
5425 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5427 if(control_frame
->type
== IF
)
5429 shader_addline(buffer
, "#} else {\n");
5430 if(!priv
->muted
&& !control_frame
->muting
)
5433 control_frame
->muting
= TRUE
;
5435 else if(control_frame
->muting
) priv
->muted
= FALSE
;
5436 return; /* Instruction is handled. */
5438 /* In case of an ifc, generate a HW shader instruction */
5440 else if(ins
->handler_idx
== WINED3DSIH_ENDIF
)
5442 struct list
*e
= list_head(&priv
->control_frames
);
5443 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5445 if(control_frame
->type
== IF
)
5447 shader_addline(buffer
, "#} endif\n");
5448 if(control_frame
->muting
) priv
->muted
= FALSE
;
5449 list_remove(&control_frame
->entry
);
5450 HeapFree(GetProcessHeap(), 0, control_frame
);
5451 return; /* Instruction is handled */
5455 if(priv
->muted
) return;
5457 /* Select handler */
5458 hw_fct
= shader_arb_instruction_handler_table
[ins
->handler_idx
];
5460 /* Unhandled opcode */
5463 FIXME("Backend can't handle opcode %#x\n", ins
->handler_idx
);
5468 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
|| ins
->handler_idx
== WINED3DSIH_ENDREP
)
5470 struct list
*e
= list_head(&priv
->control_frames
);
5471 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5472 list_remove(&control_frame
->entry
);
5473 HeapFree(GetProcessHeap(), 0, control_frame
);
5476 else if(ins
->handler_idx
== WINED3DSIH_ENDIF
)
5478 /* Non-ifc ENDIFs don't reach that place because of the return in the if block above */
5479 struct list
*e
= list_head(&priv
->control_frames
);
5480 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5481 list_remove(&control_frame
->entry
);
5482 HeapFree(GetProcessHeap(), 0, control_frame
);
5486 shader_arb_add_instruction_modifiers(ins
);
5489 const struct wined3d_shader_backend_ops arb_program_shader_backend
=
5491 shader_arb_handle_instruction
,
5493 shader_arb_select_depth_blt
,
5494 shader_arb_deselect_depth_blt
,
5495 shader_arb_update_float_vertex_constants
,
5496 shader_arb_update_float_pixel_constants
,
5497 shader_arb_load_constants
,
5498 shader_arb_load_np2fixup_constants
,
5502 shader_arb_dirty_const
,
5503 shader_arb_get_caps
,
5504 shader_arb_color_fixup_supported
,
5507 /* ARB_fragment_program fixed function pipeline replacement definitions */
5508 #define ARB_FFP_CONST_TFACTOR 0
5509 #define ARB_FFP_CONST_SPECULAR_ENABLE ((ARB_FFP_CONST_TFACTOR) + 1)
5510 #define ARB_FFP_CONST_CONSTANT(i) ((ARB_FFP_CONST_SPECULAR_ENABLE) + 1 + i)
5511 #define ARB_FFP_CONST_BUMPMAT(i) ((ARB_FFP_CONST_CONSTANT(7)) + 1 + i)
5512 #define ARB_FFP_CONST_LUMINANCE(i) ((ARB_FFP_CONST_BUMPMAT(7)) + 1 + i)
5514 struct arbfp_ffp_desc
5516 struct ffp_frag_desc parent
;
5518 unsigned int num_textures_used
;
5521 /* Context activation is done by the caller. */
5522 static void arbfp_enable(BOOL enable
)
5526 glEnable(GL_FRAGMENT_PROGRAM_ARB
);
5527 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
5529 glDisable(GL_FRAGMENT_PROGRAM_ARB
);
5530 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
5535 static HRESULT
arbfp_alloc(struct wined3d_device
*device
)
5537 struct shader_arb_priv
*priv
;
5538 /* Share private data between the shader backend and the pipeline replacement, if both
5539 * are the arb implementation. This is needed to figure out whether ARBfp should be disabled
5540 * if no pixel shader is bound or not
5542 if (device
->shader_backend
== &arb_program_shader_backend
)
5544 device
->fragment_priv
= device
->shader_priv
;
5548 device
->fragment_priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct shader_arb_priv
));
5549 if (!device
->fragment_priv
) return E_OUTOFMEMORY
;
5551 priv
= device
->fragment_priv
;
5552 if (wine_rb_init(&priv
->fragment_shaders
, &wined3d_ffp_frag_program_rb_functions
) == -1)
5554 ERR("Failed to initialize rbtree.\n");
5555 HeapFree(GetProcessHeap(), 0, device
->fragment_priv
);
5556 return E_OUTOFMEMORY
;
5558 priv
->use_arbfp_fixed_func
= TRUE
;
5562 /* Context activation is done by the caller. */
5563 static void arbfp_free_ffpshader(struct wine_rb_entry
*entry
, void *context
)
5565 const struct wined3d_gl_info
*gl_info
= context
;
5566 struct arbfp_ffp_desc
*entry_arb
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_ffp_desc
, parent
.entry
);
5569 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb
->shader
));
5570 checkGLcall("glDeleteProgramsARB(1, &entry_arb->shader)");
5571 HeapFree(GetProcessHeap(), 0, entry_arb
);
5575 /* Context activation is done by the caller. */
5576 static void arbfp_free(struct wined3d_device
*device
)
5578 struct shader_arb_priv
*priv
= device
->fragment_priv
;
5580 wine_rb_destroy(&priv
->fragment_shaders
, arbfp_free_ffpshader
, &device
->adapter
->gl_info
);
5581 priv
->use_arbfp_fixed_func
= FALSE
;
5583 if (device
->shader_backend
!= &arb_program_shader_backend
)
5585 HeapFree(GetProcessHeap(), 0, device
->fragment_priv
);
5589 static void arbfp_get_caps(const struct wined3d_gl_info
*gl_info
, struct fragment_caps
*caps
)
5591 caps
->PrimitiveMiscCaps
= WINED3DPMISCCAPS_TSSARGTEMP
;
5592 caps
->TextureOpCaps
= WINED3DTEXOPCAPS_DISABLE
|
5593 WINED3DTEXOPCAPS_SELECTARG1
|
5594 WINED3DTEXOPCAPS_SELECTARG2
|
5595 WINED3DTEXOPCAPS_MODULATE4X
|
5596 WINED3DTEXOPCAPS_MODULATE2X
|
5597 WINED3DTEXOPCAPS_MODULATE
|
5598 WINED3DTEXOPCAPS_ADDSIGNED2X
|
5599 WINED3DTEXOPCAPS_ADDSIGNED
|
5600 WINED3DTEXOPCAPS_ADD
|
5601 WINED3DTEXOPCAPS_SUBTRACT
|
5602 WINED3DTEXOPCAPS_ADDSMOOTH
|
5603 WINED3DTEXOPCAPS_BLENDCURRENTALPHA
|
5604 WINED3DTEXOPCAPS_BLENDFACTORALPHA
|
5605 WINED3DTEXOPCAPS_BLENDTEXTUREALPHA
|
5606 WINED3DTEXOPCAPS_BLENDDIFFUSEALPHA
|
5607 WINED3DTEXOPCAPS_BLENDTEXTUREALPHAPM
|
5608 WINED3DTEXOPCAPS_MODULATEALPHA_ADDCOLOR
|
5609 WINED3DTEXOPCAPS_MODULATECOLOR_ADDALPHA
|
5610 WINED3DTEXOPCAPS_MODULATEINVCOLOR_ADDALPHA
|
5611 WINED3DTEXOPCAPS_MODULATEINVALPHA_ADDCOLOR
|
5612 WINED3DTEXOPCAPS_DOTPRODUCT3
|
5613 WINED3DTEXOPCAPS_MULTIPLYADD
|
5614 WINED3DTEXOPCAPS_LERP
|
5615 WINED3DTEXOPCAPS_BUMPENVMAP
|
5616 WINED3DTEXOPCAPS_BUMPENVMAPLUMINANCE
;
5618 /* TODO: Implement WINED3DTEXOPCAPS_PREMODULATE */
5620 caps
->MaxTextureBlendStages
= 8;
5621 caps
->MaxSimultaneousTextures
= min(gl_info
->limits
.fragment_samplers
, 8);
5624 static void state_texfactor_arbfp(DWORD state_id
,
5625 struct wined3d_stateblock
*stateblock
, struct wined3d_context
*context
)
5627 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5628 const struct wined3d_state
*state
= &stateblock
->state
;
5629 struct wined3d_device
*device
= stateblock
->device
;
5632 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
5633 * application provided constants
5635 if (device
->shader_backend
== &arb_program_shader_backend
)
5637 if (use_ps(state
)) return;
5639 context
->pshader_const_dirty
[ARB_FFP_CONST_TFACTOR
] = 1;
5640 device
->highest_dirty_ps_const
= max(device
->highest_dirty_ps_const
, ARB_FFP_CONST_TFACTOR
+ 1);
5643 D3DCOLORTOGLFLOAT4(state
->render_states
[WINED3DRS_TEXTUREFACTOR
], col
);
5644 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_TFACTOR
, col
));
5645 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_TFACTOR, col)");
5649 static void state_arb_specularenable(DWORD state_id
,
5650 struct wined3d_stateblock
*stateblock
, struct wined3d_context
*context
)
5652 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5653 const struct wined3d_state
*state
= &stateblock
->state
;
5654 struct wined3d_device
*device
= stateblock
->device
;
5657 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
5658 * application provided constants
5660 if (device
->shader_backend
== &arb_program_shader_backend
)
5662 if (use_ps(state
)) return;
5664 context
->pshader_const_dirty
[ARB_FFP_CONST_SPECULAR_ENABLE
] = 1;
5665 device
->highest_dirty_ps_const
= max(device
->highest_dirty_ps_const
, ARB_FFP_CONST_SPECULAR_ENABLE
+ 1);
5668 if (state
->render_states
[WINED3DRS_SPECULARENABLE
])
5670 /* The specular color has no alpha */
5671 col
[0] = 1.0f
; col
[1] = 1.0f
;
5672 col
[2] = 1.0f
; col
[3] = 0.0f
;
5674 col
[0] = 0.0f
; col
[1] = 0.0f
;
5675 col
[2] = 0.0f
; col
[3] = 0.0f
;
5677 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_SPECULAR_ENABLE
, col
));
5678 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_SPECULAR_ENABLE, col)");
5681 static void set_bumpmat_arbfp(DWORD state_id
, struct wined3d_stateblock
*stateblock
, struct wined3d_context
*context
)
5683 DWORD stage
= (state_id
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
5684 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5685 const struct wined3d_state
*state
= &stateblock
->state
;
5686 struct wined3d_device
*device
= stateblock
->device
;
5691 if (stage
&& (state
->pixel_shader
->reg_maps
.bumpmat
& (1 << stage
)))
5693 /* The pixel shader has to know the bump env matrix. Do a constants update if it isn't scheduled
5696 if (!isStateDirty(context
, STATE_PIXELSHADERCONSTANT
))
5697 stateblock_apply_state(STATE_PIXELSHADERCONSTANT
, stateblock
, context
);
5700 if(device
->shader_backend
== &arb_program_shader_backend
) {
5701 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants */
5704 } else if(device
->shader_backend
== &arb_program_shader_backend
) {
5705 context
->pshader_const_dirty
[ARB_FFP_CONST_BUMPMAT(stage
)] = 1;
5706 device
->highest_dirty_ps_const
= max(device
->highest_dirty_ps_const
, ARB_FFP_CONST_BUMPMAT(stage
) + 1);
5709 mat
[0][0] = *((float *)&state
->texture_states
[stage
][WINED3DTSS_BUMPENVMAT00
]);
5710 mat
[0][1] = *((float *)&state
->texture_states
[stage
][WINED3DTSS_BUMPENVMAT01
]);
5711 mat
[1][0] = *((float *)&state
->texture_states
[stage
][WINED3DTSS_BUMPENVMAT10
]);
5712 mat
[1][1] = *((float *)&state
->texture_states
[stage
][WINED3DTSS_BUMPENVMAT11
]);
5714 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_BUMPMAT(stage
), &mat
[0][0]));
5715 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_BUMPMAT(stage), &mat[0][0])");
5718 static void tex_bumpenvlum_arbfp(DWORD state_id
,
5719 struct wined3d_stateblock
*stateblock
, struct wined3d_context
*context
)
5721 DWORD stage
= (state_id
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
5722 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5723 const struct wined3d_state
*state
= &stateblock
->state
;
5724 struct wined3d_device
*device
= stateblock
->device
;
5729 if (stage
&& (state
->pixel_shader
->reg_maps
.luminanceparams
& (1 << stage
)))
5731 /* The pixel shader has to know the luminance offset. Do a constants update if it
5732 * isn't scheduled anyway
5734 if (!isStateDirty(context
, STATE_PIXELSHADERCONSTANT
))
5735 stateblock_apply_state(STATE_PIXELSHADERCONSTANT
, stateblock
, context
);
5738 if(device
->shader_backend
== &arb_program_shader_backend
) {
5739 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants */
5742 } else if(device
->shader_backend
== &arb_program_shader_backend
) {
5743 context
->pshader_const_dirty
[ARB_FFP_CONST_LUMINANCE(stage
)] = 1;
5744 device
->highest_dirty_ps_const
= max(device
->highest_dirty_ps_const
, ARB_FFP_CONST_LUMINANCE(stage
) + 1);
5747 param
[0] = *((float *)&state
->texture_states
[stage
][WINED3DTSS_BUMPENVLSCALE
]);
5748 param
[1] = *((float *)&state
->texture_states
[stage
][WINED3DTSS_BUMPENVLOFFSET
]);
5752 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_LUMINANCE(stage
), param
));
5753 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_LUMINANCE(stage), param)");
5756 static const char *get_argreg(struct wined3d_shader_buffer
*buffer
, DWORD argnum
, unsigned int stage
, DWORD arg
)
5760 if(arg
== ARG_UNUSED
) return "unused"; /* This is the marker for unused registers */
5762 switch(arg
& WINED3DTA_SELECTMASK
) {
5763 case WINED3DTA_DIFFUSE
:
5764 ret
= "fragment.color.primary"; break;
5766 case WINED3DTA_CURRENT
:
5767 if (!stage
) ret
= "fragment.color.primary";
5771 case WINED3DTA_TEXTURE
:
5773 case 0: ret
= "tex0"; break;
5774 case 1: ret
= "tex1"; break;
5775 case 2: ret
= "tex2"; break;
5776 case 3: ret
= "tex3"; break;
5777 case 4: ret
= "tex4"; break;
5778 case 5: ret
= "tex5"; break;
5779 case 6: ret
= "tex6"; break;
5780 case 7: ret
= "tex7"; break;
5781 default: ret
= "unknown texture";
5785 case WINED3DTA_TFACTOR
:
5786 ret
= "tfactor"; break;
5788 case WINED3DTA_SPECULAR
:
5789 ret
= "fragment.color.secondary"; break;
5791 case WINED3DTA_TEMP
:
5792 ret
= "tempreg"; break;
5794 case WINED3DTA_CONSTANT
:
5795 FIXME("Implement perstage constants\n");
5797 case 0: ret
= "const0"; break;
5798 case 1: ret
= "const1"; break;
5799 case 2: ret
= "const2"; break;
5800 case 3: ret
= "const3"; break;
5801 case 4: ret
= "const4"; break;
5802 case 5: ret
= "const5"; break;
5803 case 6: ret
= "const6"; break;
5804 case 7: ret
= "const7"; break;
5805 default: ret
= "unknown constant";
5813 if(arg
& WINED3DTA_COMPLEMENT
) {
5814 shader_addline(buffer
, "SUB arg%u, const.x, %s;\n", argnum
, ret
);
5815 if(argnum
== 0) ret
= "arg0";
5816 if(argnum
== 1) ret
= "arg1";
5817 if(argnum
== 2) ret
= "arg2";
5819 if(arg
& WINED3DTA_ALPHAREPLICATE
) {
5820 shader_addline(buffer
, "MOV arg%u, %s.w;\n", argnum
, ret
);
5821 if(argnum
== 0) ret
= "arg0";
5822 if(argnum
== 1) ret
= "arg1";
5823 if(argnum
== 2) ret
= "arg2";
5828 static void gen_ffp_instr(struct wined3d_shader_buffer
*buffer
, unsigned int stage
, BOOL color
,
5829 BOOL alpha
, DWORD dst
, DWORD op
, DWORD dw_arg0
, DWORD dw_arg1
, DWORD dw_arg2
)
5831 const char *dstmask
, *dstreg
, *arg0
, *arg1
, *arg2
;
5832 unsigned int mul
= 1;
5833 BOOL mul_final_dest
= FALSE
;
5835 if(color
&& alpha
) dstmask
= "";
5836 else if(color
) dstmask
= ".xyz";
5837 else dstmask
= ".w";
5839 if(dst
== tempreg
) dstreg
= "tempreg";
5840 else dstreg
= "ret";
5842 arg0
= get_argreg(buffer
, 0, stage
, dw_arg0
);
5843 arg1
= get_argreg(buffer
, 1, stage
, dw_arg1
);
5844 arg2
= get_argreg(buffer
, 2, stage
, dw_arg2
);
5847 case WINED3DTOP_DISABLE
:
5848 if (!stage
) shader_addline(buffer
, "MOV %s%s, fragment.color.primary;\n", dstreg
, dstmask
);
5851 case WINED3DTOP_SELECTARG2
:
5854 case WINED3DTOP_SELECTARG1
:
5855 shader_addline(buffer
, "MOV %s%s, %s;\n", dstreg
, dstmask
, arg1
);
5858 case WINED3DTOP_MODULATE4X
:
5861 case WINED3DTOP_MODULATE2X
:
5863 if (!strcmp(dstreg
, "result.color"))
5866 mul_final_dest
= TRUE
;
5869 case WINED3DTOP_MODULATE
:
5870 shader_addline(buffer
, "MUL %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
5873 case WINED3DTOP_ADDSIGNED2X
:
5875 if (!strcmp(dstreg
, "result.color"))
5878 mul_final_dest
= TRUE
;
5881 case WINED3DTOP_ADDSIGNED
:
5882 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
5885 case WINED3DTOP_ADD
:
5886 shader_addline(buffer
, "ADD_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
5889 case WINED3DTOP_SUBTRACT
:
5890 shader_addline(buffer
, "SUB_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
5893 case WINED3DTOP_ADDSMOOTH
:
5894 shader_addline(buffer
, "SUB arg1, const.x, %s;\n", arg1
);
5895 shader_addline(buffer
, "MAD_SAT %s%s, arg1, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
5898 case WINED3DTOP_BLENDCURRENTALPHA
:
5899 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_CURRENT
);
5900 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
5902 case WINED3DTOP_BLENDFACTORALPHA
:
5903 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TFACTOR
);
5904 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
5906 case WINED3DTOP_BLENDTEXTUREALPHA
:
5907 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
5908 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
5910 case WINED3DTOP_BLENDDIFFUSEALPHA
:
5911 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_DIFFUSE
);
5912 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
5915 case WINED3DTOP_BLENDTEXTUREALPHAPM
:
5916 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
5917 shader_addline(buffer
, "SUB arg0.w, const.x, %s.w;\n", arg0
);
5918 shader_addline(buffer
, "MAD_SAT %s%s, %s, arg0.w, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
5921 /* D3DTOP_PREMODULATE ???? */
5923 case WINED3DTOP_MODULATEINVALPHA_ADDCOLOR
:
5924 shader_addline(buffer
, "SUB arg0.w, const.x, %s;\n", arg1
);
5925 shader_addline(buffer
, "MAD_SAT %s%s, arg0.w, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
5927 case WINED3DTOP_MODULATEALPHA_ADDCOLOR
:
5928 shader_addline(buffer
, "MAD_SAT %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
5930 case WINED3DTOP_MODULATEINVCOLOR_ADDALPHA
:
5931 shader_addline(buffer
, "SUB arg0, const.x, %s;\n", arg1
);
5932 shader_addline(buffer
, "MAD_SAT %s%s, arg0, %s, %s.w;\n", dstreg
, dstmask
, arg2
, arg1
);
5934 case WINED3DTOP_MODULATECOLOR_ADDALPHA
:
5935 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s.w;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
5938 case WINED3DTOP_DOTPRODUCT3
:
5940 if (!strcmp(dstreg
, "result.color"))
5943 mul_final_dest
= TRUE
;
5945 shader_addline(buffer
, "SUB arg1, %s, const.w;\n", arg1
);
5946 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
5947 shader_addline(buffer
, "DP3_SAT %s%s, arg1, arg2;\n", dstreg
, dstmask
);
5950 case WINED3DTOP_MULTIPLYADD
:
5951 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg0
);
5954 case WINED3DTOP_LERP
:
5955 /* The msdn is not quite right here */
5956 shader_addline(buffer
, "LRP %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
5959 case WINED3DTOP_BUMPENVMAP
:
5960 case WINED3DTOP_BUMPENVMAPLUMINANCE
:
5961 /* Those are handled in the first pass of the shader(generation pass 1 and 2) already */
5965 FIXME("Unhandled texture op %08x\n", op
);
5969 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.y;\n", mul_final_dest
? "result.color" : dstreg
, dstmask
, dstreg
);
5970 } else if(mul
== 4) {
5971 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.z;\n", mul_final_dest
? "result.color" : dstreg
, dstmask
, dstreg
);
5975 static GLuint
gen_arbfp_ffp_shader(const struct ffp_frag_settings
*settings
, struct wined3d_stateblock
*stateblock
)
5977 const struct wined3d_gl_info
*gl_info
= &stateblock
->device
->adapter
->gl_info
;
5979 struct wined3d_shader_buffer buffer
;
5980 BOOL tex_read
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
5981 BOOL bump_used
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
5982 BOOL luminance_used
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
5983 const char *textype
;
5984 const char *instr
, *sat
;
5985 char colorcor_dst
[8];
5987 DWORD arg0
, arg1
, arg2
;
5988 BOOL tempreg_used
= FALSE
, tfactor_used
= FALSE
;
5990 const char *final_combiner_src
= "ret";
5993 /* Find out which textures are read */
5994 for(stage
= 0; stage
< MAX_TEXTURES
; stage
++) {
5995 if(settings
->op
[stage
].cop
== WINED3DTOP_DISABLE
) break;
5996 arg0
= settings
->op
[stage
].carg0
& WINED3DTA_SELECTMASK
;
5997 arg1
= settings
->op
[stage
].carg1
& WINED3DTA_SELECTMASK
;
5998 arg2
= settings
->op
[stage
].carg2
& WINED3DTA_SELECTMASK
;
5999 if(arg0
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6000 if(arg1
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6001 if(arg2
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6003 if(settings
->op
[stage
].cop
== WINED3DTOP_BLENDTEXTUREALPHA
) tex_read
[stage
] = TRUE
;
6004 if(settings
->op
[stage
].cop
== WINED3DTOP_BLENDTEXTUREALPHAPM
) tex_read
[stage
] = TRUE
;
6005 if(settings
->op
[stage
].cop
== WINED3DTOP_BUMPENVMAP
) {
6006 bump_used
[stage
] = TRUE
;
6007 tex_read
[stage
] = TRUE
;
6009 if(settings
->op
[stage
].cop
== WINED3DTOP_BUMPENVMAPLUMINANCE
) {
6010 bump_used
[stage
] = TRUE
;
6011 tex_read
[stage
] = TRUE
;
6012 luminance_used
[stage
] = TRUE
;
6013 } else if(settings
->op
[stage
].cop
== WINED3DTOP_BLENDFACTORALPHA
) {
6014 tfactor_used
= TRUE
;
6017 if(arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
) {
6018 tfactor_used
= TRUE
;
6021 if(settings
->op
[stage
].dst
== tempreg
) tempreg_used
= TRUE
;
6022 if(arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
) {
6023 tempreg_used
= TRUE
;
6026 if(settings
->op
[stage
].aop
== WINED3DTOP_DISABLE
) continue;
6027 arg0
= settings
->op
[stage
].aarg0
& WINED3DTA_SELECTMASK
;
6028 arg1
= settings
->op
[stage
].aarg1
& WINED3DTA_SELECTMASK
;
6029 arg2
= settings
->op
[stage
].aarg2
& WINED3DTA_SELECTMASK
;
6030 if(arg0
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6031 if(arg1
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6032 if(arg2
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6034 if(arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
) {
6035 tempreg_used
= TRUE
;
6037 if(arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
) {
6038 tfactor_used
= TRUE
;
6043 if (!shader_buffer_init(&buffer
))
6045 ERR("Failed to initialize shader buffer.\n");
6049 shader_addline(&buffer
, "!!ARBfp1.0\n");
6051 switch(settings
->fog
) {
6052 case FOG_OFF
: break;
6053 case FOG_LINEAR
: shader_addline(&buffer
, "OPTION ARB_fog_linear;\n"); break;
6054 case FOG_EXP
: shader_addline(&buffer
, "OPTION ARB_fog_exp;\n"); break;
6055 case FOG_EXP2
: shader_addline(&buffer
, "OPTION ARB_fog_exp2;\n"); break;
6056 default: FIXME("Unexpected fog setting %d\n", settings
->fog
);
6059 shader_addline(&buffer
, "PARAM const = {1, 2, 4, 0.5};\n");
6060 shader_addline(&buffer
, "TEMP TMP;\n");
6061 shader_addline(&buffer
, "TEMP ret;\n");
6062 if(tempreg_used
|| settings
->sRGB_write
) shader_addline(&buffer
, "TEMP tempreg;\n");
6063 shader_addline(&buffer
, "TEMP arg0;\n");
6064 shader_addline(&buffer
, "TEMP arg1;\n");
6065 shader_addline(&buffer
, "TEMP arg2;\n");
6066 for(stage
= 0; stage
< MAX_TEXTURES
; stage
++) {
6067 if(!tex_read
[stage
]) continue;
6068 shader_addline(&buffer
, "TEMP tex%u;\n", stage
);
6069 if(!bump_used
[stage
]) continue;
6070 shader_addline(&buffer
, "PARAM bumpmat%u = program.env[%u];\n", stage
, ARB_FFP_CONST_BUMPMAT(stage
));
6071 if(!luminance_used
[stage
]) continue;
6072 shader_addline(&buffer
, "PARAM luminance%u = program.env[%u];\n", stage
, ARB_FFP_CONST_LUMINANCE(stage
));
6075 shader_addline(&buffer
, "PARAM tfactor = program.env[%u];\n", ARB_FFP_CONST_TFACTOR
);
6077 shader_addline(&buffer
, "PARAM specular_enable = program.env[%u];\n", ARB_FFP_CONST_SPECULAR_ENABLE
);
6079 if(settings
->sRGB_write
) {
6080 shader_addline(&buffer
, "PARAM srgb_consts1 = {%f, %f, %f, %f};\n",
6081 srgb_mul_low
, srgb_cmp
, srgb_pow
, srgb_mul_high
);
6082 shader_addline(&buffer
, "PARAM srgb_consts2 = {%f, %f, %f, %f};\n",
6083 srgb_sub_high
, 0.0, 0.0, 0.0);
6086 if (ffp_clip_emul(&stateblock
->state
) && settings
->emul_clipplanes
)
6087 shader_addline(&buffer
, "KIL fragment.texcoord[7];\n");
6089 /* Generate texture sampling instructions) */
6090 for(stage
= 0; stage
< MAX_TEXTURES
&& settings
->op
[stage
].cop
!= WINED3DTOP_DISABLE
; stage
++) {
6091 if(!tex_read
[stage
]) continue;
6093 switch(settings
->op
[stage
].tex_type
) {
6094 case tex_1d
: textype
= "1D"; break;
6095 case tex_2d
: textype
= "2D"; break;
6096 case tex_3d
: textype
= "3D"; break;
6097 case tex_cube
: textype
= "CUBE"; break;
6098 case tex_rect
: textype
= "RECT"; break;
6099 default: textype
= "unexpected_textype"; break;
6102 if(settings
->op
[stage
].cop
== WINED3DTOP_BUMPENVMAP
||
6103 settings
->op
[stage
].cop
== WINED3DTOP_BUMPENVMAPLUMINANCE
) {
6109 if(settings
->op
[stage
].projected
== proj_none
) {
6111 } else if(settings
->op
[stage
].projected
== proj_count4
||
6112 settings
->op
[stage
].projected
== proj_count3
) {
6115 FIXME("Unexpected projection mode %d\n", settings
->op
[stage
].projected
);
6120 (settings
->op
[stage
- 1].cop
== WINED3DTOP_BUMPENVMAP
||
6121 settings
->op
[stage
- 1].cop
== WINED3DTOP_BUMPENVMAPLUMINANCE
)) {
6122 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, x, z, 0, 0;\n", stage
- 1);
6123 shader_addline(&buffer
, "DP3 ret.x, arg1, tex%u;\n", stage
- 1);
6124 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, y, w, 0, 0;\n", stage
- 1);
6125 shader_addline(&buffer
, "DP3 ret.y, arg1, tex%u;\n", stage
- 1);
6127 /* with projective textures, texbem only divides the static texture coord, not the displacement,
6128 * so multiply the displacement with the dividing parameter before passing it to TXP
6130 if (settings
->op
[stage
].projected
!= proj_none
) {
6131 if(settings
->op
[stage
].projected
== proj_count4
) {
6132 shader_addline(&buffer
, "MOV ret.w, fragment.texcoord[%u].w;\n", stage
);
6133 shader_addline(&buffer
, "MUL ret.xyz, ret, fragment.texcoord[%u].w, fragment.texcoord[%u];\n", stage
, stage
);
6135 shader_addline(&buffer
, "MOV ret.w, fragment.texcoord[%u].z;\n", stage
);
6136 shader_addline(&buffer
, "MAD ret.xyz, ret, fragment.texcoord[%u].z, fragment.texcoord[%u];\n", stage
, stage
);
6139 shader_addline(&buffer
, "ADD ret, ret, fragment.texcoord[%u];\n", stage
);
6142 shader_addline(&buffer
, "%s%s tex%u, ret, texture[%u], %s;\n",
6143 instr
, sat
, stage
, stage
, textype
);
6144 if(settings
->op
[stage
- 1].cop
== WINED3DTOP_BUMPENVMAPLUMINANCE
) {
6145 shader_addline(&buffer
, "MAD_SAT ret.x, tex%u.z, luminance%u.x, luminance%u.y;\n",
6146 stage
- 1, stage
- 1, stage
- 1);
6147 shader_addline(&buffer
, "MUL tex%u, tex%u, ret.x;\n", stage
, stage
);
6149 } else if(settings
->op
[stage
].projected
== proj_count3
) {
6150 shader_addline(&buffer
, "MOV ret, fragment.texcoord[%u];\n", stage
);
6151 shader_addline(&buffer
, "MOV ret.w, ret.z;\n");
6152 shader_addline(&buffer
, "%s%s tex%u, ret, texture[%u], %s;\n",
6153 instr
, sat
, stage
, stage
, textype
);
6155 shader_addline(&buffer
, "%s%s tex%u, fragment.texcoord[%u], texture[%u], %s;\n",
6156 instr
, sat
, stage
, stage
, stage
, textype
);
6159 sprintf(colorcor_dst
, "tex%u", stage
);
6160 gen_color_correction(&buffer
, colorcor_dst
, WINED3DSP_WRITEMASK_ALL
, "const.x", "const.y",
6161 settings
->op
[stage
].color_fixup
);
6164 /* Generate the main shader */
6165 for (stage
= 0; stage
< MAX_TEXTURES
; ++stage
)
6167 if (settings
->op
[stage
].cop
== WINED3DTOP_DISABLE
)
6169 if (!stage
) final_combiner_src
= "fragment.color.primary";
6173 if(settings
->op
[stage
].cop
== WINED3DTOP_SELECTARG1
&&
6174 settings
->op
[stage
].aop
== WINED3DTOP_SELECTARG1
) {
6175 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
;
6176 } else if(settings
->op
[stage
].cop
== WINED3DTOP_SELECTARG1
&&
6177 settings
->op
[stage
].aop
== WINED3DTOP_SELECTARG2
) {
6178 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg2
;
6179 } else if(settings
->op
[stage
].cop
== WINED3DTOP_SELECTARG2
&&
6180 settings
->op
[stage
].aop
== WINED3DTOP_SELECTARG1
) {
6181 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg1
;
6182 } else if(settings
->op
[stage
].cop
== WINED3DTOP_SELECTARG2
&&
6183 settings
->op
[stage
].aop
== WINED3DTOP_SELECTARG2
) {
6184 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
6186 op_equal
= settings
->op
[stage
].aop
== settings
->op
[stage
].cop
&&
6187 settings
->op
[stage
].carg0
== settings
->op
[stage
].aarg0
&&
6188 settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
&&
6189 settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
6192 if(settings
->op
[stage
].aop
== WINED3DTOP_DISABLE
) {
6193 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].dst
,
6194 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6195 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6197 shader_addline(&buffer
, "MOV ret.w, fragment.color.primary.w;\n");
6201 gen_ffp_instr(&buffer
, stage
, TRUE
, TRUE
, settings
->op
[stage
].dst
,
6202 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6203 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6205 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].dst
,
6206 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6207 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6208 gen_ffp_instr(&buffer
, stage
, FALSE
, TRUE
, settings
->op
[stage
].dst
,
6209 settings
->op
[stage
].aop
, settings
->op
[stage
].aarg0
,
6210 settings
->op
[stage
].aarg1
, settings
->op
[stage
].aarg2
);
6214 if(settings
->sRGB_write
) {
6215 shader_addline(&buffer
, "MAD ret, fragment.color.secondary, specular_enable, %s;\n", final_combiner_src
);
6216 arbfp_add_sRGB_correction(&buffer
, "ret", "arg0", "arg1", "arg2", "tempreg", FALSE
);
6217 shader_addline(&buffer
, "MOV result.color, ret;\n");
6219 shader_addline(&buffer
, "MAD result.color, fragment.color.secondary, specular_enable, %s;\n", final_combiner_src
);
6223 shader_addline(&buffer
, "END\n");
6225 /* Generate the shader */
6226 GL_EXTCALL(glGenProgramsARB(1, &ret
));
6227 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, ret
));
6228 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
6229 strlen(buffer
.buffer
), buffer
.buffer
));
6230 checkGLcall("glProgramStringARB()");
6232 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
6235 FIXME("Fragment program error at position %d: %s\n\n", pos
,
6236 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
6237 shader_arb_dump_program_source(buffer
.buffer
);
6243 GL_EXTCALL(glGetProgramivARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
6244 checkGLcall("glGetProgramivARB()");
6245 if (!native
) WARN("Program exceeds native resource limits.\n");
6248 shader_buffer_free(&buffer
);
6252 static void fragment_prog_arbfp(DWORD state_id
, struct wined3d_stateblock
*stateblock
, struct wined3d_context
*context
)
6254 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
6255 const struct wined3d_state
*state
= &stateblock
->state
;
6256 struct wined3d_device
*device
= stateblock
->device
;
6257 struct shader_arb_priv
*priv
= device
->fragment_priv
;
6258 BOOL use_vshader
= use_vs(state
);
6259 BOOL use_pshader
= use_ps(state
);
6260 struct ffp_frag_settings settings
;
6261 const struct arbfp_ffp_desc
*desc
;
6264 TRACE("state_id %#x, stateblock %p, context %p\n", state_id
, stateblock
, context
);
6266 if(isStateDirty(context
, STATE_RENDER(WINED3DRS_FOGENABLE
))) {
6267 if(!use_pshader
&& device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
) {
6268 /* Reload fixed function constants since they collide with the pixel shader constants */
6269 for(i
= 0; i
< MAX_TEXTURES
; i
++) {
6270 set_bumpmat_arbfp(STATE_TEXTURESTAGE(i
, WINED3DTSS_BUMPENVMAT00
), stateblock
, context
);
6272 state_texfactor_arbfp(STATE_RENDER(WINED3DRS_TEXTUREFACTOR
), stateblock
, context
);
6273 state_arb_specularenable(STATE_RENDER(WINED3DRS_SPECULARENABLE
), stateblock
, context
);
6274 } else if(use_pshader
&& !isStateDirty(context
, device
->StateTable
[STATE_VSHADER
].representative
)) {
6275 device
->shader_backend
->shader_select(context
, use_pshader
, use_vshader
);
6281 /* Find or create a shader implementing the fixed function pipeline settings, then activate it */
6282 gen_ffp_frag_op(stateblock
, &settings
, FALSE
);
6283 desc
= (const struct arbfp_ffp_desc
*)find_ffp_frag_shader(&priv
->fragment_shaders
, &settings
);
6285 struct arbfp_ffp_desc
*new_desc
= HeapAlloc(GetProcessHeap(), 0, sizeof(*new_desc
));
6288 ERR("Out of memory\n");
6291 new_desc
->num_textures_used
= 0;
6292 for (i
= 0; i
< gl_info
->limits
.texture_stages
; ++i
)
6294 if(settings
.op
[i
].cop
== WINED3DTOP_DISABLE
) break;
6295 new_desc
->num_textures_used
= i
;
6298 memcpy(&new_desc
->parent
.settings
, &settings
, sizeof(settings
));
6299 new_desc
->shader
= gen_arbfp_ffp_shader(&settings
, stateblock
);
6300 add_ffp_frag_shader(&priv
->fragment_shaders
, &new_desc
->parent
);
6301 TRACE("Allocated fixed function replacement shader descriptor %p\n", new_desc
);
6305 /* Now activate the replacement program. GL_FRAGMENT_PROGRAM_ARB is already active(however, note the
6306 * comment above the shader_select call below). If e.g. GLSL is active, the shader_select call will
6309 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, desc
->shader
));
6310 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, desc->shader)");
6311 priv
->current_fprogram_id
= desc
->shader
;
6313 if(device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
) {
6314 /* Reload fixed function constants since they collide with the pixel shader constants */
6315 for(i
= 0; i
< MAX_TEXTURES
; i
++) {
6316 set_bumpmat_arbfp(STATE_TEXTURESTAGE(i
, WINED3DTSS_BUMPENVMAT00
), stateblock
, context
);
6318 state_texfactor_arbfp(STATE_RENDER(WINED3DRS_TEXTUREFACTOR
), stateblock
, context
);
6319 state_arb_specularenable(STATE_RENDER(WINED3DRS_SPECULARENABLE
), stateblock
, context
);
6321 context
->last_was_pshader
= FALSE
;
6323 context
->last_was_pshader
= TRUE
;
6326 /* Finally, select the shader. If a pixel shader is used, it will be set and enabled by the shader backend.
6327 * If this shader backend is arbfp(most likely), then it will simply overwrite the last fixed function replace-
6328 * ment shader. If the shader backend is not ARB, it currently is important that the opengl implementation
6329 * type overwrites GL_ARB_fragment_program. This is currently the case with GLSL. If we really want to use
6330 * atifs or nvrc pixel shaders with arb fragment programs we'd have to disable GL_FRAGMENT_PROGRAM_ARB here
6332 * Don't call shader_select if the vertex shader is dirty, because it will be called later on by the vertex
6335 if(!isStateDirty(context
, device
->StateTable
[STATE_VSHADER
].representative
)) {
6336 device
->shader_backend
->shader_select(context
, use_pshader
, use_vshader
);
6338 if (!isStateDirty(context
, STATE_VERTEXSHADERCONSTANT
) && (use_vshader
|| use_pshader
))
6339 stateblock_apply_state(STATE_VERTEXSHADERCONSTANT
, stateblock
, context
);
6341 if (use_pshader
) stateblock_apply_state(STATE_PIXELSHADERCONSTANT
, stateblock
, context
);
6344 /* We can't link the fog states to the fragment state directly since the vertex pipeline links them
6345 * to FOGENABLE. A different linking in different pipeline parts can't be expressed in the combined
6346 * state table, so we need to handle that with a forwarding function. The other invisible side effect
6347 * is that changing the fog start and fog end(which links to FOGENABLE in vertex) results in the
6348 * fragment_prog_arbfp function being called because FOGENABLE is dirty, which calls this function here
6350 static void state_arbfp_fog(DWORD state_id
, struct wined3d_stateblock
*stateblock
, struct wined3d_context
*context
)
6352 const struct wined3d_state
*state
= &stateblock
->state
;
6353 enum fogsource new_source
;
6355 TRACE("state_id %#x, stateblock %p, context %p\n", state_id
, stateblock
, context
);
6357 if(!isStateDirty(context
, STATE_PIXELSHADER
)) {
6358 fragment_prog_arbfp(state_id
, stateblock
, context
);
6361 if (!state
->render_states
[WINED3DRS_FOGENABLE
]) return;
6363 if (state
->render_states
[WINED3DRS_FOGTABLEMODE
] == WINED3DFOG_NONE
)
6367 new_source
= FOGSOURCE_VS
;
6371 if (state
->render_states
[WINED3DRS_FOGVERTEXMODE
] == WINED3DFOG_NONE
|| context
->last_was_rhw
)
6372 new_source
= FOGSOURCE_COORD
;
6374 new_source
= FOGSOURCE_FFP
;
6377 new_source
= FOGSOURCE_FFP
;
6379 if(new_source
!= context
->fog_source
) {
6380 context
->fog_source
= new_source
;
6381 state_fogstartend(STATE_RENDER(WINED3DRS_FOGSTART
), stateblock
, context
);
6385 static void textransform(DWORD state
, struct wined3d_stateblock
*stateblock
, struct wined3d_context
*context
)
6387 if (!isStateDirty(context
, STATE_PIXELSHADER
))
6388 fragment_prog_arbfp(state
, stateblock
, context
);
6391 static const struct StateEntryTemplate arbfp_fragmentstate_template
[] = {
6392 {STATE_RENDER(WINED3DRS_TEXTUREFACTOR
), { STATE_RENDER(WINED3DRS_TEXTUREFACTOR
), state_texfactor_arbfp
}, WINED3D_GL_EXT_NONE
},
6393 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6394 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6395 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6396 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6397 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6398 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6399 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6400 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6401 {STATE_TEXTURESTAGE(0, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6402 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6403 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6404 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6405 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6406 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6407 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6408 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6409 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6410 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6411 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6412 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6413 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6414 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6415 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6416 {STATE_TEXTURESTAGE(1, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6417 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6418 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6419 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6420 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6421 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6422 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6423 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6424 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6425 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6426 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6427 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6428 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6429 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6430 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6431 {STATE_TEXTURESTAGE(2, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6432 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6433 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6434 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6435 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6436 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6437 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6438 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6439 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6440 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6441 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6442 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6443 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6444 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6445 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6446 {STATE_TEXTURESTAGE(3, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6447 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6448 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6449 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6450 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6451 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6452 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6453 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6454 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6455 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6456 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6457 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6458 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6459 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6460 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6461 {STATE_TEXTURESTAGE(4, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6462 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6463 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6464 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6465 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6466 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6467 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6468 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6469 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6470 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6471 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6472 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6473 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6474 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6475 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6476 {STATE_TEXTURESTAGE(5, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6477 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6478 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6479 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6480 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6481 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6482 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6483 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6484 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6485 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6486 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6487 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6488 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6489 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6490 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6491 {STATE_TEXTURESTAGE(6, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6492 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6493 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6494 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6495 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6496 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6497 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6498 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6499 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6500 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6501 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6502 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6503 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6504 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6505 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6506 {STATE_TEXTURESTAGE(7, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6507 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6508 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6509 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6510 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6511 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6512 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6513 {STATE_PIXELSHADER
, { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
6514 {STATE_RENDER(WINED3DRS_FOGENABLE
), { STATE_RENDER(WINED3DRS_FOGENABLE
), state_arbfp_fog
}, WINED3D_GL_EXT_NONE
},
6515 {STATE_RENDER(WINED3DRS_FOGTABLEMODE
), { STATE_RENDER(WINED3DRS_FOGENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6516 {STATE_RENDER(WINED3DRS_FOGVERTEXMODE
), { STATE_RENDER(WINED3DRS_FOGENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6517 {STATE_RENDER(WINED3DRS_FOGSTART
), { STATE_RENDER(WINED3DRS_FOGSTART
), state_fogstartend
}, WINED3D_GL_EXT_NONE
},
6518 {STATE_RENDER(WINED3DRS_FOGEND
), { STATE_RENDER(WINED3DRS_FOGSTART
), NULL
}, WINED3D_GL_EXT_NONE
},
6519 {STATE_RENDER(WINED3DRS_SRGBWRITEENABLE
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6520 {STATE_RENDER(WINED3DRS_FOGCOLOR
), { STATE_RENDER(WINED3DRS_FOGCOLOR
), state_fogcolor
}, WINED3D_GL_EXT_NONE
},
6521 {STATE_RENDER(WINED3DRS_FOGDENSITY
), { STATE_RENDER(WINED3DRS_FOGDENSITY
), state_fogdensity
}, WINED3D_GL_EXT_NONE
},
6522 {STATE_TEXTURESTAGE(0,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(0, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6523 {STATE_TEXTURESTAGE(1,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(1, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6524 {STATE_TEXTURESTAGE(2,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(2, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6525 {STATE_TEXTURESTAGE(3,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(3, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6526 {STATE_TEXTURESTAGE(4,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(4, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6527 {STATE_TEXTURESTAGE(5,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(5, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6528 {STATE_TEXTURESTAGE(6,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(6, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6529 {STATE_TEXTURESTAGE(7,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(7, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6530 {STATE_RENDER(WINED3DRS_SPECULARENABLE
), { STATE_RENDER(WINED3DRS_SPECULARENABLE
), state_arb_specularenable
}, WINED3D_GL_EXT_NONE
},
6531 {0 /* Terminate */, { 0, 0 }, WINED3D_GL_EXT_NONE
},
6534 const struct fragment_pipeline arbfp_fragment_pipeline
= {
6539 shader_arb_color_fixup_supported
,
6540 arbfp_fragmentstate_template
,
6541 TRUE
/* We can disable projected textures */
6544 struct arbfp_blit_priv
{
6545 GLenum yuy2_rect_shader
, yuy2_2d_shader
;
6546 GLenum uyvy_rect_shader
, uyvy_2d_shader
;
6547 GLenum yv12_rect_shader
, yv12_2d_shader
;
6548 GLenum p8_rect_shader
, p8_2d_shader
;
6549 GLuint palette_texture
;
6552 static HRESULT
arbfp_blit_alloc(struct wined3d_device
*device
)
6554 device
->blit_priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct arbfp_blit_priv
));
6555 if(!device
->blit_priv
) {
6556 ERR("Out of memory\n");
6557 return E_OUTOFMEMORY
;
6562 /* Context activation is done by the caller. */
6563 static void arbfp_blit_free(struct wined3d_device
*device
)
6565 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
6566 struct arbfp_blit_priv
*priv
= device
->blit_priv
;
6569 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->yuy2_rect_shader
));
6570 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->yuy2_2d_shader
));
6571 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->uyvy_rect_shader
));
6572 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->uyvy_2d_shader
));
6573 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->yv12_rect_shader
));
6574 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->yv12_2d_shader
));
6575 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->p8_rect_shader
));
6576 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->p8_2d_shader
));
6577 checkGLcall("Delete yuv and p8 programs");
6579 if(priv
->palette_texture
) glDeleteTextures(1, &priv
->palette_texture
);
6582 HeapFree(GetProcessHeap(), 0, device
->blit_priv
);
6583 device
->blit_priv
= NULL
;
6586 static BOOL
gen_planar_yuv_read(struct wined3d_shader_buffer
*buffer
, enum complex_fixup fixup
,
6587 GLenum textype
, char *luminance
)
6590 const char *tex
, *texinstr
;
6592 if (fixup
== COMPLEX_FIXUP_UYVY
) {
6600 case GL_TEXTURE_2D
: tex
= "2D"; texinstr
= "TXP"; break;
6601 case GL_TEXTURE_RECTANGLE_ARB
: tex
= "RECT"; texinstr
= "TEX"; break;
6603 /* This is more tricky than just replacing the texture type - we have to navigate
6604 * properly in the texture to find the correct chroma values
6606 FIXME("Implement yuv correction for non-2d, non-rect textures\n");
6610 /* First we have to read the chroma values. This means we need at least two pixels(no filtering),
6611 * or 4 pixels(with filtering). To get the unmodified chromas, we have to rid ourselves of the
6612 * filtering when we sample the texture.
6614 * These are the rules for reading the chroma:
6620 * So we have to get the sampling x position in non-normalized coordinates in integers
6622 if(textype
!= GL_TEXTURE_RECTANGLE_ARB
) {
6623 shader_addline(buffer
, "MUL texcrd.xy, fragment.texcoord[0], size.x;\n");
6624 shader_addline(buffer
, "MOV texcrd.w, size.x;\n");
6626 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
6628 /* We must not allow filtering between pixel x and x+1, this would mix U and V
6629 * Vertical filtering is ok. However, bear in mind that the pixel center is at
6632 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
6633 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, coef.y;\n");
6635 /* Divide the x coordinate by 0.5 and get the fraction. This gives 0.25 and 0.75 for the
6636 * even and odd pixels respectively
6638 shader_addline(buffer
, "MUL texcrd2, texcrd, coef.y;\n");
6639 shader_addline(buffer
, "FRC texcrd2, texcrd2;\n");
6641 /* Sample Pixel 1 */
6642 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
6644 /* Put the value into either of the chroma values */
6645 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
6646 shader_addline(buffer
, "MUL chroma.x, luminance.%c, temp.x;\n", chroma
);
6647 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
6648 shader_addline(buffer
, "MUL chroma.y, luminance.%c, temp.x;\n", chroma
);
6650 /* Sample pixel 2. If we read an even pixel(SLT above returned 1), sample
6651 * the pixel right to the current one. Otherwise, sample the left pixel.
6652 * Bias and scale the SLT result to -1;1 and add it to the texcrd.x.
6654 shader_addline(buffer
, "MAD temp.x, temp.x, coef.z, -coef.x;\n");
6655 shader_addline(buffer
, "ADD texcrd.x, texcrd, temp.x;\n");
6656 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
6658 /* Put the value into the other chroma */
6659 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
6660 shader_addline(buffer
, "MAD chroma.y, luminance.%c, temp.x, chroma.y;\n", chroma
);
6661 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
6662 shader_addline(buffer
, "MAD chroma.x, luminance.%c, temp.x, chroma.x;\n", chroma
);
6664 /* TODO: If filtering is enabled, sample a 2nd pair of pixels left or right of
6665 * the current one and lerp the two U and V values
6668 /* This gives the correctly filtered luminance value */
6669 shader_addline(buffer
, "TEX luminance, fragment.texcoord[0], texture[0], %s;\n", tex
);
6674 static BOOL
gen_yv12_read(struct wined3d_shader_buffer
*buffer
, GLenum textype
, char *luminance
)
6679 case GL_TEXTURE_2D
: tex
= "2D"; break;
6680 case GL_TEXTURE_RECTANGLE_ARB
: tex
= "RECT"; break;
6682 FIXME("Implement yv12 correction for non-2d, non-rect textures\n");
6686 /* YV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
6687 * V and a (W/2)x(H/2) U plane, each with 8 bit per pixel. So the effective
6688 * bitdepth is 12 bits per pixel. Since the U and V planes have only half the
6689 * pitch of the luminance plane, the packing into the gl texture is a bit
6690 * unfortunate. If the whole texture is interpreted as luminance data it looks
6691 * approximately like this:
6693 * +----------------------------------+----
6705 * +----------------+-----------------+----
6707 * | U even rows | U odd rows |
6709 * +----------------+------------------ -
6711 * | V even rows | V odd rows |
6713 * +----------------+-----------------+----
6717 * So it appears as if there are 4 chroma images, but in fact the odd rows
6718 * in the chroma images are in the same row as the even ones. So its is
6719 * kinda tricky to read
6721 * When reading from rectangle textures, keep in mind that the input y coordinates
6722 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height
6724 shader_addline(buffer
, "PARAM yv12_coef = {%f, %f, %f, %f};\n",
6725 2.0f
/ 3.0f
, 1.0f
/ 6.0f
, (2.0f
/ 3.0f
) + (1.0f
/ 6.0f
), 1.0f
/ 3.0f
);
6727 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
6728 /* the chroma planes have only half the width */
6729 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, coef.y;\n");
6731 /* The first value is between 2/3 and 5/6th of the texture's height, so scale+bias
6732 * the coordinate. Also read the right side of the image when reading odd lines
6734 * Don't forget to clamp the y values in into the range, otherwise we'll get filtering
6737 if(textype
== GL_TEXTURE_2D
) {
6739 shader_addline(buffer
, "RCP chroma.w, size.y;\n");
6741 shader_addline(buffer
, "MUL texcrd2.y, texcrd.y, size.y;\n");
6743 shader_addline(buffer
, "FLR texcrd2.y, texcrd2.y;\n");
6744 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, yv12_coef.y, yv12_coef.x;\n");
6746 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
6747 shader_addline(buffer
, "ADD texcrd2.x, texcrd2.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
6748 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
6749 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
6750 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
6752 /* clamp, keep the half pixel origin in mind */
6753 shader_addline(buffer
, "MAD temp.y, coef.y, chroma.w, yv12_coef.x;\n");
6754 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
6755 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.z;\n");
6756 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
6758 /* Read from [size - size+size/4] */
6759 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
6760 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, coef.w, size.y;\n");
6762 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
6763 shader_addline(buffer
, "ADD texcrd2.x, texcrd.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
6764 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
6765 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
6766 shader_addline(buffer
, "MUL texcrd2.x, texcrd2.x, size.x;\n");
6767 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
6769 /* Make sure to read exactly from the pixel center */
6770 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
6771 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, coef.y;\n");
6774 shader_addline(buffer
, "MAD temp.y, size.y, coef.w, size.y;\n");
6775 shader_addline(buffer
, "ADD temp.y, temp.y, -coef.y;\n");
6776 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
6777 shader_addline(buffer
, "ADD temp.y, size.y, -coef.y;\n");
6778 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
6780 /* Read the texture, put the result into the output register */
6781 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
6782 shader_addline(buffer
, "MOV chroma.x, temp.w;\n");
6784 /* The other chroma value is 1/6th of the texture lower, from 5/6th to 6/6th
6785 * No need to clamp because we're just reusing the already clamped value from above
6787 if(textype
== GL_TEXTURE_2D
) {
6788 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, yv12_coef.y;\n");
6790 shader_addline(buffer
, "MAD texcrd.y, size.y, coef.w, texcrd.y;\n");
6792 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
6793 shader_addline(buffer
, "MOV chroma.y, temp.w;\n");
6795 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
6796 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
6797 * values due to filtering
6799 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
6800 if(textype
== GL_TEXTURE_2D
) {
6801 /* Multiply the y coordinate by 2/3 and clamp it */
6802 shader_addline(buffer
, "MUL texcrd.y, texcrd.y, yv12_coef.x;\n");
6803 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.x;\n");
6804 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
6805 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
6807 /* Reading from texture_rectangles is pretty straightforward, just use the unmodified
6808 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
6811 shader_addline(buffer
, "ADD temp.x, size.y, -coef.y;\n");
6812 shader_addline(buffer
, "MIN texcrd.y, texcrd.y, size.x;\n");
6813 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
6820 static GLuint
gen_p8_shader(struct arbfp_blit_priv
*priv
,
6821 const struct wined3d_gl_info
*gl_info
, GLenum textype
)
6824 struct wined3d_shader_buffer buffer
;
6828 if (!shader_buffer_init(&buffer
))
6830 ERR("Failed to initialize shader buffer.\n");
6835 GL_EXTCALL(glGenProgramsARB(1, &shader
));
6836 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
6839 shader_buffer_free(&buffer
);
6843 shader_addline(&buffer
, "!!ARBfp1.0\n");
6844 shader_addline(&buffer
, "TEMP index;\n");
6846 /* { 255/256, 0.5/255*255/256, 0, 0 } */
6847 shader_addline(&buffer
, "PARAM constants = { 0.996, 0.00195, 0, 0 };\n");
6849 /* The alpha-component contains the palette index */
6850 if(textype
== GL_TEXTURE_RECTANGLE_ARB
)
6851 shader_addline(&buffer
, "TXP index, fragment.texcoord[0], texture[0], RECT;\n");
6853 shader_addline(&buffer
, "TEX index, fragment.texcoord[0], texture[0], 2D;\n");
6855 /* Scale the index by 255/256 and add a bias of '0.5' in order to sample in the middle */
6856 shader_addline(&buffer
, "MAD index.a, index.a, constants.x, constants.y;\n");
6858 /* Use the alpha-component as an index in the palette to get the final color */
6859 shader_addline(&buffer
, "TEX result.color, index.a, texture[1], 1D;\n");
6860 shader_addline(&buffer
, "END\n");
6863 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
6864 strlen(buffer
.buffer
), buffer
.buffer
));
6865 checkGLcall("glProgramStringARB()");
6867 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
6870 FIXME("Fragment program error at position %d: %s\n\n", pos
,
6871 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
6872 shader_arb_dump_program_source(buffer
.buffer
);
6875 if (textype
== GL_TEXTURE_RECTANGLE_ARB
)
6876 priv
->p8_rect_shader
= shader
;
6878 priv
->p8_2d_shader
= shader
;
6880 shader_buffer_free(&buffer
);
6886 /* Context activation is done by the caller. */
6887 static void upload_palette(struct wined3d_surface
*surface
)
6890 struct wined3d_device
*device
= surface
->resource
.device
;
6891 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
6892 struct arbfp_blit_priv
*priv
= device
->blit_priv
;
6893 BOOL colorkey
= (surface
->CKeyFlags
& WINEDDSD_CKSRCBLT
) ? TRUE
: FALSE
;
6895 d3dfmt_p8_init_palette(surface
, table
, colorkey
);
6898 if (!priv
->palette_texture
)
6899 glGenTextures(1, &priv
->palette_texture
);
6901 GL_EXTCALL(glActiveTextureARB(GL_TEXTURE1
));
6902 glBindTexture(GL_TEXTURE_1D
, priv
->palette_texture
);
6904 glTexEnvi(GL_TEXTURE_ENV
, GL_TEXTURE_ENV_MODE
, GL_REPLACE
);
6906 glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_MIN_FILTER
, GL_NEAREST
);
6907 /* Make sure we have discrete color levels. */
6908 glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_MAG_FILTER
, GL_NEAREST
);
6909 glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_WRAP_S
, GL_CLAMP_TO_EDGE
);
6910 /* Upload the palette */
6911 /* TODO: avoid unneeed uploads in the future by adding some SFLAG_PALETTE_DIRTY mechanism */
6912 glTexImage1D(GL_TEXTURE_1D
, 0, GL_RGBA
, 256, 0, GL_RGBA
, GL_UNSIGNED_BYTE
, table
);
6914 /* Switch back to unit 0 in which the 2D texture will be stored. */
6915 GL_EXTCALL(glActiveTextureARB(GL_TEXTURE0
));
6919 /* Context activation is done by the caller. */
6920 static GLuint
gen_yuv_shader(struct arbfp_blit_priv
*priv
, const struct wined3d_gl_info
*gl_info
,
6921 enum complex_fixup yuv_fixup
, GLenum textype
)
6924 struct wined3d_shader_buffer buffer
;
6925 char luminance_component
;
6929 if (!shader_buffer_init(&buffer
))
6931 ERR("Failed to initialize shader buffer.\n");
6936 GL_EXTCALL(glGenProgramsARB(1, &shader
));
6937 checkGLcall("GL_EXTCALL(glGenProgramsARB(1, &shader))");
6938 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
6939 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
6942 shader_buffer_free(&buffer
);
6946 /* The YUY2 and UYVY formats contain two pixels packed into a 32 bit macropixel,
6947 * giving effectively 16 bit per pixel. The color consists of a luminance(Y) and
6948 * two chroma(U and V) values. Each macropixel has two luminance values, one for
6949 * each single pixel it contains, and one U and one V value shared between both
6952 * The data is loaded into an A8L8 texture. With YUY2, the luminance component
6953 * contains the luminance and alpha the chroma. With UYVY it is vice versa. Thus
6954 * take the format into account when generating the read swizzles
6956 * Reading the Y value is straightforward - just sample the texture. The hardware
6957 * takes care of filtering in the horizontal and vertical direction.
6959 * Reading the U and V values is harder. We have to avoid filtering horizontally,
6960 * because that would mix the U and V values of one pixel or two adjacent pixels.
6961 * Thus floor the texture coordinate and add 0.5 to get an unfiltered read,
6962 * regardless of the filtering setting. Vertical filtering works automatically
6963 * though - the U and V values of two rows are mixed nicely.
6965 * Appart of avoiding filtering issues, the code has to know which value it just
6966 * read, and where it can find the other one. To determine this, it checks if
6967 * it sampled an even or odd pixel, and shifts the 2nd read accordingly.
6969 * Handling horizontal filtering of U and V values requires reading a 2nd pair
6970 * of pixels, extracting U and V and mixing them. This is not implemented yet.
6972 * An alternative implementation idea is to load the texture as A8R8G8B8 texture,
6973 * with width / 2. This way one read gives all 3 values, finding U and V is easy
6974 * in an unfiltered situation. Finding the luminance on the other hand requires
6975 * finding out if it is an odd or even pixel. The real drawback of this approach
6976 * is filtering. This would have to be emulated completely in the shader, reading
6977 * up two 2 packed pixels in up to 2 rows and interpolating both horizontally and
6978 * vertically. Beyond that it would require adjustments to the texture handling
6979 * code to deal with the width scaling
6981 shader_addline(&buffer
, "!!ARBfp1.0\n");
6982 shader_addline(&buffer
, "TEMP luminance;\n");
6983 shader_addline(&buffer
, "TEMP temp;\n");
6984 shader_addline(&buffer
, "TEMP chroma;\n");
6985 shader_addline(&buffer
, "TEMP texcrd;\n");
6986 shader_addline(&buffer
, "TEMP texcrd2;\n");
6987 shader_addline(&buffer
, "PARAM coef = {1.0, 0.5, 2.0, 0.25};\n");
6988 shader_addline(&buffer
, "PARAM yuv_coef = {1.403, 0.344, 0.714, 1.770};\n");
6989 shader_addline(&buffer
, "PARAM size = program.local[0];\n");
6993 case COMPLEX_FIXUP_UYVY
:
6994 case COMPLEX_FIXUP_YUY2
:
6995 if (!gen_planar_yuv_read(&buffer
, yuv_fixup
, textype
, &luminance_component
))
6997 shader_buffer_free(&buffer
);
7002 case COMPLEX_FIXUP_YV12
:
7003 if (!gen_yv12_read(&buffer
, textype
, &luminance_component
))
7005 shader_buffer_free(&buffer
);
7011 FIXME("Unsupported YUV fixup %#x\n", yuv_fixup
);
7012 shader_buffer_free(&buffer
);
7016 /* Calculate the final result. Formula is taken from
7017 * http://www.fourcc.org/fccyvrgb.php. Note that the chroma
7018 * ranges from -0.5 to 0.5
7020 shader_addline(&buffer
, "SUB chroma.xy, chroma, coef.y;\n");
7022 shader_addline(&buffer
, "MAD result.color.x, chroma.x, yuv_coef.x, luminance.%c;\n", luminance_component
);
7023 shader_addline(&buffer
, "MAD temp.x, -chroma.y, yuv_coef.y, luminance.%c;\n", luminance_component
);
7024 shader_addline(&buffer
, "MAD result.color.y, -chroma.x, yuv_coef.z, temp.x;\n");
7025 shader_addline(&buffer
, "MAD result.color.z, chroma.y, yuv_coef.w, luminance.%c;\n", luminance_component
);
7026 shader_addline(&buffer
, "END\n");
7029 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
7030 strlen(buffer
.buffer
), buffer
.buffer
));
7031 checkGLcall("glProgramStringARB()");
7033 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
7036 FIXME("Fragment program error at position %d: %s\n\n", pos
,
7037 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
7038 shader_arb_dump_program_source(buffer
.buffer
);
7044 GL_EXTCALL(glGetProgramivARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
7045 checkGLcall("glGetProgramivARB()");
7046 if (!native
) WARN("Program exceeds native resource limits.\n");
7049 shader_buffer_free(&buffer
);
7054 case COMPLEX_FIXUP_YUY2
:
7055 if (textype
== GL_TEXTURE_RECTANGLE_ARB
) priv
->yuy2_rect_shader
= shader
;
7056 else priv
->yuy2_2d_shader
= shader
;
7059 case COMPLEX_FIXUP_UYVY
:
7060 if (textype
== GL_TEXTURE_RECTANGLE_ARB
) priv
->uyvy_rect_shader
= shader
;
7061 else priv
->uyvy_2d_shader
= shader
;
7064 case COMPLEX_FIXUP_YV12
:
7065 if (textype
== GL_TEXTURE_RECTANGLE_ARB
) priv
->yv12_rect_shader
= shader
;
7066 else priv
->yv12_2d_shader
= shader
;
7069 ERR("Unsupported complex fixup: %d\n", yuv_fixup
);
7075 /* Context activation is done by the caller. */
7076 static HRESULT
arbfp_blit_set(void *blit_priv
, const struct wined3d_gl_info
*gl_info
, struct wined3d_surface
*surface
)
7079 float size
[4] = {(float) surface
->pow2Width
, (float) surface
->pow2Height
, 1.0f
, 1.0f
};
7080 struct arbfp_blit_priv
*priv
= blit_priv
;
7081 enum complex_fixup fixup
;
7082 GLenum textype
= surface
->texture_target
;
7084 if (!is_complex_fixup(surface
->resource
.format
->color_fixup
))
7087 dump_color_fixup_desc(surface
->resource
.format
->color_fixup
);
7088 /* Don't bother setting up a shader for unconverted formats */
7091 checkGLcall("glEnable(textype)");
7096 fixup
= get_complex_fixup(surface
->resource
.format
->color_fixup
);
7100 case COMPLEX_FIXUP_YUY2
:
7101 shader
= textype
== GL_TEXTURE_RECTANGLE_ARB
? priv
->yuy2_rect_shader
: priv
->yuy2_2d_shader
;
7104 case COMPLEX_FIXUP_UYVY
:
7105 shader
= textype
== GL_TEXTURE_RECTANGLE_ARB
? priv
->uyvy_rect_shader
: priv
->uyvy_2d_shader
;
7108 case COMPLEX_FIXUP_YV12
:
7109 shader
= textype
== GL_TEXTURE_RECTANGLE_ARB
? priv
->yv12_rect_shader
: priv
->yv12_2d_shader
;
7112 case COMPLEX_FIXUP_P8
:
7113 shader
= textype
== GL_TEXTURE_RECTANGLE_ARB
? priv
->p8_rect_shader
: priv
->p8_2d_shader
;
7114 if (!shader
) shader
= gen_p8_shader(priv
, gl_info
, textype
);
7116 upload_palette(surface
);
7120 FIXME("Unsupported complex fixup %#x, not setting a shader\n", fixup
);
7123 checkGLcall("glEnable(textype)");
7128 if (!shader
) shader
= gen_yuv_shader(priv
, gl_info
, fixup
, textype
);
7131 glEnable(GL_FRAGMENT_PROGRAM_ARB
);
7132 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
7133 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7134 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
7135 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, 0, size
));
7136 checkGLcall("glProgramLocalParameter4fvARB");
7142 /* Context activation is done by the caller. */
7143 static void arbfp_blit_unset(const struct wined3d_gl_info
*gl_info
)
7146 glDisable(GL_FRAGMENT_PROGRAM_ARB
);
7147 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
7148 glDisable(GL_TEXTURE_2D
);
7149 checkGLcall("glDisable(GL_TEXTURE_2D)");
7150 if (gl_info
->supported
[ARB_TEXTURE_CUBE_MAP
])
7152 glDisable(GL_TEXTURE_CUBE_MAP_ARB
);
7153 checkGLcall("glDisable(GL_TEXTURE_CUBE_MAP_ARB)");
7155 if (gl_info
->supported
[ARB_TEXTURE_RECTANGLE
])
7157 glDisable(GL_TEXTURE_RECTANGLE_ARB
);
7158 checkGLcall("glDisable(GL_TEXTURE_RECTANGLE_ARB)");
7163 static BOOL
arbfp_blit_supported(const struct wined3d_gl_info
*gl_info
, enum wined3d_blit_op blit_op
,
7164 const RECT
*src_rect
, DWORD src_usage
, WINED3DPOOL src_pool
, const struct wined3d_format
*src_format
,
7165 const RECT
*dst_rect
, DWORD dst_usage
, WINED3DPOOL dst_pool
, const struct wined3d_format
*dst_format
)
7167 enum complex_fixup src_fixup
;
7169 if (!gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
7172 if (blit_op
!= WINED3D_BLIT_OP_COLOR_BLIT
)
7174 TRACE("Unsupported blit_op=%d\n", blit_op
);
7178 src_fixup
= get_complex_fixup(src_format
->color_fixup
);
7179 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
7181 TRACE("Checking support for fixup:\n");
7182 dump_color_fixup_desc(src_format
->color_fixup
);
7185 if (!is_identity_fixup(dst_format
->color_fixup
))
7187 TRACE("Destination fixups are not supported\n");
7191 if (is_identity_fixup(src_format
->color_fixup
))
7197 /* We only support YUV conversions. */
7198 if (!is_complex_fixup(src_format
->color_fixup
))
7200 TRACE("[FAILED]\n");
7206 case COMPLEX_FIXUP_YUY2
:
7207 case COMPLEX_FIXUP_UYVY
:
7208 case COMPLEX_FIXUP_YV12
:
7209 case COMPLEX_FIXUP_P8
:
7214 FIXME("Unsupported YUV fixup %#x\n", src_fixup
);
7215 TRACE("[FAILED]\n");
7220 HRESULT
arbfp_blit_surface(struct wined3d_device
*device
, struct wined3d_surface
*src_surface
, const RECT
*src_rect
,
7221 struct wined3d_surface
*dst_surface
, const RECT
*dst_rect_in
, enum wined3d_blit_op blit_op
, DWORD Filter
)
7223 struct wined3d_context
*context
;
7224 RECT dst_rect
= *dst_rect_in
;
7226 /* Now load the surface */
7227 surface_internal_preload(src_surface
, SRGB_RGB
);
7229 /* Activate the destination context, set it up for blitting */
7230 context
= context_acquire(device
, dst_surface
);
7231 context_apply_blit_state(context
, device
);
7233 if (!surface_is_offscreen(dst_surface
))
7234 surface_translate_drawable_coords(dst_surface
, context
->win_handle
, &dst_rect
);
7236 arbfp_blit_set(device
->blit_priv
, context
->gl_info
, src_surface
);
7240 /* Draw a textured quad */
7241 draw_textured_quad(src_surface
, src_rect
, &dst_rect
, Filter
);
7245 /* Leave the opengl state valid for blitting */
7246 arbfp_blit_unset(context
->gl_info
);
7248 if (wined3d_settings
.strict_draw_ordering
7249 || (dst_surface
->container
.type
== WINED3D_CONTAINER_SWAPCHAIN
7250 && (dst_surface
->container
.u
.swapchain
->front_buffer
== dst_surface
)))
7251 wglFlush(); /* Flush to ensure ordering across contexts. */
7253 context_release(context
);
7255 surface_modify_location(dst_surface
, SFLAG_INDRAWABLE
, TRUE
);
7259 /* Do not call while under the GL lock. */
7260 static HRESULT
arbfp_blit_color_fill(struct wined3d_device
*device
, struct wined3d_surface
*dst_surface
,
7261 const RECT
*dst_rect
, const WINED3DCOLORVALUE
*color
)
7263 FIXME("Color filling not implemented by arbfp_blit\n");
7264 return WINED3DERR_INVALIDCALL
;
7267 /* Do not call while under the GL lock. */
7268 static HRESULT
arbfp_blit_depth_fill(struct wined3d_device
*device
,
7269 struct wined3d_surface
*surface
, const RECT
*rect
, float depth
)
7271 FIXME("Depth filling not implemented by arbfp_blit.\n");
7272 return WINED3DERR_INVALIDCALL
;
7275 const struct blit_shader arbfp_blit
= {
7280 arbfp_blit_supported
,
7281 arbfp_blit_color_fill
,
7282 arbfp_blit_depth_fill
,