2 * GLSL pixel and vertex shader implementation
4 * Copyright 2006 Jason Green
5 * Copyright 2006-2007 Henri Verbeet
6 * Copyright 2007-2008 Stefan Dösinger for CodeWeavers
8 * This library is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
13 * This library is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with this library; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
24 * D3D shader asm has swizzles on source parameters, and write masks for
25 * destination parameters. GLSL uses swizzles for both. The result of this is
26 * that for example "mov dst.xw, src.zyxw" becomes "dst.xw = src.zw" in GLSL.
27 * Ie, to generate a proper GLSL source swizzle, we need to take the D3D write
28 * mask for the destination parameter into account.
33 #include "wined3d_private.h"
35 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader
);
36 WINE_DECLARE_DEBUG_CHANNEL(d3d_constants
);
37 WINE_DECLARE_DEBUG_CHANNEL(d3d_caps
);
39 #define GLINFO_LOCATION (*gl_info)
54 } glsl_sample_function_t
;
56 /** Prints the GLSL info log which will contain error messages if they exist */
57 void print_glsl_info_log(WineD3D_GL_Info
*gl_info
, GLhandleARB obj
) {
59 int infologLength
= 0;
64 const char *spam
[] = {
65 "Vertex shader was successfully compiled to run on hardware.\n", /* fglrx */
66 "Fragment shader was successfully compiled to run on hardware.\n", /* fglrx */
67 "Fragment shader(s) linked, vertex shader(s) linked.", /* fglrx, no \n */
68 "Vertex shader(s) linked, no fragment shader(s) defined.", /* fglrx, no \n */
69 "Fragment shader was successfully compiled to run on hardware.\nWARNING: 0:1: extension 'GL_ARB_draw_buffers' is not supported",
70 "Fragment shader(s) linked, no vertex shader(s) defined." /* fglrx, no \n */
73 GL_EXTCALL(glGetObjectParameterivARB(obj
,
74 GL_OBJECT_INFO_LOG_LENGTH_ARB
,
77 /* A size of 1 is just a null-terminated string, so the log should be bigger than
78 * that if there are errors. */
79 if (infologLength
> 1)
81 /* Fglrx doesn't terminate the string properly, but it tells us the proper length.
82 * So use HEAP_ZERO_MEMORY to avoid uninitialized bytes
84 infoLog
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, infologLength
);
85 GL_EXTCALL(glGetInfoLogARB(obj
, infologLength
, NULL
, infoLog
));
88 for(i
= 0; i
< sizeof(spam
) / sizeof(spam
[0]); i
++) {
89 if(strcmp(infoLog
, spam
[i
]) == 0) {
95 TRACE("Spam received from GLSL shader #%u: %s\n", obj
, debugstr_a(infoLog
));
97 FIXME("Error received from GLSL shader #%u: %s\n", obj
, debugstr_a(infoLog
));
99 HeapFree(GetProcessHeap(), 0, infoLog
);
104 * Loads (pixel shader) samplers
106 static void shader_glsl_load_psamplers(
107 WineD3D_GL_Info
*gl_info
,
108 IWineD3DStateBlock
* iface
,
109 GLhandleARB programId
) {
111 IWineD3DStateBlockImpl
* stateBlock
= (IWineD3DStateBlockImpl
*) iface
;
112 GLhandleARB name_loc
;
114 char sampler_name
[20];
116 for (i
= 0; i
< MAX_FRAGMENT_SAMPLERS
; ++i
) {
117 snprintf(sampler_name
, sizeof(sampler_name
), "Psampler%d", i
);
118 name_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, sampler_name
));
119 if (name_loc
!= -1) {
120 int mapped_unit
= stateBlock
->wineD3DDevice
->texUnitMap
[i
];
121 if (mapped_unit
!= -1 && mapped_unit
< GL_LIMITS(fragment_samplers
)) {
122 TRACE("Loading %s for texture %d\n", sampler_name
, mapped_unit
);
123 GL_EXTCALL(glUniform1iARB(name_loc
, mapped_unit
));
124 checkGLcall("glUniform1iARB");
126 ERR("Trying to load sampler %s on unsupported unit %d\n", sampler_name
, mapped_unit
);
132 static void shader_glsl_load_vsamplers(WineD3D_GL_Info
*gl_info
, IWineD3DStateBlock
* iface
, GLhandleARB programId
) {
133 IWineD3DStateBlockImpl
* stateBlock
= (IWineD3DStateBlockImpl
*) iface
;
134 GLhandleARB name_loc
;
135 char sampler_name
[20];
138 for (i
= 0; i
< MAX_VERTEX_SAMPLERS
; ++i
) {
139 snprintf(sampler_name
, sizeof(sampler_name
), "Vsampler%d", i
);
140 name_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, sampler_name
));
141 if (name_loc
!= -1) {
142 int mapped_unit
= stateBlock
->wineD3DDevice
->texUnitMap
[MAX_FRAGMENT_SAMPLERS
+ i
];
143 if (mapped_unit
!= -1 && mapped_unit
< GL_LIMITS(combined_samplers
)) {
144 TRACE("Loading %s for texture %d\n", sampler_name
, mapped_unit
);
145 GL_EXTCALL(glUniform1iARB(name_loc
, mapped_unit
));
146 checkGLcall("glUniform1iARB");
148 ERR("Trying to load sampler %s on unsupported unit %d\n", sampler_name
, mapped_unit
);
155 * Loads floating point constants (aka uniforms) into the currently set GLSL program.
156 * When constant_list == NULL, it will load all the constants.
158 static void shader_glsl_load_constantsF(IWineD3DBaseShaderImpl
* This
, WineD3D_GL_Info
*gl_info
,
159 unsigned int max_constants
, float* constants
, GLhandleARB
*constant_locations
,
160 struct list
*constant_list
) {
161 constants_entry
*constant
;
162 local_constant
* lconst
;
167 if (TRACE_ON(d3d_shader
)) {
168 LIST_FOR_EACH_ENTRY(constant
, constant_list
, constants_entry
, entry
) {
173 tmp_loc
= constant_locations
[i
];
175 TRACE_(d3d_constants
)("Loading constants %i: %f, %f, %f, %f\n", i
,
176 constants
[i
* 4 + 0], constants
[i
* 4 + 1],
177 constants
[i
* 4 + 2], constants
[i
* 4 + 3]);
183 /* 1.X pshaders have the constants clamped to [-1;1] implicitly. */
184 if(WINED3DSHADER_VERSION_MAJOR(This
->baseShader
.hex_version
) == 1 &&
185 shader_is_pshader_version(This
->baseShader
.hex_version
)) {
188 LIST_FOR_EACH_ENTRY(constant
, constant_list
, constants_entry
, entry
) {
193 tmp_loc
= constant_locations
[i
];
195 /* We found this uniform name in the program - go ahead and send the data */
197 if(constants
[k
+ 0] < -1.0) lcl_const
[0] = -1.0;
198 else if(constants
[k
+ 0] > 1.0) lcl_const
[0] = 1.0;
199 else lcl_const
[0] = constants
[k
+ 0];
200 if(constants
[k
+ 1] < -1.0) lcl_const
[1] = -1.0;
201 else if(constants
[k
+ 1] > 1.0) lcl_const
[1] = 1.0;
202 else lcl_const
[1] = constants
[k
+ 1];
203 if(constants
[k
+ 2] < -1.0) lcl_const
[2] = -1.0;
204 else if(constants
[k
+ 2] > 1.0) lcl_const
[2] = 1.0;
205 else lcl_const
[2] = constants
[k
+ 2];
206 if(constants
[k
+ 3] < -1.0) lcl_const
[3] = -1.0;
207 else if(constants
[k
+ 3] > 1.0) lcl_const
[3] = 1.0;
208 else lcl_const
[3] = constants
[k
+ 3];
210 GL_EXTCALL(glUniform4fvARB(tmp_loc
, 1, lcl_const
));
215 LIST_FOR_EACH_ENTRY(constant
, constant_list
, constants_entry
, entry
) {
220 tmp_loc
= constant_locations
[i
];
222 /* We found this uniform name in the program - go ahead and send the data */
223 GL_EXTCALL(glUniform4fvARB(tmp_loc
, 1, constants
+ (i
* 4)));
228 checkGLcall("glUniform4fvARB()");
230 if(!This
->baseShader
.load_local_constsF
) {
231 TRACE("No need to load local float constants for this shader\n");
235 /* Load immediate constants */
236 if (TRACE_ON(d3d_shader
)) {
237 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
238 tmp_loc
= constant_locations
[lconst
->idx
];
240 GLfloat
* values
= (GLfloat
*)lconst
->value
;
241 TRACE_(d3d_constants
)("Loading local constants %i: %f, %f, %f, %f\n", lconst
->idx
,
242 values
[0], values
[1], values
[2], values
[3]);
246 /* Immediate constants are clamped to [-1;1] at shader creation time if needed */
247 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
248 tmp_loc
= constant_locations
[lconst
->idx
];
250 /* We found this uniform name in the program - go ahead and send the data */
251 GL_EXTCALL(glUniform4fvARB(tmp_loc
, 1, (GLfloat
*)lconst
->value
));
254 checkGLcall("glUniform4fvARB()");
258 * Loads integer constants (aka uniforms) into the currently set GLSL program.
259 * When @constants_set == NULL, it will load all the constants.
261 static void shader_glsl_load_constantsI(
262 IWineD3DBaseShaderImpl
* This
,
263 WineD3D_GL_Info
*gl_info
,
264 GLhandleARB programId
,
265 GLhandleARB locations
[MAX_CONST_I
],
266 unsigned max_constants
,
268 BOOL
* constants_set
) {
273 for (i
=0; i
<max_constants
; ++i
) {
274 if (NULL
== constants_set
|| constants_set
[i
]) {
276 TRACE_(d3d_constants
)("Loading constants %i: %i, %i, %i, %i\n",
277 i
, constants
[i
*4], constants
[i
*4+1], constants
[i
*4+2], constants
[i
*4+3]);
279 /* We found this uniform name in the program - go ahead and send the data */
280 GL_EXTCALL(glUniform4ivARB(locations
[i
], 1, &constants
[i
*4]));
281 checkGLcall("glUniform4ivARB");
285 /* Load immediate constants */
286 ptr
= list_head(&This
->baseShader
.constantsI
);
288 local_constant
* lconst
= LIST_ENTRY(ptr
, struct local_constant
, entry
);
289 unsigned int idx
= lconst
->idx
;
290 GLint
* values
= (GLint
*) lconst
->value
;
292 TRACE_(d3d_constants
)("Loading local constants %i: %i, %i, %i, %i\n", idx
,
293 values
[0], values
[1], values
[2], values
[3]);
295 /* We found this uniform name in the program - go ahead and send the data */
296 GL_EXTCALL(glUniform4ivARB(locations
[idx
], 1, values
));
297 checkGLcall("glUniform4ivARB");
298 ptr
= list_next(&This
->baseShader
.constantsI
, ptr
);
303 * Loads boolean constants (aka uniforms) into the currently set GLSL program.
304 * When @constants_set == NULL, it will load all the constants.
306 static void shader_glsl_load_constantsB(
307 IWineD3DBaseShaderImpl
* This
,
308 WineD3D_GL_Info
*gl_info
,
309 GLhandleARB programId
,
310 unsigned max_constants
,
312 BOOL
* constants_set
) {
317 char is_pshader
= shader_is_pshader_version(This
->baseShader
.hex_version
);
318 const char* prefix
= is_pshader
? "PB":"VB";
321 for (i
=0; i
<max_constants
; ++i
) {
322 if (NULL
== constants_set
|| constants_set
[i
]) {
324 TRACE_(d3d_constants
)("Loading constants %i: %i;\n", i
, constants
[i
]);
326 /* TODO: Benchmark and see if it would be beneficial to store the
327 * locations of the constants to avoid looking up each time */
328 snprintf(tmp_name
, sizeof(tmp_name
), "%s[%i]", prefix
, i
);
329 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, tmp_name
));
331 /* We found this uniform name in the program - go ahead and send the data */
332 GL_EXTCALL(glUniform1ivARB(tmp_loc
, 1, &constants
[i
]));
333 checkGLcall("glUniform1ivARB");
338 /* Load immediate constants */
339 ptr
= list_head(&This
->baseShader
.constantsB
);
341 local_constant
* lconst
= LIST_ENTRY(ptr
, struct local_constant
, entry
);
342 unsigned int idx
= lconst
->idx
;
343 GLint
* values
= (GLint
*) lconst
->value
;
345 TRACE_(d3d_constants
)("Loading local constants %i: %i\n", idx
, values
[0]);
347 snprintf(tmp_name
, sizeof(tmp_name
), "%s[%i]", prefix
, idx
);
348 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, tmp_name
));
350 /* We found this uniform name in the program - go ahead and send the data */
351 GL_EXTCALL(glUniform1ivARB(tmp_loc
, 1, values
));
352 checkGLcall("glUniform1ivARB");
354 ptr
= list_next(&This
->baseShader
.constantsB
, ptr
);
361 * Loads the app-supplied constants into the currently set GLSL program.
363 void shader_glsl_load_constants(
364 IWineD3DDevice
* device
,
366 char useVertexShader
) {
368 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) device
;
369 IWineD3DStateBlockImpl
* stateBlock
= deviceImpl
->stateBlock
;
370 WineD3D_GL_Info
*gl_info
= &deviceImpl
->adapter
->gl_info
;
372 GLhandleARB
*constant_locations
;
373 struct list
*constant_list
;
374 GLhandleARB programId
;
375 struct glsl_shader_prog_link
*prog
= stateBlock
->glsl_program
;
379 /* No GLSL program set - nothing to do. */
382 programId
= prog
->programId
;
384 if (useVertexShader
) {
385 IWineD3DBaseShaderImpl
* vshader
= (IWineD3DBaseShaderImpl
*) stateBlock
->vertexShader
;
387 constant_locations
= prog
->vuniformF_locations
;
388 constant_list
= &stateBlock
->set_vconstantsF
;
390 /* Load DirectX 9 float constants/uniforms for vertex shader */
391 shader_glsl_load_constantsF(vshader
, gl_info
, GL_LIMITS(vshader_constantsF
),
392 stateBlock
->vertexShaderConstantF
, constant_locations
, constant_list
);
394 /* Load DirectX 9 integer constants/uniforms for vertex shader */
395 shader_glsl_load_constantsI(vshader
, gl_info
, programId
,
396 prog
->vuniformI_locations
, MAX_CONST_I
,
397 stateBlock
->vertexShaderConstantI
,
398 stateBlock
->changed
.vertexShaderConstantsI
);
400 /* Load DirectX 9 boolean constants/uniforms for vertex shader */
401 shader_glsl_load_constantsB(vshader
, gl_info
, programId
, MAX_CONST_B
,
402 stateBlock
->vertexShaderConstantB
,
403 stateBlock
->changed
.vertexShaderConstantsB
);
405 /* Upload the position fixup params */
406 GL_EXTCALL(glUniform4fvARB(prog
->posFixup_location
, 1, &deviceImpl
->posFixup
[0]));
407 checkGLcall("glUniform4fvARB");
410 if (usePixelShader
) {
412 IWineD3DBaseShaderImpl
* pshader
= (IWineD3DBaseShaderImpl
*) stateBlock
->pixelShader
;
414 constant_locations
= prog
->puniformF_locations
;
415 constant_list
= &stateBlock
->set_pconstantsF
;
417 /* Load DirectX 9 float constants/uniforms for pixel shader */
418 shader_glsl_load_constantsF(pshader
, gl_info
, GL_LIMITS(pshader_constantsF
),
419 stateBlock
->pixelShaderConstantF
, constant_locations
, constant_list
);
421 /* Load DirectX 9 integer constants/uniforms for pixel shader */
422 shader_glsl_load_constantsI(pshader
, gl_info
, programId
,
423 prog
->puniformI_locations
, MAX_CONST_I
,
424 stateBlock
->pixelShaderConstantI
,
425 stateBlock
->changed
.pixelShaderConstantsI
);
427 /* Load DirectX 9 boolean constants/uniforms for pixel shader */
428 shader_glsl_load_constantsB(pshader
, gl_info
, programId
, MAX_CONST_B
,
429 stateBlock
->pixelShaderConstantB
,
430 stateBlock
->changed
.pixelShaderConstantsB
);
432 /* Upload the environment bump map matrix if needed. The needsbumpmat member specifies the texture stage to load the matrix from.
433 * It can't be 0 for a valid texbem instruction.
435 for(i
= 0; i
< ((IWineD3DPixelShaderImpl
*) pshader
)->numbumpenvmatconsts
; i
++) {
436 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) pshader
;
437 int stage
= ps
->luminanceconst
[i
].texunit
;
439 float *data
= (float *) &stateBlock
->textureState
[(int) ps
->bumpenvmatconst
[i
].texunit
][WINED3DTSS_BUMPENVMAT00
];
440 GL_EXTCALL(glUniformMatrix2fvARB(prog
->bumpenvmat_location
[i
], 1, 0, data
));
441 checkGLcall("glUniformMatrix2fvARB");
443 /* texbeml needs the luminance scale and offset too. If texbeml is used, needsbumpmat
444 * is set too, so we can check that in the needsbumpmat check
446 if(ps
->baseShader
.reg_maps
.luminanceparams
[stage
]) {
447 GLfloat
*scale
= (GLfloat
*) &stateBlock
->textureState
[stage
][WINED3DTSS_BUMPENVLSCALE
];
448 GLfloat
*offset
= (GLfloat
*) &stateBlock
->textureState
[stage
][WINED3DTSS_BUMPENVLOFFSET
];
450 GL_EXTCALL(glUniform1fvARB(prog
->luminancescale_location
[i
], 1, scale
));
451 checkGLcall("glUniform1fvARB");
452 GL_EXTCALL(glUniform1fvARB(prog
->luminanceoffset_location
[i
], 1, offset
));
453 checkGLcall("glUniform1fvARB");
457 if(((IWineD3DPixelShaderImpl
*) pshader
)->srgb_enabled
&&
458 !((IWineD3DPixelShaderImpl
*) pshader
)->srgb_mode_hardcoded
) {
462 if(stateBlock
->renderState
[WINED3DRS_SRGBWRITEENABLE
]) {
463 comparison
[0] = srgb_cmp
; comparison
[1] = srgb_cmp
;
464 comparison
[2] = srgb_cmp
; comparison
[3] = srgb_cmp
;
466 mul_low
[0] = srgb_mul_low
; mul_low
[1] = srgb_mul_low
;
467 mul_low
[2] = srgb_mul_low
; mul_low
[3] = srgb_mul_low
;
469 comparison
[0] = 1.0 / 0.0; comparison
[1] = 1.0 / 0.0;
470 comparison
[2] = 1.0 / 0.0; comparison
[3] = 1.0 / 0.0;
472 mul_low
[0] = 1.0; mul_low
[1] = 1.0;
473 mul_low
[2] = 1.0; mul_low
[3] = 1.0;
476 GL_EXTCALL(glUniform4fvARB(prog
->srgb_comparison_location
, 1, comparison
));
477 GL_EXTCALL(glUniform4fvARB(prog
->srgb_mul_low_location
, 1, mul_low
));
479 if(((IWineD3DPixelShaderImpl
*) pshader
)->vpos_uniform
) {
480 float correction_params
[4];
481 if(deviceImpl
->render_offscreen
) {
482 correction_params
[0] = 0.0;
483 correction_params
[1] = 1.0;
485 /* position is window relative, not viewport relative */
486 correction_params
[0] = ((IWineD3DSurfaceImpl
*) deviceImpl
->render_targets
[0])->currentDesc
.Height
;
487 correction_params
[1] = -1.0;
489 GL_EXTCALL(glUniform4fvARB(prog
->ycorrection_location
, 1, correction_params
));
494 /** Generate the variable & register declarations for the GLSL output target */
495 void shader_generate_glsl_declarations(
496 IWineD3DBaseShader
*iface
,
497 shader_reg_maps
* reg_maps
,
498 SHADER_BUFFER
* buffer
,
499 WineD3D_GL_Info
* gl_info
) {
501 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*) iface
;
502 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
504 unsigned int extra_constants_needed
= 0;
505 local_constant
* lconst
;
507 /* There are some minor differences between pixel and vertex shaders */
508 char pshader
= shader_is_pshader_version(This
->baseShader
.hex_version
);
509 char prefix
= pshader
? 'P' : 'V';
511 /* Prototype the subroutines */
512 for (i
= 0; i
< This
->baseShader
.limits
.label
; i
++) {
513 if (reg_maps
->labels
[i
])
514 shader_addline(buffer
, "void subroutine%u();\n", i
);
517 /* Declare the constants (aka uniforms) */
518 if (This
->baseShader
.limits
.constant_float
> 0) {
519 unsigned max_constantsF
= min(This
->baseShader
.limits
.constant_float
,
520 (pshader
? GL_LIMITS(pshader_constantsF
) : GL_LIMITS(vshader_constantsF
)));
521 shader_addline(buffer
, "uniform vec4 %cC[%u];\n", prefix
, max_constantsF
);
524 if (This
->baseShader
.limits
.constant_int
> 0)
525 shader_addline(buffer
, "uniform ivec4 %cI[%u];\n", prefix
, This
->baseShader
.limits
.constant_int
);
527 if (This
->baseShader
.limits
.constant_bool
> 0)
528 shader_addline(buffer
, "uniform bool %cB[%u];\n", prefix
, This
->baseShader
.limits
.constant_bool
);
531 shader_addline(buffer
, "uniform vec4 posFixup;\n");
532 /* Predeclaration; This function is added at link time based on the pixel shader.
533 * VS 3.0 shaders have an array OUT[] the shader writes to, earlier versions don't have
534 * that. We know the input to the reorder function at vertex shader compile time, so
535 * we can deal with that. The reorder function for a 1.x and 2.x vertex shader can just
536 * read gl_FrontColor. The output depends on the pixel shader. The reorder function for a
537 * 1.x and 2.x pshader or for fixed function will write gl_FrontColor, and for a 3.0 shader
538 * it will write to the varying array. Here we depend on the shader optimizer on sorting that
539 * out. The nvidia driver only does that if the parameter is inout instead of out, hence the
542 if(This
->baseShader
.hex_version
>= WINED3DVS_VERSION(3, 0)) {
543 shader_addline(buffer
, "void order_ps_input(in vec4[%u]);\n", MAX_REG_OUTPUT
);
545 shader_addline(buffer
, "void order_ps_input();\n");
548 IWineD3DPixelShaderImpl
*ps_impl
= (IWineD3DPixelShaderImpl
*) This
;
550 ps_impl
->numbumpenvmatconsts
= 0;
551 for(i
= 0; i
< (sizeof(reg_maps
->bumpmat
) / sizeof(reg_maps
->bumpmat
[0])); i
++) {
552 if(!reg_maps
->bumpmat
[i
]) {
556 ps_impl
->bumpenvmatconst
[(int) ps_impl
->numbumpenvmatconsts
].texunit
= i
;
557 shader_addline(buffer
, "uniform mat2 bumpenvmat%d;\n", i
);
559 if(reg_maps
->luminanceparams
) {
560 ps_impl
->luminanceconst
[(int) ps_impl
->numbumpenvmatconsts
].texunit
= i
;
561 shader_addline(buffer
, "uniform float luminancescale%d;\n", i
);
562 shader_addline(buffer
, "uniform float luminanceoffset%d;\n", i
);
563 extra_constants_needed
++;
565 ps_impl
->luminanceconst
[(int) ps_impl
->numbumpenvmatconsts
].texunit
= -1;
568 extra_constants_needed
++;
569 ps_impl
->numbumpenvmatconsts
++;
572 if(device
->stateBlock
->renderState
[WINED3DRS_SRGBWRITEENABLE
]) {
573 ps_impl
->srgb_enabled
= 1;
574 if(This
->baseShader
.limits
.constant_float
+ extra_constants_needed
+ 1 < GL_LIMITS(pshader_constantsF
)) {
575 shader_addline(buffer
, "uniform vec4 srgb_mul_low;\n");
576 shader_addline(buffer
, "uniform vec4 srgb_comparison;\n");
577 ps_impl
->srgb_mode_hardcoded
= 0;
578 extra_constants_needed
++;
580 ps_impl
->srgb_mode_hardcoded
= 1;
581 shader_addline(buffer
, "const vec4 srgb_mul_low = vec4(%f, %f, %f, %f);\n",
582 srgb_mul_low
, srgb_mul_low
, srgb_mul_low
, srgb_mul_low
);
583 shader_addline(buffer
, "const vec4 srgb_comparison = vec4(%f, %f, %f, %f);\n",
584 srgb_cmp
, srgb_cmp
, srgb_cmp
, srgb_cmp
);
587 IWineD3DPixelShaderImpl
*ps_impl
= (IWineD3DPixelShaderImpl
*) This
;
589 /* Do not write any srgb fixup into the shader to save shader size and processing time.
590 * As a consequence, we can't toggle srgb write on without recompilation
592 ps_impl
->srgb_enabled
= 0;
593 ps_impl
->srgb_mode_hardcoded
= 1;
595 if(reg_maps
->vpos
|| reg_maps
->usesdsy
) {
596 if(This
->baseShader
.limits
.constant_float
+ extra_constants_needed
+ 1 < GL_LIMITS(pshader_constantsF
)) {
597 shader_addline(buffer
, "uniform vec4 ycorrection;\n");
598 ((IWineD3DPixelShaderImpl
*) This
)->vpos_uniform
= 1;
599 extra_constants_needed
++;
601 /* This happens because we do not have proper tracking of the constant registers that are
602 * actually used, only the max limit of the shader version
604 FIXME("Cannot find a free uniform for vpos correction params\n");
605 shader_addline(buffer
, "const vec4 ycorrection = vec4(%f, %f, 0.0, 0.0);\n",
606 device
->render_offscreen
? 0.0 : ((IWineD3DSurfaceImpl
*) device
->render_targets
[0])->currentDesc
.Height
,
607 device
->render_offscreen
? 1.0 : -1.0);
609 shader_addline(buffer
, "vec4 vpos;\n");
613 /* Declare texture samplers */
614 for (i
= 0; i
< This
->baseShader
.limits
.sampler
; i
++) {
615 if (reg_maps
->samplers
[i
]) {
617 DWORD stype
= reg_maps
->samplers
[i
] & WINED3DSP_TEXTURETYPE_MASK
;
621 shader_addline(buffer
, "uniform sampler1D %csampler%u;\n", prefix
, i
);
624 if(device
->stateBlock
->textures
[i
] &&
625 IWineD3DBaseTexture_GetTextureDimensions(device
->stateBlock
->textures
[i
]) == GL_TEXTURE_RECTANGLE_ARB
) {
626 shader_addline(buffer
, "uniform sampler2DRect %csampler%u;\n", prefix
, i
);
628 shader_addline(buffer
, "uniform sampler2D %csampler%u;\n", prefix
, i
);
631 case WINED3DSTT_CUBE
:
632 shader_addline(buffer
, "uniform samplerCube %csampler%u;\n", prefix
, i
);
634 case WINED3DSTT_VOLUME
:
635 shader_addline(buffer
, "uniform sampler3D %csampler%u;\n", prefix
, i
);
638 shader_addline(buffer
, "uniform unsupported_sampler %csampler%u;\n", prefix
, i
);
639 FIXME("Unrecognized sampler type: %#x\n", stype
);
645 /* Declare address variables */
646 for (i
= 0; i
< This
->baseShader
.limits
.address
; i
++) {
647 if (reg_maps
->address
[i
])
648 shader_addline(buffer
, "ivec4 A%d;\n", i
);
651 /* Declare texture coordinate temporaries and initialize them */
652 for (i
= 0; i
< This
->baseShader
.limits
.texcoord
; i
++) {
653 if (reg_maps
->texcoord
[i
])
654 shader_addline(buffer
, "vec4 T%u = gl_TexCoord[%u];\n", i
, i
);
657 /* Declare input register varyings. Only pixel shader, vertex shaders have that declared in the
658 * helper function shader that is linked in at link time
660 if(pshader
&& This
->baseShader
.hex_version
>= WINED3DPS_VERSION(3, 0)) {
662 shader_addline(buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
664 /* TODO: Write a replacement shader for the fixed function vertex pipeline, so this isn't needed.
665 * For fixed function vertex processing + 3.0 pixel shader we need a separate function in the
666 * pixel shader that reads the fixed function color into the packed input registers.
668 shader_addline(buffer
, "vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
672 /* Declare output register temporaries */
673 if(This
->baseShader
.limits
.packed_output
) {
674 shader_addline(buffer
, "vec4 OUT[%u];\n", This
->baseShader
.limits
.packed_output
);
677 /* Declare temporary variables */
678 for(i
= 0; i
< This
->baseShader
.limits
.temporary
; i
++) {
679 if (reg_maps
->temporary
[i
])
680 shader_addline(buffer
, "vec4 R%u;\n", i
);
683 /* Declare attributes */
684 for (i
= 0; i
< This
->baseShader
.limits
.attributes
; i
++) {
685 if (reg_maps
->attributes
[i
])
686 shader_addline(buffer
, "attribute vec4 attrib%i;\n", i
);
689 /* Declare loop registers aLx */
690 for (i
= 0; i
< reg_maps
->loop_depth
; i
++) {
691 shader_addline(buffer
, "int aL%u;\n", i
);
692 shader_addline(buffer
, "int tmpInt%u;\n", i
);
695 /* Temporary variables for matrix operations */
696 shader_addline(buffer
, "vec4 tmp0;\n");
697 shader_addline(buffer
, "vec4 tmp1;\n");
699 /* Hardcodable local constants */
700 if(!This
->baseShader
.load_local_constsF
) {
701 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
702 float *value
= (float *) lconst
->value
;
703 shader_addline(buffer
, "const vec4 LC%u = vec4(%f, %f, %f, %f);\n", lconst
->idx
,
704 value
[0], value
[1], value
[2], value
[3]);
708 /* Start the main program */
709 shader_addline(buffer
, "void main() {\n");
710 if(pshader
&& reg_maps
->vpos
) {
711 /* DirectX apps expect integer values, while OpenGL drivers add approximately 0.5. This causes
712 * off-by-one problems as spotted by the vPos d3d9 visual test. Unfortunately the ATI cards do
713 * not add exactly 0.5, but rather something like 0.49999999 or 0.50000001, which still causes
714 * precision troubles when we just substract 0.5.
716 * To deal with that just floor() the position. This will eliminate the fraction on all cards.
718 * TODO: Test how that behaves with multisampling once we can enable multisampling in winex11.
720 * An advantage of floor is that it works even if the driver doesn't add 1/2. It is somewhat
721 * questionable if 1.5, 2.5, ... are the proper values to return in gl_FragCoord, even though
722 * coordinates specify the pixel centers instead of the pixel corners. This code will behave
723 * correctly on drivers that returns integer values.
725 shader_addline(buffer
, "vpos = floor(vec4(0, ycorrection[0], 0, 0) + gl_FragCoord * vec4(1, ycorrection[1], 1, 1));\n");
729 /*****************************************************************************
730 * Functions to generate GLSL strings from DirectX Shader bytecode begin here.
732 * For more information, see http://wiki.winehq.org/DirectX-Shaders
733 ****************************************************************************/
736 static void shader_glsl_add_src_param(SHADER_OPCODE_ARG
* arg
, const DWORD param
,
737 const DWORD addr_token
, DWORD mask
, glsl_src_param_t
*src_param
);
739 /** Used for opcode modifiers - They multiply the result by the specified amount */
740 static const char * const shift_glsl_tab
[] = {
742 "2.0 * ", /* 1 (x2) */
743 "4.0 * ", /* 2 (x4) */
744 "8.0 * ", /* 3 (x8) */
745 "16.0 * ", /* 4 (x16) */
746 "32.0 * ", /* 5 (x32) */
753 "0.0625 * ", /* 12 (d16) */
754 "0.125 * ", /* 13 (d8) */
755 "0.25 * ", /* 14 (d4) */
756 "0.5 * " /* 15 (d2) */
759 /* Generate a GLSL parameter that does the input modifier computation and return the input register/mask to use */
760 static void shader_glsl_gen_modifier (
763 const char *in_regswizzle
,
768 if (instr
== WINED3DSIO_TEXKILL
)
771 switch (instr
& WINED3DSP_SRCMOD_MASK
) {
772 case WINED3DSPSM_DZ
: /* Need to handle this in the instructions itself (texld & texcrd). */
774 case WINED3DSPSM_NONE
:
775 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
777 case WINED3DSPSM_NEG
:
778 sprintf(out_str
, "-%s%s", in_reg
, in_regswizzle
);
780 case WINED3DSPSM_NOT
:
781 sprintf(out_str
, "!%s%s", in_reg
, in_regswizzle
);
783 case WINED3DSPSM_BIAS
:
784 sprintf(out_str
, "(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
786 case WINED3DSPSM_BIASNEG
:
787 sprintf(out_str
, "-(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
789 case WINED3DSPSM_SIGN
:
790 sprintf(out_str
, "(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
792 case WINED3DSPSM_SIGNNEG
:
793 sprintf(out_str
, "-(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
795 case WINED3DSPSM_COMP
:
796 sprintf(out_str
, "(1.0 - %s%s)", in_reg
, in_regswizzle
);
799 sprintf(out_str
, "(2.0 * %s%s)", in_reg
, in_regswizzle
);
801 case WINED3DSPSM_X2NEG
:
802 sprintf(out_str
, "-(2.0 * %s%s)", in_reg
, in_regswizzle
);
804 case WINED3DSPSM_ABS
:
805 sprintf(out_str
, "abs(%s%s)", in_reg
, in_regswizzle
);
807 case WINED3DSPSM_ABSNEG
:
808 sprintf(out_str
, "-abs(%s%s)", in_reg
, in_regswizzle
);
811 FIXME("Unhandled modifier %u\n", (instr
& WINED3DSP_SRCMOD_MASK
));
812 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
816 /** Writes the GLSL variable name that corresponds to the register that the
817 * DX opcode parameter is trying to access */
818 static void shader_glsl_get_register_name(
820 const DWORD addr_token
,
823 SHADER_OPCODE_ARG
* arg
) {
825 /* oPos, oFog and oPts in D3D */
826 static const char * const hwrastout_reg_names
[] = { "gl_Position", "gl_FogFragCoord", "gl_PointSize" };
828 DWORD reg
= param
& WINED3DSP_REGNUM_MASK
;
829 DWORD regtype
= shader_get_regtype(param
);
830 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
831 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
832 WineD3D_GL_Info
* gl_info
= &deviceImpl
->adapter
->gl_info
;
834 char pshader
= shader_is_pshader_version(This
->baseShader
.hex_version
);
840 case WINED3DSPR_TEMP
:
841 sprintf(tmpStr
, "R%u", reg
);
843 case WINED3DSPR_INPUT
:
845 /* Pixel shaders >= 3.0 */
846 if (WINED3DSHADER_VERSION_MAJOR(This
->baseShader
.hex_version
) >= 3) {
847 if (param
& WINED3DSHADER_ADDRMODE_RELATIVE
) {
848 glsl_src_param_t rel_param
;
849 shader_glsl_add_src_param(arg
, addr_token
, 0, WINED3DSP_WRITEMASK_0
, &rel_param
);
851 /* Removing a + 0 would be an obvious optimization, but macos doesn't see the NOP
854 if(((IWineD3DPixelShaderImpl
*) This
)->input_reg_map
[reg
]) {
855 sprintf(tmpStr
, "IN[%s + %u]", rel_param
.param_str
,
856 ((IWineD3DPixelShaderImpl
*) This
)->input_reg_map
[reg
]);
858 sprintf(tmpStr
, "IN[%s]", rel_param
.param_str
);
861 sprintf(tmpStr
, "IN[%u]",
862 ((IWineD3DPixelShaderImpl
*) This
)->input_reg_map
[reg
]);
866 strcpy(tmpStr
, "gl_Color");
868 strcpy(tmpStr
, "gl_SecondaryColor");
871 if (vshader_input_is_color((IWineD3DVertexShader
*) This
, reg
))
873 sprintf(tmpStr
, "attrib%u", reg
);
876 case WINED3DSPR_CONST
:
878 const char* prefix
= pshader
? "PC":"VC";
880 /* Relative addressing */
881 if (param
& WINED3DSHADER_ADDRMODE_RELATIVE
) {
883 /* Relative addressing on shaders 2.0+ have a relative address token,
884 * prior to that, it was hard-coded as "A0.x" because there's only 1 register */
885 if (WINED3DSHADER_VERSION_MAJOR(This
->baseShader
.hex_version
) >= 2) {
886 glsl_src_param_t rel_param
;
887 shader_glsl_add_src_param(arg
, addr_token
, 0, WINED3DSP_WRITEMASK_0
, &rel_param
);
889 sprintf(tmpStr
, "%s[%s + %u]", prefix
, rel_param
.param_str
, reg
);
891 sprintf(tmpStr
, "%s[%s]", prefix
, rel_param
.param_str
);
895 sprintf(tmpStr
, "%s[A0.x + %u]", prefix
, reg
);
897 sprintf(tmpStr
, "%s[A0.x]", prefix
);
902 if(shader_constant_is_local(This
, reg
)) {
903 sprintf(tmpStr
, "LC%u", reg
);
905 sprintf(tmpStr
, "%s[%u]", prefix
, reg
);
911 case WINED3DSPR_CONSTINT
:
913 sprintf(tmpStr
, "PI[%u]", reg
);
915 sprintf(tmpStr
, "VI[%u]", reg
);
917 case WINED3DSPR_CONSTBOOL
:
919 sprintf(tmpStr
, "PB[%u]", reg
);
921 sprintf(tmpStr
, "VB[%u]", reg
);
923 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
925 sprintf(tmpStr
, "T%u", reg
);
927 sprintf(tmpStr
, "A%u", reg
);
930 case WINED3DSPR_LOOP
:
931 sprintf(tmpStr
, "aL%u", This
->baseShader
.cur_loop_regno
- 1);
933 case WINED3DSPR_SAMPLER
:
935 sprintf(tmpStr
, "Psampler%u", reg
);
937 sprintf(tmpStr
, "Vsampler%u", reg
);
939 case WINED3DSPR_COLOROUT
:
940 if (reg
>= GL_LIMITS(buffers
)) {
941 WARN("Write to render target %u, only %d supported\n", reg
, 4);
943 if (GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
944 sprintf(tmpStr
, "gl_FragData[%u]", reg
);
945 } else { /* On older cards with GLSL support like the GeforceFX there's only one buffer. */
946 sprintf(tmpStr
, "gl_FragColor");
949 case WINED3DSPR_RASTOUT
:
950 sprintf(tmpStr
, "%s", hwrastout_reg_names
[reg
]);
952 case WINED3DSPR_DEPTHOUT
:
953 sprintf(tmpStr
, "gl_FragDepth");
955 case WINED3DSPR_ATTROUT
:
957 sprintf(tmpStr
, "gl_FrontColor");
959 sprintf(tmpStr
, "gl_FrontSecondaryColor");
962 case WINED3DSPR_TEXCRDOUT
:
963 /* Vertex shaders >= 3.0: WINED3DSPR_OUTPUT */
964 if (WINED3DSHADER_VERSION_MAJOR(This
->baseShader
.hex_version
) >= 3)
965 sprintf(tmpStr
, "OUT[%u]", reg
);
967 sprintf(tmpStr
, "gl_TexCoord[%u]", reg
);
969 case WINED3DSPR_MISCTYPE
:
972 sprintf(tmpStr
, "vpos");
973 } else if (reg
== 1){
974 /* Note that gl_FrontFacing is a bool, while vFace is
975 * a float for which the sign determines front/back
977 sprintf(tmpStr
, "(gl_FrontFacing ? 1.0 : -1.0)");
979 FIXME("Unhandled misctype register %d\n", reg
);
980 sprintf(tmpStr
, "unrecognized_register");
984 FIXME("Unhandled register name Type(%d)\n", regtype
);
985 sprintf(tmpStr
, "unrecognized_register");
989 strcat(regstr
, tmpStr
);
992 /* Get the GLSL write mask for the destination register */
993 static DWORD
shader_glsl_get_write_mask(const DWORD param
, char *write_mask
) {
994 char *ptr
= write_mask
;
995 DWORD mask
= param
& WINED3DSP_WRITEMASK_ALL
;
997 if (shader_is_scalar(param
)) {
998 mask
= WINED3DSP_WRITEMASK_0
;
1001 if (param
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
1002 if (param
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
1003 if (param
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
1004 if (param
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
1012 static unsigned int shader_glsl_get_write_mask_size(DWORD write_mask
) {
1013 unsigned int size
= 0;
1015 if (write_mask
& WINED3DSP_WRITEMASK_0
) ++size
;
1016 if (write_mask
& WINED3DSP_WRITEMASK_1
) ++size
;
1017 if (write_mask
& WINED3DSP_WRITEMASK_2
) ++size
;
1018 if (write_mask
& WINED3DSP_WRITEMASK_3
) ++size
;
1023 static void shader_glsl_get_swizzle(const DWORD param
, BOOL fixup
, DWORD mask
, char *swizzle_str
) {
1024 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
1025 * but addressed as "rgba". To fix this we need to swap the register's x
1026 * and z components. */
1027 DWORD swizzle
= (param
& WINED3DSP_SWIZZLE_MASK
) >> WINED3DSP_SWIZZLE_SHIFT
;
1028 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
1029 char *ptr
= swizzle_str
;
1031 if (!shader_is_scalar(param
)) {
1033 /* swizzle bits fields: wwzzyyxx */
1034 if (mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = swizzle_chars
[swizzle
& 0x03];
1035 if (mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = swizzle_chars
[(swizzle
>> 2) & 0x03];
1036 if (mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = swizzle_chars
[(swizzle
>> 4) & 0x03];
1037 if (mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = swizzle_chars
[(swizzle
>> 6) & 0x03];
1043 /* From a given parameter token, generate the corresponding GLSL string.
1044 * Also, return the actual register name and swizzle in case the
1045 * caller needs this information as well. */
1046 static void shader_glsl_add_src_param(SHADER_OPCODE_ARG
* arg
, const DWORD param
,
1047 const DWORD addr_token
, DWORD mask
, glsl_src_param_t
*src_param
) {
1048 BOOL is_color
= FALSE
;
1049 char swizzle_str
[6];
1051 src_param
->reg_name
[0] = '\0';
1052 src_param
->param_str
[0] = '\0';
1053 swizzle_str
[0] = '\0';
1055 shader_glsl_get_register_name(param
, addr_token
, src_param
->reg_name
, &is_color
, arg
);
1057 shader_glsl_get_swizzle(param
, is_color
, mask
, swizzle_str
);
1058 shader_glsl_gen_modifier(param
, src_param
->reg_name
, swizzle_str
, src_param
->param_str
);
1061 /* From a given parameter token, generate the corresponding GLSL string.
1062 * Also, return the actual register name and swizzle in case the
1063 * caller needs this information as well. */
1064 static DWORD
shader_glsl_add_dst_param(SHADER_OPCODE_ARG
* arg
, const DWORD param
,
1065 const DWORD addr_token
, glsl_dst_param_t
*dst_param
) {
1066 BOOL is_color
= FALSE
;
1068 dst_param
->mask_str
[0] = '\0';
1069 dst_param
->reg_name
[0] = '\0';
1071 shader_glsl_get_register_name(param
, addr_token
, dst_param
->reg_name
, &is_color
, arg
);
1072 return shader_glsl_get_write_mask(param
, dst_param
->mask_str
);
1075 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1076 static DWORD
shader_glsl_append_dst_ext(SHADER_BUFFER
*buffer
, SHADER_OPCODE_ARG
*arg
, const DWORD param
) {
1077 glsl_dst_param_t dst_param
;
1081 mask
= shader_glsl_add_dst_param(arg
, param
, arg
->dst_addr
, &dst_param
);
1084 shift
= (param
& WINED3DSP_DSTSHIFT_MASK
) >> WINED3DSP_DSTSHIFT_SHIFT
;
1085 shader_addline(buffer
, "%s%s = %s(", dst_param
.reg_name
, dst_param
.mask_str
, shift_glsl_tab
[shift
]);
1091 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1092 static DWORD
shader_glsl_append_dst(SHADER_BUFFER
*buffer
, SHADER_OPCODE_ARG
*arg
) {
1093 return shader_glsl_append_dst_ext(buffer
, arg
, arg
->dst
);
1096 /** Process GLSL instruction modifiers */
1097 void shader_glsl_add_instruction_modifiers(SHADER_OPCODE_ARG
* arg
) {
1099 DWORD mask
= arg
->dst
& WINED3DSP_DSTMOD_MASK
;
1101 if (arg
->opcode
->dst_token
&& mask
!= 0) {
1102 glsl_dst_param_t dst_param
;
1104 shader_glsl_add_dst_param(arg
, arg
->dst
, 0, &dst_param
);
1106 if (mask
& WINED3DSPDM_SATURATE
) {
1107 /* _SAT means to clamp the value of the register to between 0 and 1 */
1108 shader_addline(arg
->buffer
, "%s%s = clamp(%s%s, 0.0, 1.0);\n", dst_param
.reg_name
,
1109 dst_param
.mask_str
, dst_param
.reg_name
, dst_param
.mask_str
);
1111 if (mask
& WINED3DSPDM_MSAMPCENTROID
) {
1112 FIXME("_centroid modifier not handled\n");
1114 if (mask
& WINED3DSPDM_PARTIALPRECISION
) {
1115 /* MSDN says this modifier can be safely ignored, so that's what we'll do. */
1120 static inline const char* shader_get_comp_op(
1121 const DWORD opcode
) {
1123 DWORD op
= (opcode
& INST_CONTROLS_MASK
) >> INST_CONTROLS_SHIFT
;
1125 case COMPARISON_GT
: return ">";
1126 case COMPARISON_EQ
: return "==";
1127 case COMPARISON_GE
: return ">=";
1128 case COMPARISON_LT
: return "<";
1129 case COMPARISON_NE
: return "!=";
1130 case COMPARISON_LE
: return "<=";
1132 FIXME("Unrecognized comparison value: %u\n", op
);
1137 static void shader_glsl_get_sample_function(DWORD sampler_type
, BOOL projected
, BOOL texrect
, glsl_sample_function_t
*sample_function
) {
1138 /* Note that there's no such thing as a projected cube texture. */
1139 switch(sampler_type
) {
1141 sample_function
->name
= projected
? "texture1DProj" : "texture1D";
1142 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
;
1146 sample_function
->name
= projected
? "texture2DRectProj" : "texture2DRect";
1148 sample_function
->name
= projected
? "texture2DProj" : "texture2D";
1150 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
;
1152 case WINED3DSTT_CUBE
:
1153 sample_function
->name
= "textureCube";
1154 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1156 case WINED3DSTT_VOLUME
:
1157 sample_function
->name
= projected
? "texture3DProj" : "texture3D";
1158 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1161 sample_function
->name
= "";
1162 FIXME("Unrecognized sampler type: %#x;\n", sampler_type
);
1167 static void shader_glsl_color_correction(SHADER_OPCODE_ARG
* arg
) {
1168 IWineD3DBaseShaderImpl
* shader
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
1169 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) shader
->baseShader
.device
;
1170 WineD3D_GL_Info
*gl_info
= &deviceImpl
->adapter
->gl_info
;
1171 glsl_dst_param_t dst_param
;
1172 glsl_dst_param_t dst_param2
;
1174 WINED3DFORMAT conversion_group
;
1175 IWineD3DBaseTextureImpl
*texture
;
1176 DWORD mask
, mask_size
;
1178 BOOL recorded
= FALSE
;
1180 DWORD hex_version
= shader
->baseShader
.hex_version
;
1182 switch(arg
->opcode
->opcode
) {
1183 case WINED3DSIO_TEX
:
1184 if (hex_version
< WINED3DPS_VERSION(2,0)) {
1185 sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
1187 sampler_idx
= arg
->src
[1] & WINED3DSP_REGNUM_MASK
;
1191 case WINED3DSIO_TEXLDL
:
1192 FIXME("Add color fixup for vertex texture WINED3DSIO_TEXLDL\n");
1195 case WINED3DSIO_TEXDP3TEX
:
1196 case WINED3DSIO_TEXM3x3TEX
:
1197 case WINED3DSIO_TEXM3x3SPEC
:
1198 case WINED3DSIO_TEXM3x3VSPEC
:
1199 case WINED3DSIO_TEXBEM
:
1200 case WINED3DSIO_TEXREG2AR
:
1201 case WINED3DSIO_TEXREG2GB
:
1202 case WINED3DSIO_TEXREG2RGB
:
1203 sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
1207 /* Not a texture sampling instruction, nothing to do */
1211 texture
= (IWineD3DBaseTextureImpl
*) deviceImpl
->stateBlock
->textures
[sampler_idx
];
1213 fmt
= texture
->resource
.format
;
1214 conversion_group
= texture
->baseTexture
.shader_conversion_group
;
1216 fmt
= WINED3DFMT_UNKNOWN
;
1217 conversion_group
= WINED3DFMT_UNKNOWN
;
1220 /* before doing anything, record the sampler with the format in the format conversion list,
1221 * but check if it's not there already
1223 for(i
= 0; i
< shader
->baseShader
.num_sampled_samplers
; i
++) {
1224 if(shader
->baseShader
.sampled_samplers
[i
] == sampler_idx
) {
1230 shader
->baseShader
.sampled_samplers
[shader
->baseShader
.num_sampled_samplers
] = sampler_idx
;
1231 shader
->baseShader
.num_sampled_samplers
++;
1232 shader
->baseShader
.sampled_format
[sampler_idx
] = conversion_group
;
1236 case WINED3DFMT_V8U8
:
1237 case WINED3DFMT_V16U16
:
1238 if(GL_SUPPORT(NV_TEXTURE_SHADER
) ||
1239 (GL_SUPPORT(ATI_ENVMAP_BUMPMAP
) && fmt
== WINED3DFMT_V8U8
)) {
1240 /* The 3rd channel returns 1.0 in d3d, but 0.0 in gl. Fix this while we're at it :-) */
1241 mask
= shader_glsl_add_dst_param(arg
, arg
->dst
, WINED3DSP_WRITEMASK_2
, &dst_param
);
1242 mask_size
= shader_glsl_get_write_mask_size(mask
);
1243 if(mask_size
>= 3) {
1244 shader_addline(arg
->buffer
, "%s.%c = 1.0;\n", dst_param
.reg_name
, dst_param
.mask_str
[3]);
1247 /* Correct the sign, but leave the blue as it is - it was loaded correctly already */
1248 mask
= shader_glsl_add_dst_param(arg
, arg
->dst
,
1249 WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
,
1251 mask_size
= shader_glsl_get_write_mask_size(mask
);
1252 if(mask_size
>= 2) {
1253 shader_addline(arg
->buffer
, "%s.%c%c = %s.%c%c * 2.0 - 1.0;\n",
1254 dst_param
.reg_name
, dst_param
.mask_str
[1], dst_param
.mask_str
[2],
1255 dst_param
.reg_name
, dst_param
.mask_str
[1], dst_param
.mask_str
[2]);
1256 } else if(mask_size
== 1) {
1257 shader_addline(arg
->buffer
, "%s.%c = %s.%c * 2.0 - 1.0;\n", dst_param
.reg_name
, dst_param
.mask_str
[1],
1258 dst_param
.reg_name
, dst_param
.mask_str
[1]);
1263 case WINED3DFMT_X8L8V8U8
:
1264 if(!GL_SUPPORT(NV_TEXTURE_SHADER
)) {
1265 /* Red and blue are the signed channels, fix them up; Blue(=L) is correct already,
1266 * and a(X) is always 1.0
1268 mask
= shader_glsl_add_dst_param(arg
, arg
->dst
, WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &dst_param
);
1269 mask_size
= shader_glsl_get_write_mask_size(mask
);
1270 if(mask_size
>= 2) {
1271 shader_addline(arg
->buffer
, "%s.%c%c = %s.%c%c * 2.0 - 1.0;\n",
1272 dst_param
.reg_name
, dst_param
.mask_str
[1], dst_param
.mask_str
[2],
1273 dst_param
.reg_name
, dst_param
.mask_str
[1], dst_param
.mask_str
[2]);
1274 } else if(mask_size
== 1) {
1275 shader_addline(arg
->buffer
, "%s.%c = %s.%c * 2.0 - 1.0;\n",
1276 dst_param
.reg_name
, dst_param
.mask_str
[1],
1277 dst_param
.reg_name
, dst_param
.mask_str
[1]);
1282 case WINED3DFMT_L6V5U5
:
1283 if(!GL_SUPPORT(NV_TEXTURE_SHADER
)) {
1284 mask
= shader_glsl_add_dst_param(arg
, arg
->dst
, WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &dst_param
);
1285 mask_size
= shader_glsl_get_write_mask_size(mask
);
1286 shader_glsl_add_dst_param(arg
, arg
->dst
, WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_2
, &dst_param2
);
1287 if(mask_size
>= 3) {
1288 /* Swap y and z (U and L), and do a sign conversion on x and the new y(V and U) */
1289 shader_addline(arg
->buffer
, "tmp0.g = %s.%c;\n",
1290 dst_param
.reg_name
, dst_param
.mask_str
[2]);
1291 shader_addline(arg
->buffer
, "%s.%c%c = %s.%c%c * 2.0 - 1.0;\n",
1292 dst_param
.reg_name
, dst_param
.mask_str
[2], dst_param
.mask_str
[1],
1293 dst_param2
.reg_name
, dst_param
.mask_str
[1], dst_param
.mask_str
[3]);
1294 shader_addline(arg
->buffer
, "%s.%c = tmp0.g;\n", dst_param
.reg_name
,
1295 dst_param
.mask_str
[3]);
1296 } else if(mask_size
== 2) {
1297 /* This is bad: We have VL, but we need VU */
1298 FIXME("2 components sampled from a converted L6V5U5 texture\n");
1300 shader_addline(arg
->buffer
, "%s.%c = %s.%c * 2.0 - 1.0;\n",
1301 dst_param
.reg_name
, dst_param
.mask_str
[1],
1302 dst_param2
.reg_name
, dst_param
.mask_str
[1]);
1307 case WINED3DFMT_Q8W8V8U8
:
1308 if(!GL_SUPPORT(NV_TEXTURE_SHADER
)) {
1309 /* Correct the sign in all channels. The writemask just applies as-is, no
1310 * need for checking the mask size
1312 shader_glsl_add_dst_param(arg
, arg
->dst
,
1313 WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
|
1314 WINED3DSP_WRITEMASK_2
| WINED3DSP_WRITEMASK_3
,
1316 shader_addline(arg
->buffer
, "%s%s = %s%s * 2.0 - 1.0;\n", dst_param
.reg_name
, dst_param
.mask_str
,
1317 dst_param
.reg_name
, dst_param
.mask_str
);
1321 /* stupid compiler */
1327 /*****************************************************************************
1329 * Begin processing individual instruction opcodes
1331 ****************************************************************************/
1333 /* Generate GLSL arithmetic functions (dst = src1 + src2) */
1334 void shader_glsl_arith(SHADER_OPCODE_ARG
* arg
) {
1335 CONST SHADER_OPCODE
* curOpcode
= arg
->opcode
;
1336 SHADER_BUFFER
* buffer
= arg
->buffer
;
1337 glsl_src_param_t src0_param
;
1338 glsl_src_param_t src1_param
;
1342 /* Determine the GLSL operator to use based on the opcode */
1343 switch (curOpcode
->opcode
) {
1344 case WINED3DSIO_MUL
: op
= '*'; break;
1345 case WINED3DSIO_ADD
: op
= '+'; break;
1346 case WINED3DSIO_SUB
: op
= '-'; break;
1349 FIXME("Opcode %s not yet handled in GLSL\n", curOpcode
->name
);
1353 write_mask
= shader_glsl_append_dst(buffer
, arg
);
1354 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
1355 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1356 shader_addline(buffer
, "%s %c %s);\n", src0_param
.param_str
, op
, src1_param
.param_str
);
1359 /* Process the WINED3DSIO_MOV opcode using GLSL (dst = src) */
1360 void shader_glsl_mov(SHADER_OPCODE_ARG
* arg
) {
1361 IWineD3DBaseShaderImpl
* shader
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
1362 SHADER_BUFFER
* buffer
= arg
->buffer
;
1363 glsl_src_param_t src0_param
;
1366 write_mask
= shader_glsl_append_dst(buffer
, arg
);
1367 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
1369 /* In vs_1_1 WINED3DSIO_MOV can write to the address register. In later
1370 * shader versions WINED3DSIO_MOVA is used for this. */
1371 if ((WINED3DSHADER_VERSION_MAJOR(shader
->baseShader
.hex_version
) == 1 &&
1372 !shader_is_pshader_version(shader
->baseShader
.hex_version
) &&
1373 shader_get_regtype(arg
->dst
) == WINED3DSPR_ADDR
)) {
1374 /* This is a simple floor() */
1375 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1376 if (mask_size
> 1) {
1377 shader_addline(buffer
, "ivec%d(floor(%s)));\n", mask_size
, src0_param
.param_str
);
1379 shader_addline(buffer
, "int(floor(%s)));\n", src0_param
.param_str
);
1381 } else if(arg
->opcode
->opcode
== WINED3DSIO_MOVA
) {
1382 /* We need to *round* to the nearest int here. */
1383 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1384 if (mask_size
> 1) {
1385 shader_addline(buffer
, "ivec%d(floor(abs(%s) + vec%d(0.5)) * sign(%s)));\n", mask_size
, src0_param
.param_str
, mask_size
, src0_param
.param_str
);
1387 shader_addline(buffer
, "int(floor(abs(%s) + 0.5) * sign(%s)));\n", src0_param
.param_str
, src0_param
.param_str
);
1390 shader_addline(buffer
, "%s);\n", src0_param
.param_str
);
1394 /* Process the dot product operators DP3 and DP4 in GLSL (dst = dot(src0, src1)) */
1395 void shader_glsl_dot(SHADER_OPCODE_ARG
* arg
) {
1396 CONST SHADER_OPCODE
* curOpcode
= arg
->opcode
;
1397 SHADER_BUFFER
* buffer
= arg
->buffer
;
1398 glsl_src_param_t src0_param
;
1399 glsl_src_param_t src1_param
;
1400 DWORD dst_write_mask
, src_write_mask
;
1401 unsigned int dst_size
= 0;
1403 dst_write_mask
= shader_glsl_append_dst(buffer
, arg
);
1404 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1406 /* dp3 works on vec3, dp4 on vec4 */
1407 if (curOpcode
->opcode
== WINED3DSIO_DP4
) {
1408 src_write_mask
= WINED3DSP_WRITEMASK_ALL
;
1410 src_write_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1413 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_write_mask
, &src0_param
);
1414 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], src_write_mask
, &src1_param
);
1417 shader_addline(buffer
, "vec%d(dot(%s, %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
1419 shader_addline(buffer
, "dot(%s, %s));\n", src0_param
.param_str
, src1_param
.param_str
);
1423 /* Note that this instruction has some restrictions. The destination write mask
1424 * can't contain the w component, and the source swizzles have to be .xyzw */
1425 void shader_glsl_cross(SHADER_OPCODE_ARG
*arg
) {
1426 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1427 glsl_src_param_t src0_param
;
1428 glsl_src_param_t src1_param
;
1431 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
1432 shader_glsl_append_dst(arg
->buffer
, arg
);
1433 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
1434 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], src_mask
, &src1_param
);
1435 shader_addline(arg
->buffer
, "cross(%s, %s)%s);\n", src0_param
.param_str
, src1_param
.param_str
, dst_mask
);
1438 /* Process the WINED3DSIO_POW instruction in GLSL (dst = |src0|^src1)
1439 * Src0 and src1 are scalars. Note that D3D uses the absolute of src0, while
1440 * GLSL uses the value as-is. */
1441 void shader_glsl_pow(SHADER_OPCODE_ARG
*arg
) {
1442 SHADER_BUFFER
*buffer
= arg
->buffer
;
1443 glsl_src_param_t src0_param
;
1444 glsl_src_param_t src1_param
;
1445 DWORD dst_write_mask
;
1446 unsigned int dst_size
;
1448 dst_write_mask
= shader_glsl_append_dst(buffer
, arg
);
1449 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1451 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1452 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
1455 shader_addline(buffer
, "vec%d(pow(abs(%s), %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
1457 shader_addline(buffer
, "pow(abs(%s), %s));\n", src0_param
.param_str
, src1_param
.param_str
);
1461 /* Process the WINED3DSIO_LOG instruction in GLSL (dst = log2(|src0|))
1462 * Src0 is a scalar. Note that D3D uses the absolute of src0, while
1463 * GLSL uses the value as-is. */
1464 void shader_glsl_log(SHADER_OPCODE_ARG
*arg
) {
1465 SHADER_BUFFER
*buffer
= arg
->buffer
;
1466 glsl_src_param_t src0_param
;
1467 DWORD dst_write_mask
;
1468 unsigned int dst_size
;
1470 dst_write_mask
= shader_glsl_append_dst(buffer
, arg
);
1471 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1473 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1476 shader_addline(buffer
, "vec%d(log2(abs(%s))));\n", dst_size
, src0_param
.param_str
);
1478 shader_addline(buffer
, "log2(abs(%s)));\n", src0_param
.param_str
);
1482 /* Map the opcode 1-to-1 to the GL code (arg->dst = instruction(src0, src1, ...) */
1483 void shader_glsl_map2gl(SHADER_OPCODE_ARG
* arg
) {
1484 CONST SHADER_OPCODE
* curOpcode
= arg
->opcode
;
1485 SHADER_BUFFER
* buffer
= arg
->buffer
;
1486 glsl_src_param_t src_param
;
1487 const char *instruction
;
1488 char arguments
[256];
1492 /* Determine the GLSL function to use based on the opcode */
1493 /* TODO: Possibly make this a table for faster lookups */
1494 switch (curOpcode
->opcode
) {
1495 case WINED3DSIO_MIN
: instruction
= "min"; break;
1496 case WINED3DSIO_MAX
: instruction
= "max"; break;
1497 case WINED3DSIO_ABS
: instruction
= "abs"; break;
1498 case WINED3DSIO_FRC
: instruction
= "fract"; break;
1499 case WINED3DSIO_NRM
: instruction
= "normalize"; break;
1500 case WINED3DSIO_LOGP
:
1501 case WINED3DSIO_LOG
: instruction
= "log2"; break;
1502 case WINED3DSIO_EXP
: instruction
= "exp2"; break;
1503 case WINED3DSIO_SGN
: instruction
= "sign"; break;
1504 case WINED3DSIO_DSX
: instruction
= "dFdx"; break;
1505 case WINED3DSIO_DSY
: instruction
= "ycorrection.y * dFdy"; break;
1506 default: instruction
= "";
1507 FIXME("Opcode %s not yet handled in GLSL\n", curOpcode
->name
);
1511 write_mask
= shader_glsl_append_dst(buffer
, arg
);
1513 arguments
[0] = '\0';
1514 if (curOpcode
->num_params
> 0) {
1515 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src_param
);
1516 strcat(arguments
, src_param
.param_str
);
1517 for (i
= 2; i
< curOpcode
->num_params
; ++i
) {
1518 strcat(arguments
, ", ");
1519 shader_glsl_add_src_param(arg
, arg
->src
[i
-1], arg
->src_addr
[i
-1], write_mask
, &src_param
);
1520 strcat(arguments
, src_param
.param_str
);
1524 shader_addline(buffer
, "%s(%s));\n", instruction
, arguments
);
1527 /** Process the WINED3DSIO_EXPP instruction in GLSL:
1528 * For shader model 1.x, do the following (and honor the writemask, so use a temporary variable):
1529 * dst.x = 2^(floor(src))
1530 * dst.y = src - floor(src)
1531 * dst.z = 2^src (partial precision is allowed, but optional)
1533 * For 2.0 shaders, just do this (honoring writemask and swizzle):
1534 * dst = 2^src; (partial precision is allowed, but optional)
1536 void shader_glsl_expp(SHADER_OPCODE_ARG
* arg
) {
1537 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)arg
->shader
;
1538 glsl_src_param_t src_param
;
1540 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src_param
);
1542 if (shader
->baseShader
.hex_version
< WINED3DPS_VERSION(2,0)) {
1545 shader_addline(arg
->buffer
, "tmp0.x = exp2(floor(%s));\n", src_param
.param_str
);
1546 shader_addline(arg
->buffer
, "tmp0.y = %s - floor(%s);\n", src_param
.param_str
, src_param
.param_str
);
1547 shader_addline(arg
->buffer
, "tmp0.z = exp2(%s);\n", src_param
.param_str
);
1548 shader_addline(arg
->buffer
, "tmp0.w = 1.0;\n");
1550 shader_glsl_append_dst(arg
->buffer
, arg
);
1551 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
1552 shader_addline(arg
->buffer
, "tmp0%s);\n", dst_mask
);
1555 unsigned int mask_size
;
1557 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1558 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1560 if (mask_size
> 1) {
1561 shader_addline(arg
->buffer
, "vec%d(exp2(%s)));\n", mask_size
, src_param
.param_str
);
1563 shader_addline(arg
->buffer
, "exp2(%s));\n", src_param
.param_str
);
1568 /** Process the RCP (reciprocal or inverse) opcode in GLSL (dst = 1 / src) */
1569 void shader_glsl_rcp(SHADER_OPCODE_ARG
* arg
) {
1570 glsl_src_param_t src_param
;
1572 unsigned int mask_size
;
1574 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1575 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1576 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
1578 if (mask_size
> 1) {
1579 shader_addline(arg
->buffer
, "vec%d(1.0 / %s));\n", mask_size
, src_param
.param_str
);
1581 shader_addline(arg
->buffer
, "1.0 / %s);\n", src_param
.param_str
);
1585 void shader_glsl_rsq(SHADER_OPCODE_ARG
* arg
) {
1586 SHADER_BUFFER
* buffer
= arg
->buffer
;
1587 glsl_src_param_t src_param
;
1589 unsigned int mask_size
;
1591 write_mask
= shader_glsl_append_dst(buffer
, arg
);
1592 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1594 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
1596 if (mask_size
> 1) {
1597 shader_addline(buffer
, "vec%d(inversesqrt(%s)));\n", mask_size
, src_param
.param_str
);
1599 shader_addline(buffer
, "inversesqrt(%s));\n", src_param
.param_str
);
1603 /** Process signed comparison opcodes in GLSL. */
1604 void shader_glsl_compare(SHADER_OPCODE_ARG
* arg
) {
1605 glsl_src_param_t src0_param
;
1606 glsl_src_param_t src1_param
;
1608 unsigned int mask_size
;
1610 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1611 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1612 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
1613 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1615 if (mask_size
> 1) {
1616 const char *compare
;
1618 switch(arg
->opcode
->opcode
) {
1619 case WINED3DSIO_SLT
: compare
= "lessThan"; break;
1620 case WINED3DSIO_SGE
: compare
= "greaterThanEqual"; break;
1621 default: compare
= "";
1622 FIXME("Can't handle opcode %s\n", arg
->opcode
->name
);
1625 shader_addline(arg
->buffer
, "vec%d(%s(%s, %s)));\n", mask_size
, compare
,
1626 src0_param
.param_str
, src1_param
.param_str
);
1628 switch(arg
->opcode
->opcode
) {
1629 case WINED3DSIO_SLT
:
1630 /* Step(src0, src1) is not suitable here because if src0 == src1 SLT is supposed,
1631 * to return 0.0 but step returns 1.0 because step is not < x
1632 * An alternative is a bvec compare padded with an unused second component.
1633 * step(src1 * -1.0, src0 * -1.0) is not an option because it suffers from the same
1634 * issue. Playing with not() is not possible either because not() does not accept
1637 shader_addline(arg
->buffer
, "(%s < %s) ? 1.0 : 0.0);\n", src0_param
.param_str
, src1_param
.param_str
);
1639 case WINED3DSIO_SGE
:
1640 /* Here we can use the step() function and safe a conditional */
1641 shader_addline(arg
->buffer
, "step(%s, %s));\n", src1_param
.param_str
, src0_param
.param_str
);
1644 FIXME("Can't handle opcode %s\n", arg
->opcode
->name
);
1650 /** Process CMP instruction in GLSL (dst = src0 >= 0.0 ? src1 : src2), per channel */
1651 void shader_glsl_cmp(SHADER_OPCODE_ARG
* arg
) {
1652 glsl_src_param_t src0_param
;
1653 glsl_src_param_t src1_param
;
1654 glsl_src_param_t src2_param
;
1655 DWORD write_mask
, cmp_channel
= 0;
1658 BOOL temp_destination
= FALSE
;
1660 if(shader_is_scalar(arg
->src
[0])) {
1661 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1663 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
1664 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1665 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1667 shader_addline(arg
->buffer
, "%s >= 0.0 ? %s : %s);\n",
1668 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1670 DWORD src0reg
= arg
->src
[0] & WINED3DSP_REGNUM_MASK
;
1671 DWORD src1reg
= arg
->src
[1] & WINED3DSP_REGNUM_MASK
;
1672 DWORD src2reg
= arg
->src
[2] & WINED3DSP_REGNUM_MASK
;
1673 DWORD src0regtype
= shader_get_regtype(arg
->src
[0]);
1674 DWORD src1regtype
= shader_get_regtype(arg
->src
[1]);
1675 DWORD src2regtype
= shader_get_regtype(arg
->src
[2]);
1676 DWORD dstreg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
1677 DWORD dstregtype
= shader_get_regtype(arg
->dst
);
1679 /* Cycle through all source0 channels */
1680 for (i
=0; i
<4; i
++) {
1682 /* Find the destination channels which use the current source0 channel */
1683 for (j
=0; j
<4; j
++) {
1684 if ( ((arg
->src
[0] >> (WINED3DSP_SWIZZLE_SHIFT
+ 2*j
)) & 0x3) == i
) {
1685 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
1686 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
1690 /* Splitting the cmp instruction up in multiple lines imposes a problem:
1691 * The first lines may overwrite source parameters of the following lines.
1692 * Deal with that by using a temporary destination register if needed
1694 if((src0reg
== dstreg
&& src0regtype
== dstregtype
) ||
1695 (src1reg
== dstreg
&& src1regtype
== dstregtype
) ||
1696 (src2reg
== dstreg
&& src2regtype
== dstregtype
)) {
1698 write_mask
= shader_glsl_get_write_mask(arg
->dst
& (~WINED3DSP_SWIZZLE_MASK
| write_mask
), mask_char
);
1699 if (!write_mask
) continue;
1700 shader_addline(arg
->buffer
, "tmp0%s = (", mask_char
);
1701 temp_destination
= TRUE
;
1703 write_mask
= shader_glsl_append_dst_ext(arg
->buffer
, arg
, arg
->dst
& (~WINED3DSP_SWIZZLE_MASK
| write_mask
));
1704 if (!write_mask
) continue;
1707 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], cmp_channel
, &src0_param
);
1708 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1709 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1711 shader_addline(arg
->buffer
, "%s >= 0.0 ? %s : %s);\n",
1712 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1715 if(temp_destination
) {
1716 shader_glsl_get_write_mask(arg
->dst
, mask_char
);
1717 shader_glsl_append_dst_ext(arg
->buffer
, arg
, arg
->dst
);
1718 shader_addline(arg
->buffer
, "tmp0%s);\n", mask_char
);
1724 /** Process the CND opcode in GLSL (dst = (src0 > 0.5) ? src1 : src2) */
1725 /* For ps 1.1-1.3, only a single component of src0 is used. For ps 1.4
1726 * the compare is done per component of src0. */
1727 void shader_glsl_cnd(SHADER_OPCODE_ARG
* arg
) {
1728 IWineD3DBaseShaderImpl
* shader
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
1729 glsl_src_param_t src0_param
;
1730 glsl_src_param_t src1_param
;
1731 glsl_src_param_t src2_param
;
1732 DWORD write_mask
, cmp_channel
= 0;
1735 if (shader
->baseShader
.hex_version
< WINED3DPS_VERSION(1, 4)) {
1736 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1737 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1738 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1739 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1741 /* Fun: The D3DSI_COISSUE flag changes the semantic of the cnd instruction for < 1.4 shaders */
1742 if(arg
->opcode_token
& WINED3DSI_COISSUE
) {
1743 shader_addline(arg
->buffer
, "%s /* COISSUE! */);\n", src1_param
.param_str
);
1745 shader_addline(arg
->buffer
, "%s > 0.5 ? %s : %s);\n",
1746 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1750 /* Cycle through all source0 channels */
1751 for (i
=0; i
<4; i
++) {
1753 /* Find the destination channels which use the current source0 channel */
1754 for (j
=0; j
<4; j
++) {
1755 if ( ((arg
->src
[0] >> (WINED3DSP_SWIZZLE_SHIFT
+ 2*j
)) & 0x3) == i
) {
1756 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
1757 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
1760 write_mask
= shader_glsl_append_dst_ext(arg
->buffer
, arg
, arg
->dst
& (~WINED3DSP_SWIZZLE_MASK
| write_mask
));
1761 if (!write_mask
) continue;
1763 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], cmp_channel
, &src0_param
);
1764 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1765 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1767 shader_addline(arg
->buffer
, "%s > 0.5 ? %s : %s);\n",
1768 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1772 /** GLSL code generation for WINED3DSIO_MAD: Multiply the first 2 opcodes, then add the last */
1773 void shader_glsl_mad(SHADER_OPCODE_ARG
* arg
) {
1774 glsl_src_param_t src0_param
;
1775 glsl_src_param_t src1_param
;
1776 glsl_src_param_t src2_param
;
1779 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1780 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
1781 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1782 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1783 shader_addline(arg
->buffer
, "(%s * %s) + %s);\n",
1784 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1787 /** Handles transforming all WINED3DSIO_M?x? opcodes for
1788 Vertex shaders to GLSL codes */
1789 void shader_glsl_mnxn(SHADER_OPCODE_ARG
* arg
) {
1791 int nComponents
= 0;
1792 SHADER_OPCODE_ARG tmpArg
;
1794 memset(&tmpArg
, 0, sizeof(SHADER_OPCODE_ARG
));
1796 /* Set constants for the temporary argument */
1797 tmpArg
.shader
= arg
->shader
;
1798 tmpArg
.buffer
= arg
->buffer
;
1799 tmpArg
.src
[0] = arg
->src
[0];
1800 tmpArg
.src_addr
[0] = arg
->src_addr
[0];
1801 tmpArg
.src_addr
[1] = arg
->src_addr
[1];
1802 tmpArg
.reg_maps
= arg
->reg_maps
;
1804 switch(arg
->opcode
->opcode
) {
1805 case WINED3DSIO_M4x4
:
1807 tmpArg
.opcode
= shader_get_opcode(arg
->shader
, WINED3DSIO_DP4
);
1809 case WINED3DSIO_M4x3
:
1811 tmpArg
.opcode
= shader_get_opcode(arg
->shader
, WINED3DSIO_DP4
);
1813 case WINED3DSIO_M3x4
:
1815 tmpArg
.opcode
= shader_get_opcode(arg
->shader
, WINED3DSIO_DP3
);
1817 case WINED3DSIO_M3x3
:
1819 tmpArg
.opcode
= shader_get_opcode(arg
->shader
, WINED3DSIO_DP3
);
1821 case WINED3DSIO_M3x2
:
1823 tmpArg
.opcode
= shader_get_opcode(arg
->shader
, WINED3DSIO_DP3
);
1829 for (i
= 0; i
< nComponents
; i
++) {
1830 tmpArg
.dst
= ((arg
->dst
) & ~WINED3DSP_WRITEMASK_ALL
)|(WINED3DSP_WRITEMASK_0
<<i
);
1831 tmpArg
.src
[1] = arg
->src
[1]+i
;
1832 shader_glsl_dot(&tmpArg
);
1837 The LRP instruction performs a component-wise linear interpolation
1838 between the second and third operands using the first operand as the
1839 blend factor. Equation: (dst = src2 + src0 * (src1 - src2))
1840 This is equivalent to mix(src2, src1, src0);
1842 void shader_glsl_lrp(SHADER_OPCODE_ARG
* arg
) {
1843 glsl_src_param_t src0_param
;
1844 glsl_src_param_t src1_param
;
1845 glsl_src_param_t src2_param
;
1848 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1850 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
1851 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1852 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1854 shader_addline(arg
->buffer
, "mix(%s, %s, %s));\n",
1855 src2_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
);
1858 /** Process the WINED3DSIO_LIT instruction in GLSL:
1859 * dst.x = dst.w = 1.0
1860 * dst.y = (src0.x > 0) ? src0.x
1861 * dst.z = (src0.x > 0) ? ((src0.y > 0) ? pow(src0.y, src.w) : 0) : 0
1862 * where src.w is clamped at +- 128
1864 void shader_glsl_lit(SHADER_OPCODE_ARG
* arg
) {
1865 glsl_src_param_t src0_param
;
1866 glsl_src_param_t src1_param
;
1867 glsl_src_param_t src3_param
;
1870 shader_glsl_append_dst(arg
->buffer
, arg
);
1871 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
1873 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1874 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_1
, &src1_param
);
1875 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &src3_param
);
1877 /* The sdk specifies the instruction like this
1879 * if(src.x > 0.0) dst.y = src.x
1881 * if(src.x > 0.0 && src.y > 0.0) dst.z = pow(src.y, power);
1885 * Obviously that has quite a few conditionals in it which we don't like. So the first step is this:
1886 * dst.x = 1.0 ... No further explanation needed
1887 * dst.y = max(src.y, 0.0); ... If x < 0.0, use 0.0, otherwise x. Same as the conditional
1888 * dst.z = x > 0.0 ? pow(max(y, 0.0), p) : 0; ... 0 ^ power is 0, and otherwise we use y anyway
1889 * dst.w = 1.0. ... Nothing fancy.
1891 * So we still have one conditional in there. So do this:
1892 * dst.z = pow(max(0.0, src.y) * step(0.0, src.x), power);
1894 * step(0.0, x) will return 1 if src.x > 0.0, and 0 otherwise. So if y is 0 we get pow(0.0 * 1.0, power),
1895 * which sets dst.z to 0. If y > 0, but x = 0.0, we get pow(y * 0.0, power), which results in 0 too.
1896 * if both x and y are > 0, we get pow(y * 1.0, power), as it is supposed to
1898 shader_addline(arg
->buffer
, "vec4(1.0, max(%s, 0.0), pow(max(0.0, %s) * step(0.0, %s), clamp(%s, -128.0, 128.0)), 1.0)%s);\n",
1899 src0_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
, src3_param
.param_str
, dst_mask
);
1902 /** Process the WINED3DSIO_DST instruction in GLSL:
1904 * dst.y = src0.x * src0.y
1908 void shader_glsl_dst(SHADER_OPCODE_ARG
* arg
) {
1909 glsl_src_param_t src0y_param
;
1910 glsl_src_param_t src0z_param
;
1911 glsl_src_param_t src1y_param
;
1912 glsl_src_param_t src1w_param
;
1915 shader_glsl_append_dst(arg
->buffer
, arg
);
1916 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
1918 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_1
, &src0y_param
);
1919 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_2
, &src0z_param
);
1920 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_1
, &src1y_param
);
1921 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_3
, &src1w_param
);
1923 shader_addline(arg
->buffer
, "vec4(1.0, %s * %s, %s, %s))%s;\n",
1924 src0y_param
.param_str
, src1y_param
.param_str
, src0z_param
.param_str
, src1w_param
.param_str
, dst_mask
);
1927 /** Process the WINED3DSIO_SINCOS instruction in GLSL:
1928 * VS 2.0 requires that specific cosine and sine constants be passed to this instruction so the hardware
1929 * can handle it. But, these functions are built-in for GLSL, so we can just ignore the last 2 params.
1931 * dst.x = cos(src0.?)
1932 * dst.y = sin(src0.?)
1936 void shader_glsl_sincos(SHADER_OPCODE_ARG
* arg
) {
1937 glsl_src_param_t src0_param
;
1940 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1941 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1943 switch (write_mask
) {
1944 case WINED3DSP_WRITEMASK_0
:
1945 shader_addline(arg
->buffer
, "cos(%s));\n", src0_param
.param_str
);
1948 case WINED3DSP_WRITEMASK_1
:
1949 shader_addline(arg
->buffer
, "sin(%s));\n", src0_param
.param_str
);
1952 case (WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
):
1953 shader_addline(arg
->buffer
, "vec2(cos(%s), sin(%s)));\n", src0_param
.param_str
, src0_param
.param_str
);
1957 ERR("Write mask should be .x, .y or .xy\n");
1962 /** Process the WINED3DSIO_LOOP instruction in GLSL:
1963 * Start a for() loop where src1.y is the initial value of aL,
1964 * increment aL by src1.z for a total of src1.x iterations.
1965 * Need to use a temporary variable for this operation.
1967 /* FIXME: I don't think nested loops will work correctly this way. */
1968 void shader_glsl_loop(SHADER_OPCODE_ARG
* arg
) {
1969 glsl_src_param_t src1_param
;
1970 IWineD3DBaseShaderImpl
* shader
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
1971 DWORD regtype
= shader_get_regtype(arg
->src
[1]);
1972 DWORD reg
= arg
->src
[1] & WINED3DSP_REGNUM_MASK
;
1973 const DWORD
*control_values
= NULL
;
1974 local_constant
*constant
;
1976 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_ALL
, &src1_param
);
1978 /* Try to hardcode the loop control parameters if possible. Direct3D 9 class hardware doesn't support real
1979 * varying indexing, but Microsoft designed this feature for Shader model 2.x+. If the loop control is
1980 * known at compile time, the GLSL compiler can unroll the loop, and replace indirect addressing with direct
1983 if(regtype
== WINED3DSPR_CONSTINT
) {
1984 LIST_FOR_EACH_ENTRY(constant
, &shader
->baseShader
.constantsI
, local_constant
, entry
) {
1985 if(constant
->idx
== reg
) {
1986 control_values
= constant
->value
;
1992 if(control_values
) {
1993 if(control_values
[2] > 0) {
1994 shader_addline(arg
->buffer
, "for (aL%u = %d; aL%u < (%d * %d + %d); aL%u += %d) {\n",
1995 shader
->baseShader
.cur_loop_depth
, control_values
[1],
1996 shader
->baseShader
.cur_loop_depth
, control_values
[0], control_values
[2], control_values
[1],
1997 shader
->baseShader
.cur_loop_depth
, control_values
[2]);
1998 } else if(control_values
[2] == 0) {
1999 shader_addline(arg
->buffer
, "for (aL%u = %d, tmpInt%u = 0; tmpInt%u < %d; tmpInt%u++) {\n",
2000 shader
->baseShader
.cur_loop_depth
, control_values
[1], shader
->baseShader
.cur_loop_depth
,
2001 shader
->baseShader
.cur_loop_depth
, control_values
[0],
2002 shader
->baseShader
.cur_loop_depth
);
2004 shader_addline(arg
->buffer
, "for (aL%u = %d; aL%u > (%d * %d + %d); aL%u += %d) {\n",
2005 shader
->baseShader
.cur_loop_depth
, control_values
[1],
2006 shader
->baseShader
.cur_loop_depth
, control_values
[0], control_values
[2], control_values
[1],
2007 shader
->baseShader
.cur_loop_depth
, control_values
[2]);
2010 shader_addline(arg
->buffer
, "for (tmpInt%u = 0, aL%u = %s.y; tmpInt%u < %s.x; tmpInt%u++, aL%u += %s.z) {\n",
2011 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
,
2012 src1_param
.reg_name
, shader
->baseShader
.cur_loop_depth
, src1_param
.reg_name
,
2013 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
, src1_param
.reg_name
);
2016 shader
->baseShader
.cur_loop_depth
++;
2017 shader
->baseShader
.cur_loop_regno
++;
2020 void shader_glsl_end(SHADER_OPCODE_ARG
* arg
) {
2021 IWineD3DBaseShaderImpl
* shader
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
2023 shader_addline(arg
->buffer
, "}\n");
2025 if(arg
->opcode
->opcode
== WINED3DSIO_ENDLOOP
) {
2026 shader
->baseShader
.cur_loop_depth
--;
2027 shader
->baseShader
.cur_loop_regno
--;
2029 if(arg
->opcode
->opcode
== WINED3DSIO_ENDREP
) {
2030 shader
->baseShader
.cur_loop_depth
--;
2034 void shader_glsl_rep(SHADER_OPCODE_ARG
* arg
) {
2035 IWineD3DBaseShaderImpl
* shader
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
2036 glsl_src_param_t src0_param
;
2038 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2039 shader_addline(arg
->buffer
, "for (tmpInt%d = 0; tmpInt%d < %s; tmpInt%d++) {\n",
2040 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_depth
,
2041 src0_param
.param_str
, shader
->baseShader
.cur_loop_depth
);
2042 shader
->baseShader
.cur_loop_depth
++;
2045 void shader_glsl_if(SHADER_OPCODE_ARG
* arg
) {
2046 glsl_src_param_t src0_param
;
2048 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2049 shader_addline(arg
->buffer
, "if (%s) {\n", src0_param
.param_str
);
2052 void shader_glsl_ifc(SHADER_OPCODE_ARG
* arg
) {
2053 glsl_src_param_t src0_param
;
2054 glsl_src_param_t src1_param
;
2056 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2057 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2059 shader_addline(arg
->buffer
, "if (%s %s %s) {\n",
2060 src0_param
.param_str
, shader_get_comp_op(arg
->opcode_token
), src1_param
.param_str
);
2063 void shader_glsl_else(SHADER_OPCODE_ARG
* arg
) {
2064 shader_addline(arg
->buffer
, "} else {\n");
2067 void shader_glsl_break(SHADER_OPCODE_ARG
* arg
) {
2068 shader_addline(arg
->buffer
, "break;\n");
2071 /* FIXME: According to MSDN the compare is done per component. */
2072 void shader_glsl_breakc(SHADER_OPCODE_ARG
* arg
) {
2073 glsl_src_param_t src0_param
;
2074 glsl_src_param_t src1_param
;
2076 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2077 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2079 shader_addline(arg
->buffer
, "if (%s %s %s) break;\n",
2080 src0_param
.param_str
, shader_get_comp_op(arg
->opcode_token
), src1_param
.param_str
);
2083 void shader_glsl_label(SHADER_OPCODE_ARG
* arg
) {
2085 DWORD snum
= (arg
->src
[0]) & WINED3DSP_REGNUM_MASK
;
2086 shader_addline(arg
->buffer
, "}\n");
2087 shader_addline(arg
->buffer
, "void subroutine%u () {\n", snum
);
2090 void shader_glsl_call(SHADER_OPCODE_ARG
* arg
) {
2091 DWORD snum
= (arg
->src
[0]) & WINED3DSP_REGNUM_MASK
;
2092 shader_addline(arg
->buffer
, "subroutine%u();\n", snum
);
2095 void shader_glsl_callnz(SHADER_OPCODE_ARG
* arg
) {
2096 glsl_src_param_t src1_param
;
2098 DWORD snum
= (arg
->src
[0]) & WINED3DSP_REGNUM_MASK
;
2099 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2100 shader_addline(arg
->buffer
, "if (%s) subroutine%u();\n", src1_param
.param_str
, snum
);
2103 /*********************************************
2104 * Pixel Shader Specific Code begins here
2105 ********************************************/
2106 void pshader_glsl_tex(SHADER_OPCODE_ARG
* arg
) {
2107 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2108 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2109 DWORD hex_version
= This
->baseShader
.hex_version
;
2110 char dst_swizzle
[6];
2111 glsl_sample_function_t sample_function
;
2114 BOOL projected
, texrect
= FALSE
;
2117 /* All versions have a destination register */
2118 shader_glsl_append_dst(arg
->buffer
, arg
);
2120 /* 1.0-1.4: Use destination register as sampler source.
2121 * 2.0+: Use provided sampler source. */
2122 if (hex_version
< WINED3DPS_VERSION(1,4)) {
2125 sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2126 flags
= deviceImpl
->stateBlock
->textureState
[sampler_idx
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
2128 if (flags
& WINED3DTTFF_PROJECTED
) {
2130 switch (flags
& ~WINED3DTTFF_PROJECTED
) {
2131 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
2132 case WINED3DTTFF_COUNT2
: mask
= WINED3DSP_WRITEMASK_1
; break;
2133 case WINED3DTTFF_COUNT3
: mask
= WINED3DSP_WRITEMASK_2
; break;
2134 case WINED3DTTFF_COUNT4
:
2135 case WINED3DTTFF_DISABLE
: mask
= WINED3DSP_WRITEMASK_3
; break;
2140 } else if (hex_version
< WINED3DPS_VERSION(2,0)) {
2141 DWORD src_mod
= arg
->src
[0] & WINED3DSP_SRCMOD_MASK
;
2142 sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2144 if (src_mod
== WINED3DSPSM_DZ
) {
2146 mask
= WINED3DSP_WRITEMASK_2
;
2147 } else if (src_mod
== WINED3DSPSM_DW
) {
2149 mask
= WINED3DSP_WRITEMASK_3
;
2154 sampler_idx
= arg
->src
[1] & WINED3DSP_REGNUM_MASK
;
2155 if(arg
->opcode_token
& WINED3DSI_TEXLD_PROJECT
) {
2156 /* ps 2.0 texldp instruction always divides by the fourth component. */
2158 mask
= WINED3DSP_WRITEMASK_3
;
2164 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
2165 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
2169 sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2170 shader_glsl_get_sample_function(sampler_type
, projected
, texrect
, &sample_function
);
2171 mask
|= sample_function
.coord_mask
;
2173 if (hex_version
< WINED3DPS_VERSION(2,0)) {
2174 shader_glsl_get_write_mask(arg
->dst
, dst_swizzle
);
2176 shader_glsl_get_swizzle(arg
->src
[1], FALSE
, arg
->dst
, dst_swizzle
);
2179 /* 1.0-1.3: Use destination register as coordinate source.
2180 1.4+: Use provided coordinate source register. */
2181 if (hex_version
< WINED3DPS_VERSION(1,4)) {
2183 shader_glsl_get_write_mask(mask
, coord_mask
);
2184 shader_addline(arg
->buffer
, "%s(Psampler%u, T%u%s)%s);\n",
2185 sample_function
.name
, sampler_idx
, sampler_idx
, coord_mask
, dst_swizzle
);
2187 glsl_src_param_t coord_param
;
2188 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], mask
, &coord_param
);
2189 if(arg
->opcode_token
& WINED3DSI_TEXLD_BIAS
) {
2190 glsl_src_param_t bias
;
2191 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &bias
);
2193 shader_addline(arg
->buffer
, "%s(Psampler%u, %s, %s)%s);\n",
2194 sample_function
.name
, sampler_idx
, coord_param
.param_str
,
2195 bias
.param_str
, dst_swizzle
);
2197 shader_addline(arg
->buffer
, "%s(Psampler%u, %s)%s);\n",
2198 sample_function
.name
, sampler_idx
, coord_param
.param_str
, dst_swizzle
);
2203 void shader_glsl_texldl(SHADER_OPCODE_ARG
* arg
) {
2204 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*)arg
->shader
;
2205 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2206 glsl_sample_function_t sample_function
;
2207 glsl_src_param_t coord_param
, lod_param
;
2208 char dst_swizzle
[6];
2211 BOOL texrect
= FALSE
;
2213 shader_glsl_append_dst(arg
->buffer
, arg
);
2214 shader_glsl_get_swizzle(arg
->src
[1], FALSE
, arg
->dst
, dst_swizzle
);
2216 sampler_idx
= arg
->src
[1] & WINED3DSP_REGNUM_MASK
;
2217 sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2218 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
2219 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
2222 shader_glsl_get_sample_function(sampler_type
, FALSE
, texrect
, &sample_function
); shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], sample_function
.coord_mask
, &coord_param
);
2224 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &lod_param
);
2226 if (shader_is_pshader_version(This
->baseShader
.hex_version
)) {
2227 /* The GLSL spec claims the Lod sampling functions are only supported in vertex shaders.
2228 * However, they seem to work just fine in fragment shaders as well. */
2229 WARN("Using %sLod in fragment shader.\n", sample_function
.name
);
2230 shader_addline(arg
->buffer
, "%sLod(Psampler%u, %s, %s)%s);\n",
2231 sample_function
.name
, sampler_idx
, coord_param
.param_str
, lod_param
.param_str
, dst_swizzle
);
2233 shader_addline(arg
->buffer
, "%sLod(Vsampler%u, %s, %s)%s);\n",
2234 sample_function
.name
, sampler_idx
, coord_param
.param_str
, lod_param
.param_str
, dst_swizzle
);
2238 void pshader_glsl_texcoord(SHADER_OPCODE_ARG
* arg
) {
2240 /* FIXME: Make this work for more than just 2D textures */
2242 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2243 SHADER_BUFFER
* buffer
= arg
->buffer
;
2244 DWORD hex_version
= This
->baseShader
.hex_version
;
2248 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
2249 shader_glsl_get_write_mask(write_mask
, dst_mask
);
2251 if (hex_version
!= WINED3DPS_VERSION(1,4)) {
2252 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2253 shader_addline(buffer
, "clamp(gl_TexCoord[%u], 0.0, 1.0)%s);\n", reg
, dst_mask
);
2255 DWORD reg
= arg
->src
[0] & WINED3DSP_REGNUM_MASK
;
2256 DWORD src_mod
= arg
->src
[0] & WINED3DSP_SRCMOD_MASK
;
2257 char dst_swizzle
[6];
2259 shader_glsl_get_swizzle(arg
->src
[0], FALSE
, write_mask
, dst_swizzle
);
2261 if (src_mod
== WINED3DSPSM_DZ
) {
2262 glsl_src_param_t div_param
;
2263 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2264 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_2
, &div_param
);
2266 if (mask_size
> 1) {
2267 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
2269 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
2271 } else if (src_mod
== WINED3DSPSM_DW
) {
2272 glsl_src_param_t div_param
;
2273 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2274 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &div_param
);
2276 if (mask_size
> 1) {
2277 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
2279 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
2282 shader_addline(buffer
, "gl_TexCoord[%u]%s);\n", reg
, dst_swizzle
);
2287 /** Process the WINED3DSIO_TEXDP3TEX instruction in GLSL:
2288 * Take a 3-component dot product of the TexCoord[dstreg] and src,
2289 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
2290 void pshader_glsl_texdp3tex(SHADER_OPCODE_ARG
* arg
) {
2291 glsl_src_param_t src0_param
;
2293 glsl_sample_function_t sample_function
;
2294 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2295 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2296 DWORD sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2298 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2300 shader_glsl_append_dst(arg
->buffer
, arg
);
2301 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2303 /* Do I have to take care about the projected bit? I don't think so, since the dp3 returns only one
2304 * scalar, and projected sampling would require 4.
2306 * It is a dependent read - not valid with conditional NP2 textures
2308 shader_glsl_get_sample_function(sampler_type
, FALSE
, FALSE
, &sample_function
);
2310 switch(count_bits(sample_function
.coord_mask
)) {
2312 shader_addline(arg
->buffer
, "%s(Psampler%u, dot(gl_TexCoord[%u].xyz, %s))%s);\n",
2313 sample_function
.name
, sampler_idx
, sampler_idx
, src0_param
.param_str
, dst_mask
);
2317 shader_addline(arg
->buffer
, "%s(Psampler%u, vec2(dot(gl_TexCoord[%u].xyz, %s), 0.0))%s);\n",
2318 sample_function
.name
, sampler_idx
, sampler_idx
, src0_param
.param_str
, dst_mask
);
2322 shader_addline(arg
->buffer
, "%s(Psampler%u, vec3(dot(gl_TexCoord[%u].xyz, %s), 0.0, 0.0))%s);\n",
2323 sample_function
.name
, sampler_idx
, sampler_idx
, src0_param
.param_str
, dst_mask
);
2326 FIXME("Unexpected mask bitcount %d\n", count_bits(sample_function
.coord_mask
));
2330 /** Process the WINED3DSIO_TEXDP3 instruction in GLSL:
2331 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
2332 void pshader_glsl_texdp3(SHADER_OPCODE_ARG
* arg
) {
2333 glsl_src_param_t src0_param
;
2334 DWORD dstreg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2335 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2337 unsigned int mask_size
;
2339 dst_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
2340 mask_size
= shader_glsl_get_write_mask_size(dst_mask
);
2341 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2343 if (mask_size
> 1) {
2344 shader_addline(arg
->buffer
, "vec%d(dot(T%u.xyz, %s)));\n", mask_size
, dstreg
, src0_param
.param_str
);
2346 shader_addline(arg
->buffer
, "dot(T%u.xyz, %s));\n", dstreg
, src0_param
.param_str
);
2350 /** Process the WINED3DSIO_TEXDEPTH instruction in GLSL:
2351 * Calculate the depth as dst.x / dst.y */
2352 void pshader_glsl_texdepth(SHADER_OPCODE_ARG
* arg
) {
2353 glsl_dst_param_t dst_param
;
2355 shader_glsl_add_dst_param(arg
, arg
->dst
, 0, &dst_param
);
2357 /* Tests show that texdepth never returns anything below 0.0, and that r5.y is clamped to 1.0.
2358 * Negative input is accepted, -0.25 / -0.5 returns 0.5. GL should clamp gl_FragDepth to [0;1], but
2359 * this doesn't always work, so clamp the results manually. Whether or not the x value is clamped at 1
2360 * too is irrelevant, since if x = 0, any y value < 1.0 (and > 1.0 is not allowed) results in a result
2363 shader_addline(arg
->buffer
, "gl_FragDepth = clamp((%s.x / min(%s.y, 1.0)), 0.0, 1.0);\n", dst_param
.reg_name
, dst_param
.reg_name
);
2366 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in GLSL:
2367 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
2368 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
2369 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
2371 void pshader_glsl_texm3x2depth(SHADER_OPCODE_ARG
* arg
) {
2372 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2373 DWORD dstreg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2374 glsl_src_param_t src0_param
;
2376 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2378 shader_addline(arg
->buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", dstreg
, src0_param
.param_str
);
2379 shader_addline(arg
->buffer
, "gl_FragDepth = (tmp0.y == 0.0) ? 1.0 : clamp(tmp0.x / tmp0.y, 0.0, 1.0);\n");
2382 /** Process the WINED3DSIO_TEXM3X2PAD instruction in GLSL
2383 * Calculate the 1st of a 2-row matrix multiplication. */
2384 void pshader_glsl_texm3x2pad(SHADER_OPCODE_ARG
* arg
) {
2385 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2386 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2387 SHADER_BUFFER
* buffer
= arg
->buffer
;
2388 glsl_src_param_t src0_param
;
2390 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2391 shader_addline(buffer
, "tmp0.x = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2394 /** Process the WINED3DSIO_TEXM3X3PAD instruction in GLSL
2395 * Calculate the 1st or 2nd row of a 3-row matrix multiplication. */
2396 void pshader_glsl_texm3x3pad(SHADER_OPCODE_ARG
* arg
) {
2398 IWineD3DPixelShaderImpl
* shader
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2399 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2400 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2401 SHADER_BUFFER
* buffer
= arg
->buffer
;
2402 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2403 glsl_src_param_t src0_param
;
2405 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2406 shader_addline(buffer
, "tmp0.%c = dot(T%u.xyz, %s);\n", 'x' + current_state
->current_row
, reg
, src0_param
.param_str
);
2407 current_state
->texcoord_w
[current_state
->current_row
++] = reg
;
2410 void pshader_glsl_texm3x2tex(SHADER_OPCODE_ARG
* arg
) {
2411 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2412 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2413 SHADER_BUFFER
* buffer
= arg
->buffer
;
2414 glsl_src_param_t src0_param
;
2417 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2418 shader_addline(buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2420 shader_glsl_append_dst(buffer
, arg
);
2421 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2423 /* Sample the texture using the calculated coordinates */
2424 shader_addline(buffer
, "texture2D(Psampler%u, tmp0.xy)%s);\n", reg
, dst_mask
);
2427 /** Process the WINED3DSIO_TEXM3X3TEX instruction in GLSL
2428 * Perform the 3rd row of a 3x3 matrix multiply, then sample the texture using the calculated coordinates */
2429 void pshader_glsl_texm3x3tex(SHADER_OPCODE_ARG
* arg
) {
2430 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2431 glsl_src_param_t src0_param
;
2433 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2434 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2435 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
2436 DWORD sampler_type
= arg
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2437 glsl_sample_function_t sample_function
;
2439 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2440 shader_addline(arg
->buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2442 shader_glsl_append_dst(arg
->buffer
, arg
);
2443 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2444 /* Dependent read, not valid with conditional NP2 */
2445 shader_glsl_get_sample_function(sampler_type
, FALSE
, FALSE
, &sample_function
);
2447 /* Sample the texture using the calculated coordinates */
2448 shader_addline(arg
->buffer
, "%s(Psampler%u, tmp0.xyz)%s);\n", sample_function
.name
, reg
, dst_mask
);
2450 current_state
->current_row
= 0;
2453 /** Process the WINED3DSIO_TEXM3X3 instruction in GLSL
2454 * Perform the 3rd row of a 3x3 matrix multiply */
2455 void pshader_glsl_texm3x3(SHADER_OPCODE_ARG
* arg
) {
2456 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2457 glsl_src_param_t src0_param
;
2459 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2460 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2461 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
2463 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2465 shader_glsl_append_dst(arg
->buffer
, arg
);
2466 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2467 shader_addline(arg
->buffer
, "vec4(tmp0.xy, dot(T%u.xyz, %s), 1.0)%s);\n", reg
, src0_param
.param_str
, dst_mask
);
2469 current_state
->current_row
= 0;
2472 /** Process the WINED3DSIO_TEXM3X3SPEC instruction in GLSL
2473 * Peform the final texture lookup based on the previous 2 3x3 matrix multiplies */
2474 void pshader_glsl_texm3x3spec(SHADER_OPCODE_ARG
* arg
) {
2476 IWineD3DPixelShaderImpl
* shader
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2477 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2478 glsl_src_param_t src0_param
;
2479 glsl_src_param_t src1_param
;
2481 SHADER_BUFFER
* buffer
= arg
->buffer
;
2482 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2483 DWORD stype
= arg
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2484 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2485 glsl_sample_function_t sample_function
;
2487 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2488 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], src_mask
, &src1_param
);
2490 /* Perform the last matrix multiply operation */
2491 shader_addline(buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2492 /* Reflection calculation */
2493 shader_addline(buffer
, "tmp0.xyz = -reflect((%s), normalize(tmp0.xyz));\n", src1_param
.param_str
);
2495 shader_glsl_append_dst(buffer
, arg
);
2496 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2497 /* Dependent read, not valid with conditional NP2 */
2498 shader_glsl_get_sample_function(stype
, FALSE
, FALSE
, &sample_function
);
2500 /* Sample the texture */
2501 shader_addline(buffer
, "%s(Psampler%u, tmp0.xyz)%s);\n", sample_function
.name
, reg
, dst_mask
);
2503 current_state
->current_row
= 0;
2506 /** Process the WINED3DSIO_TEXM3X3VSPEC instruction in GLSL
2507 * Peform the final texture lookup based on the previous 2 3x3 matrix multiplies */
2508 void pshader_glsl_texm3x3vspec(SHADER_OPCODE_ARG
* arg
) {
2510 IWineD3DPixelShaderImpl
* shader
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2511 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2512 SHADER_BUFFER
* buffer
= arg
->buffer
;
2513 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2514 glsl_src_param_t src0_param
;
2516 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2517 DWORD sampler_type
= arg
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2518 glsl_sample_function_t sample_function
;
2520 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2522 /* Perform the last matrix multiply operation */
2523 shader_addline(buffer
, "tmp0.z = dot(vec3(T%u), vec3(%s));\n", reg
, src0_param
.param_str
);
2525 /* Construct the eye-ray vector from w coordinates */
2526 shader_addline(buffer
, "tmp1.xyz = normalize(vec3(gl_TexCoord[%u].w, gl_TexCoord[%u].w, gl_TexCoord[%u].w));\n",
2527 current_state
->texcoord_w
[0], current_state
->texcoord_w
[1], reg
);
2528 shader_addline(buffer
, "tmp0.xyz = -reflect(tmp1.xyz, normalize(tmp0.xyz));\n");
2530 shader_glsl_append_dst(buffer
, arg
);
2531 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2532 /* Dependent read, not valid with conditional NP2 */
2533 shader_glsl_get_sample_function(sampler_type
, FALSE
, FALSE
, &sample_function
);
2535 /* Sample the texture using the calculated coordinates */
2536 shader_addline(buffer
, "%s(Psampler%u, tmp0.xyz)%s);\n", sample_function
.name
, reg
, dst_mask
);
2538 current_state
->current_row
= 0;
2541 /** Process the WINED3DSIO_TEXBEM instruction in GLSL.
2542 * Apply a fake bump map transform.
2543 * texbem is pshader <= 1.3 only, this saves a few version checks
2545 void pshader_glsl_texbem(SHADER_OPCODE_ARG
* arg
) {
2546 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2547 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2548 char dst_swizzle
[6];
2549 glsl_sample_function_t sample_function
;
2550 glsl_src_param_t coord_param
;
2557 sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2558 flags
= deviceImpl
->stateBlock
->textureState
[sampler_idx
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
2560 sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2561 /* Dependent read, not valid with conditional NP2 */
2562 shader_glsl_get_sample_function(sampler_type
, FALSE
, FALSE
, &sample_function
);
2563 mask
= sample_function
.coord_mask
;
2565 shader_glsl_get_write_mask(arg
->dst
, dst_swizzle
);
2567 shader_glsl_get_write_mask(mask
, coord_mask
);
2569 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2570 * so we can't let the GL handle this.
2572 if (flags
& WINED3DTTFF_PROJECTED
) {
2574 char coord_div_mask
[3];
2575 switch (flags
& ~WINED3DTTFF_PROJECTED
) {
2576 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
2577 case WINED3DTTFF_COUNT2
: div_mask
= WINED3DSP_WRITEMASK_1
; break;
2578 case WINED3DTTFF_COUNT3
: div_mask
= WINED3DSP_WRITEMASK_2
; break;
2579 case WINED3DTTFF_COUNT4
:
2580 case WINED3DTTFF_DISABLE
: div_mask
= WINED3DSP_WRITEMASK_3
; break;
2582 shader_glsl_get_write_mask(div_mask
, coord_div_mask
);
2583 shader_addline(arg
->buffer
, "T%u%s /= T%u%s;\n", sampler_idx
, coord_mask
, sampler_idx
, coord_div_mask
);
2586 shader_glsl_append_dst(arg
->buffer
, arg
);
2587 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
|WINED3DSP_WRITEMASK_1
, &coord_param
);
2588 if(arg
->opcode
->opcode
== WINED3DSIO_TEXBEML
) {
2589 glsl_src_param_t luminance_param
;
2590 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_2
, &luminance_param
);
2591 shader_addline(arg
->buffer
, "(%s(Psampler%u, T%u%s + vec4(bumpenvmat%d * %s, 0.0, 0.0)%s )*(%s * luminancescale%d + luminanceoffset%d))%s);\n",
2592 sample_function
.name
, sampler_idx
, sampler_idx
, coord_mask
, sampler_idx
, coord_param
.param_str
, coord_mask
,
2593 luminance_param
.param_str
, sampler_idx
, sampler_idx
, dst_swizzle
);
2595 shader_addline(arg
->buffer
, "%s(Psampler%u, T%u%s + vec4(bumpenvmat%d * %s, 0.0, 0.0)%s )%s);\n",
2596 sample_function
.name
, sampler_idx
, sampler_idx
, coord_mask
, sampler_idx
, coord_param
.param_str
, coord_mask
, dst_swizzle
);
2600 void pshader_glsl_bem(SHADER_OPCODE_ARG
* arg
) {
2601 glsl_src_param_t src0_param
, src1_param
;
2602 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2604 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
|WINED3DSP_WRITEMASK_1
, &src0_param
);
2605 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
|WINED3DSP_WRITEMASK_1
, &src1_param
);
2607 shader_glsl_append_dst(arg
->buffer
, arg
);
2608 shader_addline(arg
->buffer
, "%s + bumpenvmat%d * %s);\n",
2609 src0_param
.param_str
, sampler_idx
, src1_param
.param_str
);
2612 /** Process the WINED3DSIO_TEXREG2AR instruction in GLSL
2613 * Sample 2D texture at dst using the alpha & red (wx) components of src as texture coordinates */
2614 void pshader_glsl_texreg2ar(SHADER_OPCODE_ARG
* arg
) {
2616 glsl_src_param_t src0_param
;
2617 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2620 shader_glsl_append_dst(arg
->buffer
, arg
);
2621 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2622 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
2624 shader_addline(arg
->buffer
, "texture2D(Psampler%u, %s.wx)%s);\n", sampler_idx
, src0_param
.reg_name
, dst_mask
);
2627 /** Process the WINED3DSIO_TEXREG2GB instruction in GLSL
2628 * Sample 2D texture at dst using the green & blue (yz) components of src as texture coordinates */
2629 void pshader_glsl_texreg2gb(SHADER_OPCODE_ARG
* arg
) {
2630 glsl_src_param_t src0_param
;
2631 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2634 shader_glsl_append_dst(arg
->buffer
, arg
);
2635 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2636 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
2638 shader_addline(arg
->buffer
, "texture2D(Psampler%u, %s.yz)%s);\n", sampler_idx
, src0_param
.reg_name
, dst_mask
);
2641 /** Process the WINED3DSIO_TEXREG2RGB instruction in GLSL
2642 * Sample texture at dst using the rgb (xyz) components of src as texture coordinates */
2643 void pshader_glsl_texreg2rgb(SHADER_OPCODE_ARG
* arg
) {
2644 glsl_src_param_t src0_param
;
2646 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2647 DWORD sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2648 glsl_sample_function_t sample_function
;
2650 shader_glsl_append_dst(arg
->buffer
, arg
);
2651 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2652 /* Dependent read, not valid with conditional NP2 */
2653 shader_glsl_get_sample_function(sampler_type
, FALSE
, FALSE
, &sample_function
);
2654 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], sample_function
.coord_mask
, &src0_param
);
2656 shader_addline(arg
->buffer
, "%s(Psampler%u, %s)%s);\n", sample_function
.name
, sampler_idx
, src0_param
.param_str
, dst_mask
);
2659 /** Process the WINED3DSIO_TEXKILL instruction in GLSL.
2660 * If any of the first 3 components are < 0, discard this pixel */
2661 void pshader_glsl_texkill(SHADER_OPCODE_ARG
* arg
) {
2662 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2663 DWORD hex_version
= This
->baseShader
.hex_version
;
2664 glsl_dst_param_t dst_param
;
2666 /* The argument is a destination parameter, and no writemasks are allowed */
2667 shader_glsl_add_dst_param(arg
, arg
->dst
, 0, &dst_param
);
2668 if((hex_version
>= WINED3DPS_VERSION(2,0))) {
2669 /* 2.0 shaders compare all 4 components in texkill */
2670 shader_addline(arg
->buffer
, "if (any(lessThan(%s.xyzw, vec4(0.0)))) discard;\n", dst_param
.reg_name
);
2672 /* 1.X shaders only compare the first 3 components, probably due to the nature of the texkill
2673 * instruction as a tex* instruction, and phase, which kills all a / w components. Even if all
2674 * 4 components are defined, only the first 3 are used
2676 shader_addline(arg
->buffer
, "if (any(lessThan(%s.xyz, vec3(0.0)))) discard;\n", dst_param
.reg_name
);
2680 /** Process the WINED3DSIO_DP2ADD instruction in GLSL.
2681 * dst = dot2(src0, src1) + src2 */
2682 void pshader_glsl_dp2add(SHADER_OPCODE_ARG
* arg
) {
2683 glsl_src_param_t src0_param
;
2684 glsl_src_param_t src1_param
;
2685 glsl_src_param_t src2_param
;
2687 unsigned int mask_size
;
2689 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
2690 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2692 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src0_param
);
2693 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src1_param
);
2694 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], WINED3DSP_WRITEMASK_0
, &src2_param
);
2696 shader_addline(arg
->buffer
, "dot(%s, %s) + %s);\n", src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2699 void pshader_glsl_input_pack(
2700 SHADER_BUFFER
* buffer
,
2701 semantic
* semantics_in
,
2702 IWineD3DPixelShader
*iface
) {
2705 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*) iface
;
2707 for (i
= 0; i
< MAX_REG_INPUT
; i
++) {
2709 DWORD usage_token
= semantics_in
[i
].usage
;
2710 DWORD register_token
= semantics_in
[i
].reg
;
2711 DWORD usage
, usage_idx
;
2715 if (!usage_token
) continue;
2716 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
2717 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
2718 shader_glsl_get_write_mask(register_token
, reg_mask
);
2722 case WINED3DDECLUSAGE_TEXCOORD
:
2723 if(usage_idx
< 8 && This
->vertexprocessing
== pretransformed
) {
2724 shader_addline(buffer
, "IN[%u]%s = gl_TexCoord[%u]%s;\n",
2725 This
->input_reg_map
[i
], reg_mask
, usage_idx
, reg_mask
);
2727 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2728 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2732 case WINED3DDECLUSAGE_COLOR
:
2734 shader_addline(buffer
, "IN[%u]%s = vec4(gl_Color)%s;\n",
2735 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2736 else if (usage_idx
== 1)
2737 shader_addline(buffer
, "IN[%u]%s = vec4(gl_SecondaryColor)%s;\n",
2738 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2740 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2741 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2745 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2746 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2751 /*********************************************
2752 * Vertex Shader Specific Code begins here
2753 ********************************************/
2755 static void add_glsl_program_entry(IWineD3DDeviceImpl
*device
, struct glsl_shader_prog_link
*entry
) {
2756 glsl_program_key_t
*key
;
2758 key
= HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t
));
2759 key
->vshader
= entry
->vshader
;
2760 key
->pshader
= entry
->pshader
;
2762 hash_table_put(device
->glsl_program_lookup
, key
, entry
);
2765 static struct glsl_shader_prog_link
*get_glsl_program_entry(IWineD3DDeviceImpl
*device
,
2766 GLhandleARB vshader
, GLhandleARB pshader
) {
2767 glsl_program_key_t key
;
2769 key
.vshader
= vshader
;
2770 key
.pshader
= pshader
;
2772 return (struct glsl_shader_prog_link
*)hash_table_get(device
->glsl_program_lookup
, &key
);
2775 void delete_glsl_program_entry(IWineD3DDevice
*iface
, struct glsl_shader_prog_link
*entry
) {
2776 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
2777 WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
2778 glsl_program_key_t
*key
;
2780 key
= HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t
));
2781 key
->vshader
= entry
->vshader
;
2782 key
->pshader
= entry
->pshader
;
2783 hash_table_remove(This
->glsl_program_lookup
, key
);
2785 GL_EXTCALL(glDeleteObjectARB(entry
->programId
));
2786 if (entry
->vshader
) list_remove(&entry
->vshader_entry
);
2787 if (entry
->pshader
) list_remove(&entry
->pshader_entry
);
2788 HeapFree(GetProcessHeap(), 0, entry
->vuniformF_locations
);
2789 HeapFree(GetProcessHeap(), 0, entry
->puniformF_locations
);
2790 HeapFree(GetProcessHeap(), 0, entry
);
2793 static void handle_ps3_input(SHADER_BUFFER
*buffer
, semantic
*semantics_in
, semantic
*semantics_out
, WineD3D_GL_Info
*gl_info
, DWORD
*map
) {
2795 DWORD usage_token
, usage_token_out
;
2796 DWORD register_token
, register_token_out
;
2797 DWORD usage
, usage_idx
, usage_out
, usage_idx_out
;
2799 char reg_mask
[6], reg_mask_out
[6];
2801 set
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*set
) * (GL_LIMITS(glsl_varyings
) / 4));
2803 for(i
= 0; i
< MAX_REG_INPUT
; i
++) {
2804 usage_token
= semantics_in
[i
].usage
;
2805 if (!usage_token
) continue;
2806 if(map
[i
] >= (GL_LIMITS(glsl_varyings
) / 4)) {
2807 FIXME("More input varyings declared than supported, expect issues\n");
2809 } else if(map
[i
] == -1) {
2810 /* Declared, but not read register */
2813 register_token
= semantics_in
[i
].reg
;
2815 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
2816 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
2817 set
[map
[i
]] = shader_glsl_get_write_mask(register_token
, reg_mask
);
2819 if(!semantics_out
) {
2821 case WINED3DDECLUSAGE_COLOR
:
2823 shader_addline(buffer
, "IN[%u]%s = gl_FrontColor%s;\n",
2824 map
[i
], reg_mask
, reg_mask
);
2825 else if (usage_idx
== 1)
2826 shader_addline(buffer
, "IN[%u]%s = gl_FrontSecondaryColor%s;\n",
2827 map
[i
], reg_mask
, reg_mask
);
2829 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2830 map
[i
], reg_mask
, reg_mask
);
2833 case WINED3DDECLUSAGE_TEXCOORD
:
2834 if (usage_idx
< 8) {
2835 shader_addline(buffer
, "IN[%u]%s = gl_TexCoord[%u]%s;\n",
2836 map
[i
], reg_mask
, usage_idx
, reg_mask
);
2838 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2839 map
[i
], reg_mask
, reg_mask
);
2843 case WINED3DDECLUSAGE_FOG
:
2844 shader_addline(buffer
, "IN[%u]%s = vec4(gl_FogFragCoord, 0.0, 0.0, 0.0)%s;\n",
2845 map
[i
], reg_mask
, reg_mask
);
2849 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2850 map
[i
], reg_mask
, reg_mask
);
2854 for(j
= 0; j
< MAX_REG_OUTPUT
; j
++) {
2855 usage_token_out
= semantics_out
[j
].usage
;
2856 if (!usage_token_out
) continue;
2857 register_token_out
= semantics_out
[j
].reg
;
2859 usage_out
= (usage_token_out
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
2860 usage_idx_out
= (usage_token_out
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
2861 shader_glsl_get_write_mask(register_token_out
, reg_mask_out
);
2863 if(usage
== usage_out
&&
2864 usage_idx
== usage_idx_out
) {
2865 shader_addline(buffer
, "IN[%u]%s = OUT[%u]%s;\n",
2866 map
[i
], reg_mask
, j
, reg_mask
);
2871 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2872 map
[i
], reg_mask
, reg_mask
);
2877 /* This is solely to make the compiler / linker happy and avoid warning about undefined
2878 * varyings. It shouldn't result in any real code executed on the GPU, since all read
2879 * input varyings are assigned above, if the optimizer works properly.
2881 for(i
= 0; i
< GL_LIMITS(glsl_varyings
) / 4; i
++) {
2882 if(set
[i
] != WINED3DSP_WRITEMASK_ALL
) {
2883 unsigned int size
= 0;
2884 memset(reg_mask
, 0, sizeof(reg_mask
));
2885 if(!(set
[i
] & WINED3DSP_WRITEMASK_0
)) {
2886 reg_mask
[size
] = 'x';
2889 if(!(set
[i
] & WINED3DSP_WRITEMASK_1
)) {
2890 reg_mask
[size
] = 'y';
2893 if(!(set
[i
] & WINED3DSP_WRITEMASK_2
)) {
2894 reg_mask
[size
] = 'z';
2897 if(!(set
[i
] & WINED3DSP_WRITEMASK_3
)) {
2898 reg_mask
[size
] = 'w';
2903 shader_addline(buffer
, "IN[%u].%s = 0.0;\n", i
, reg_mask
);
2906 shader_addline(buffer
, "IN[%u].%s = vec2(0.0, 0.0);\n", i
, reg_mask
);
2909 shader_addline(buffer
, "IN[%u].%s = vec3(0.0, 0.0, 0.0);\n", i
, reg_mask
);
2912 shader_addline(buffer
, "IN[%u].%s = vec4(0.0, 0.0, 0.0, 0.0);\n", i
, reg_mask
);
2918 HeapFree(GetProcessHeap(), 0, set
);
2921 static GLhandleARB
generate_param_reorder_function(IWineD3DVertexShader
*vertexshader
,
2922 IWineD3DPixelShader
*pixelshader
,
2923 WineD3D_GL_Info
*gl_info
) {
2924 GLhandleARB ret
= 0;
2925 IWineD3DVertexShaderImpl
*vs
= (IWineD3DVertexShaderImpl
*) vertexshader
;
2926 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) pixelshader
;
2927 DWORD vs_major
= vs
? WINED3DSHADER_VERSION_MAJOR(vs
->baseShader
.hex_version
) : 0;
2928 DWORD ps_major
= ps
? WINED3DSHADER_VERSION_MAJOR(ps
->baseShader
.hex_version
) : 0;
2930 SHADER_BUFFER buffer
;
2932 DWORD register_token
;
2933 DWORD usage
, usage_idx
, writemask
;
2935 semantic
*semantics_out
, *semantics_in
;
2937 buffer
.buffer
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, SHADER_PGMSIZE
);
2940 buffer
.newline
= TRUE
;
2942 if(vs_major
< 3 && ps_major
< 3) {
2943 /* That one is easy: The vertex shader writes to the builtin varyings, the pixel shader reads from them.
2944 * Take care about the texcoord .w fixup though if we're using the fixed function fragment pipeline
2946 if((GLINFO_LOCATION
).set_texcoord_w
&& ps_major
== 0 && vs_major
> 0) {
2947 shader_addline(&buffer
, "void order_ps_input() {\n");
2948 for(i
= 0; i
< min(8, MAX_REG_TEXCRD
); i
++) {
2949 if(vs
->baseShader
.reg_maps
.texcoord_mask
[i
] != 0 &&
2950 vs
->baseShader
.reg_maps
.texcoord_mask
[i
] != WINED3DSP_WRITEMASK_ALL
) {
2951 shader_addline(&buffer
, "gl_TexCoord[%u].w = 1.0;\n", i
);
2954 shader_addline(&buffer
, "}\n");
2956 shader_addline(&buffer
, "void order_ps_input() { /* do nothing */ }\n");
2958 } else if(ps_major
< 3 && vs_major
>= 3) {
2959 /* The vertex shader writes to its own varyings, the pixel shader needs them in the builtin ones */
2960 semantics_out
= vs
->semantics_out
;
2962 shader_addline(&buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
2963 for(i
= 0; i
< MAX_REG_OUTPUT
; i
++) {
2964 usage_token
= semantics_out
[i
].usage
;
2965 if (!usage_token
) continue;
2966 register_token
= semantics_out
[i
].reg
;
2968 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
2969 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
2970 writemask
= shader_glsl_get_write_mask(register_token
, reg_mask
);
2973 case WINED3DDECLUSAGE_COLOR
:
2975 shader_addline(&buffer
, "gl_FrontColor%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
2976 else if (usage_idx
== 1)
2977 shader_addline(&buffer
, "gl_FrontSecondaryColor%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
2980 case WINED3DDECLUSAGE_POSITION
:
2981 shader_addline(&buffer
, "gl_Position%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
2984 case WINED3DDECLUSAGE_TEXCOORD
:
2985 if (usage_idx
< 8) {
2986 if(!(GLINFO_LOCATION
).set_texcoord_w
|| ps_major
> 0) writemask
|= WINED3DSP_WRITEMASK_3
;
2988 shader_addline(&buffer
, "gl_TexCoord[%u]%s = OUT[%u]%s;\n",
2989 usage_idx
, reg_mask
, i
, reg_mask
);
2990 if(!(writemask
& WINED3DSP_WRITEMASK_3
)) {
2991 shader_addline(&buffer
, "gl_TexCoord[%u].w = 1.0;\n", usage_idx
);
2996 case WINED3DDECLUSAGE_PSIZE
:
2997 shader_addline(&buffer
, "gl_PointSize = OUT[%u].x;\n", i
);
3000 case WINED3DDECLUSAGE_FOG
:
3001 shader_addline(&buffer
, "gl_FogFragCoord = OUT[%u].%c;\n", i
, reg_mask
[1]);
3008 shader_addline(&buffer
, "}\n");
3010 } else if(ps_major
>= 3 && vs_major
>= 3) {
3011 semantics_out
= vs
->semantics_out
;
3012 semantics_in
= ps
->semantics_in
;
3014 /* This one is tricky: a 3.0 pixel shader reads from a 3.0 vertex shader */
3015 shader_addline(&buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
3016 shader_addline(&buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3018 /* First, sort out position and point size. Those are not passed to the pixel shader */
3019 for(i
= 0; i
< MAX_REG_OUTPUT
; i
++) {
3020 usage_token
= semantics_out
[i
].usage
;
3021 if (!usage_token
) continue;
3022 register_token
= semantics_out
[i
].reg
;
3024 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
3025 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
3026 shader_glsl_get_write_mask(register_token
, reg_mask
);
3029 case WINED3DDECLUSAGE_POSITION
:
3030 shader_addline(&buffer
, "gl_Position%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3033 case WINED3DDECLUSAGE_PSIZE
:
3034 shader_addline(&buffer
, "gl_PointSize = OUT[%u].x;\n", i
);
3042 /* Then, fix the pixel shader input */
3043 handle_ps3_input(&buffer
, semantics_in
, semantics_out
, gl_info
, ps
->input_reg_map
);
3045 shader_addline(&buffer
, "}\n");
3046 } else if(ps_major
>= 3 && vs_major
< 3) {
3047 semantics_in
= ps
->semantics_in
;
3049 shader_addline(&buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
3050 shader_addline(&buffer
, "void order_ps_input() {\n");
3051 /* The vertex shader wrote to the builtin varyings. There is no need to figure out position and
3052 * point size, but we depend on the optimizers kindness to find out that the pixel shader doesn't
3053 * read gl_TexCoord and gl_ColorX, otherwise we'll run out of varyings
3055 handle_ps3_input(&buffer
, semantics_in
, NULL
, gl_info
, ps
->input_reg_map
);
3056 shader_addline(&buffer
, "}\n");
3058 ERR("Unexpected vertex and pixel shader version condition: vs: %d, ps: %d\n", vs_major
, ps_major
);
3061 ret
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3062 checkGLcall("glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB)");
3063 GL_EXTCALL(glShaderSourceARB(ret
, 1, (const char**)&buffer
.buffer
, NULL
));
3064 checkGLcall("glShaderSourceARB(ret, 1, (const char**)&buffer.buffer, NULL)");
3065 GL_EXTCALL(glCompileShaderARB(ret
));
3066 checkGLcall("glCompileShaderARB(ret)");
3068 HeapFree(GetProcessHeap(), 0, buffer
.buffer
);
3072 /** Sets the GLSL program ID for the given pixel and vertex shader combination.
3073 * It sets the programId on the current StateBlock (because it should be called
3074 * inside of the DrawPrimitive() part of the render loop).
3076 * If a program for the given combination does not exist, create one, and store
3077 * the program in the hash table. If it creates a program, it will link the
3078 * given objects, too.
3080 static void set_glsl_shader_program(IWineD3DDevice
*iface
, BOOL use_ps
, BOOL use_vs
) {
3081 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3082 WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3083 IWineD3DPixelShader
*pshader
= This
->stateBlock
->pixelShader
;
3084 IWineD3DVertexShader
*vshader
= This
->stateBlock
->vertexShader
;
3085 struct glsl_shader_prog_link
*entry
= NULL
;
3086 GLhandleARB programId
= 0;
3087 GLhandleARB reorder_shader_id
= 0;
3091 GLhandleARB vshader_id
= use_vs
? ((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.prgId
: 0;
3092 GLhandleARB pshader_id
= use_ps
? ((IWineD3DBaseShaderImpl
*)pshader
)->baseShader
.prgId
: 0;
3093 entry
= get_glsl_program_entry(This
, vshader_id
, pshader_id
);
3095 This
->stateBlock
->glsl_program
= entry
;
3099 /* If we get to this point, then no matching program exists, so we create one */
3100 programId
= GL_EXTCALL(glCreateProgramObjectARB());
3101 TRACE("Created new GLSL shader program %u\n", programId
);
3103 /* Create the entry */
3104 entry
= HeapAlloc(GetProcessHeap(), 0, sizeof(struct glsl_shader_prog_link
));
3105 entry
->programId
= programId
;
3106 entry
->vshader
= vshader_id
;
3107 entry
->pshader
= pshader_id
;
3108 /* Add the hash table entry */
3109 add_glsl_program_entry(This
, entry
);
3111 /* Set the current program */
3112 This
->stateBlock
->glsl_program
= entry
;
3114 /* Attach GLSL vshader */
3116 int max_attribs
= 16; /* TODO: Will this always be the case? It is at the moment... */
3119 reorder_shader_id
= generate_param_reorder_function(vshader
, pshader
, gl_info
);
3120 TRACE("Attaching GLSL shader object %u to program %u\n", reorder_shader_id
, programId
);
3121 GL_EXTCALL(glAttachObjectARB(programId
, reorder_shader_id
));
3122 checkGLcall("glAttachObjectARB");
3123 /* Flag the reorder function for deletion, then it will be freed automatically when the program
3126 GL_EXTCALL(glDeleteObjectARB(reorder_shader_id
));
3128 TRACE("Attaching GLSL shader object %u to program %u\n", vshader_id
, programId
);
3129 GL_EXTCALL(glAttachObjectARB(programId
, vshader_id
));
3130 checkGLcall("glAttachObjectARB");
3132 /* Bind vertex attributes to a corresponding index number to match
3133 * the same index numbers as ARB_vertex_programs (makes loading
3134 * vertex attributes simpler). With this method, we can use the
3135 * exact same code to load the attributes later for both ARB and
3138 * We have to do this here because we need to know the Program ID
3139 * in order to make the bindings work, and it has to be done prior
3140 * to linking the GLSL program. */
3141 for (i
= 0; i
< max_attribs
; ++i
) {
3142 if (((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.reg_maps
.attributes
[i
]) {
3143 snprintf(tmp_name
, sizeof(tmp_name
), "attrib%i", i
);
3144 GL_EXTCALL(glBindAttribLocationARB(programId
, i
, tmp_name
));
3147 checkGLcall("glBindAttribLocationARB");
3149 list_add_head(&((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.linked_programs
, &entry
->vshader_entry
);
3152 /* Attach GLSL pshader */
3154 TRACE("Attaching GLSL shader object %u to program %u\n", pshader_id
, programId
);
3155 GL_EXTCALL(glAttachObjectARB(programId
, pshader_id
));
3156 checkGLcall("glAttachObjectARB");
3158 list_add_head(&((IWineD3DBaseShaderImpl
*)pshader
)->baseShader
.linked_programs
, &entry
->pshader_entry
);
3161 /* Link the program */
3162 TRACE("Linking GLSL shader program %u\n", programId
);
3163 GL_EXTCALL(glLinkProgramARB(programId
));
3164 print_glsl_info_log(&GLINFO_LOCATION
, programId
);
3166 entry
->vuniformF_locations
= HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB
) * GL_LIMITS(vshader_constantsF
));
3167 for (i
= 0; i
< GL_LIMITS(vshader_constantsF
); ++i
) {
3168 snprintf(glsl_name
, sizeof(glsl_name
), "VC[%i]", i
);
3169 entry
->vuniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3171 for (i
= 0; i
< MAX_CONST_I
; ++i
) {
3172 snprintf(glsl_name
, sizeof(glsl_name
), "VI[%i]", i
);
3173 entry
->vuniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3175 entry
->puniformF_locations
= HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB
) * GL_LIMITS(pshader_constantsF
));
3176 for (i
= 0; i
< GL_LIMITS(pshader_constantsF
); ++i
) {
3177 snprintf(glsl_name
, sizeof(glsl_name
), "PC[%i]", i
);
3178 entry
->puniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3180 for (i
= 0; i
< MAX_CONST_I
; ++i
) {
3181 snprintf(glsl_name
, sizeof(glsl_name
), "PI[%i]", i
);
3182 entry
->puniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3186 for(i
= 0; i
< ((IWineD3DPixelShaderImpl
*)pshader
)->numbumpenvmatconsts
; i
++) {
3188 sprintf(name
, "bumpenvmat%d", ((IWineD3DPixelShaderImpl
*)pshader
)->bumpenvmatconst
[i
].texunit
);
3189 entry
->bumpenvmat_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3190 sprintf(name
, "luminancescale%d", ((IWineD3DPixelShaderImpl
*)pshader
)->luminanceconst
[i
].texunit
);
3191 entry
->luminancescale_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3192 sprintf(name
, "luminanceoffset%d", ((IWineD3DPixelShaderImpl
*)pshader
)->luminanceconst
[i
].texunit
);
3193 entry
->luminanceoffset_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3198 entry
->posFixup_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "posFixup"));
3199 entry
->srgb_comparison_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "srgb_comparison"));
3200 entry
->srgb_mul_low_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "srgb_mul_low"));
3201 entry
->ycorrection_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "ycorrection"));
3202 checkGLcall("Find glsl program uniform locations");
3204 /* Set the shader to allow uniform loading on it */
3205 GL_EXTCALL(glUseProgramObjectARB(programId
));
3206 checkGLcall("glUseProgramObjectARB(programId)");
3208 /* Load the vertex and pixel samplers now. The function that finds the mappings makes sure
3209 * that it stays the same for each vertexshader-pixelshader pair(=linked glsl program). If
3210 * a pshader with fixed function pipeline is used there are no vertex samplers, and if a
3211 * vertex shader with fixed function pixel processing is used we make sure that the card
3212 * supports enough samplers to allow the max number of vertex samplers with all possible
3213 * fixed function fragment processing setups. So once the program is linked these samplers
3217 /* Load vertex shader samplers */
3218 shader_glsl_load_vsamplers(gl_info
, (IWineD3DStateBlock
*)This
->stateBlock
, programId
);
3221 /* Load pixel shader samplers */
3222 shader_glsl_load_psamplers(gl_info
, (IWineD3DStateBlock
*)This
->stateBlock
, programId
);
3226 static GLhandleARB
create_glsl_blt_shader(WineD3D_GL_Info
*gl_info
) {
3227 GLhandleARB program_id
;
3228 GLhandleARB vshader_id
, pshader_id
;
3229 const char *blt_vshader
[] = {
3232 " gl_Position = gl_Vertex;\n"
3233 " gl_FrontColor = vec4(1.0);\n"
3234 " gl_TexCoord[0].x = (gl_Vertex.x * 0.5) + 0.5;\n"
3235 " gl_TexCoord[0].y = (-gl_Vertex.y * 0.5) + 0.5;\n"
3239 const char *blt_pshader
[] = {
3240 "uniform sampler2D sampler;\n"
3243 " gl_FragDepth = texture2D(sampler, gl_TexCoord[0].xy).x;\n"
3247 vshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3248 GL_EXTCALL(glShaderSourceARB(vshader_id
, 1, blt_vshader
, NULL
));
3249 GL_EXTCALL(glCompileShaderARB(vshader_id
));
3251 pshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
3252 GL_EXTCALL(glShaderSourceARB(pshader_id
, 1, blt_pshader
, NULL
));
3253 GL_EXTCALL(glCompileShaderARB(pshader_id
));
3255 program_id
= GL_EXTCALL(glCreateProgramObjectARB());
3256 GL_EXTCALL(glAttachObjectARB(program_id
, vshader_id
));
3257 GL_EXTCALL(glAttachObjectARB(program_id
, pshader_id
));
3258 GL_EXTCALL(glLinkProgramARB(program_id
));
3260 print_glsl_info_log(&GLINFO_LOCATION
, program_id
);
3262 /* Once linked we can mark the shaders for deletion. They will be deleted once the program
3265 GL_EXTCALL(glDeleteObjectARB(vshader_id
));
3266 GL_EXTCALL(glDeleteObjectARB(pshader_id
));
3270 static void shader_glsl_select(IWineD3DDevice
*iface
, BOOL usePS
, BOOL useVS
) {
3271 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3272 WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3273 GLhandleARB program_id
= 0;
3275 if (useVS
|| usePS
) set_glsl_shader_program(iface
, usePS
, useVS
);
3276 else This
->stateBlock
->glsl_program
= NULL
;
3278 program_id
= This
->stateBlock
->glsl_program
? This
->stateBlock
->glsl_program
->programId
: 0;
3279 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
3280 GL_EXTCALL(glUseProgramObjectARB(program_id
));
3281 checkGLcall("glUseProgramObjectARB");
3284 static void shader_glsl_select_depth_blt(IWineD3DDevice
*iface
) {
3285 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3286 WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3287 struct shader_glsl_priv
*priv
= (struct shader_glsl_priv
*) This
->shader_priv
;
3288 static GLhandleARB loc
= -1;
3290 if (!priv
->depth_blt_glsl_program_id
) {
3291 priv
->depth_blt_glsl_program_id
= create_glsl_blt_shader(gl_info
);
3292 loc
= GL_EXTCALL(glGetUniformLocationARB(priv
->depth_blt_glsl_program_id
, "sampler"));
3295 GL_EXTCALL(glUseProgramObjectARB(priv
->depth_blt_glsl_program_id
));
3296 GL_EXTCALL(glUniform1iARB(loc
, 0));
3299 static void shader_glsl_destroy_depth_blt(IWineD3DDevice
*iface
) {
3300 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3301 struct shader_glsl_priv
*priv
= (struct shader_glsl_priv
*) This
->shader_priv
;
3302 WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3304 if(priv
->depth_blt_glsl_program_id
) {
3305 GL_EXTCALL(glDeleteObjectARB(priv
->depth_blt_glsl_program_id
));
3306 priv
->depth_blt_glsl_program_id
= 0;
3310 static void shader_glsl_cleanup(IWineD3DDevice
*iface
) {
3311 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3312 WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3313 GL_EXTCALL(glUseProgramObjectARB(0));
3316 static void shader_glsl_destroy(IWineD3DBaseShader
*iface
) {
3317 struct list
*linked_programs
;
3318 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*) iface
;
3319 WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*) This
->baseShader
.device
)->adapter
->gl_info
;
3321 /* Note: Do not use QueryInterface here to find out which shader type this is because this code
3322 * can be called from IWineD3DBaseShader::Release
3324 char pshader
= shader_is_pshader_version(This
->baseShader
.hex_version
);
3326 if(This
->baseShader
.prgId
== 0) return;
3327 linked_programs
= &This
->baseShader
.linked_programs
;
3329 TRACE("Deleting linked programs\n");
3330 if (linked_programs
->next
) {
3331 struct glsl_shader_prog_link
*entry
, *entry2
;
3334 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, pshader_entry
) {
3335 delete_glsl_program_entry(This
->baseShader
.device
, entry
);
3338 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, vshader_entry
) {
3339 delete_glsl_program_entry(This
->baseShader
.device
, entry
);
3344 TRACE("Deleting shader object %u\n", This
->baseShader
.prgId
);
3345 GL_EXTCALL(glDeleteObjectARB(This
->baseShader
.prgId
));
3346 checkGLcall("glDeleteObjectARB");
3347 This
->baseShader
.prgId
= 0;
3348 This
->baseShader
.is_compiled
= FALSE
;
3351 static unsigned int glsl_program_key_hash(void *key
) {
3352 glsl_program_key_t
*k
= (glsl_program_key_t
*)key
;
3354 unsigned int hash
= k
->vshader
| k
->pshader
<< 16;
3355 hash
+= ~(hash
<< 15);
3356 hash
^= (hash
>> 10);
3357 hash
+= (hash
<< 3);
3358 hash
^= (hash
>> 6);
3359 hash
+= ~(hash
<< 11);
3360 hash
^= (hash
>> 16);
3365 static BOOL
glsl_program_key_compare(void *keya
, void *keyb
) {
3366 glsl_program_key_t
*ka
= (glsl_program_key_t
*)keya
;
3367 glsl_program_key_t
*kb
= (glsl_program_key_t
*)keyb
;
3369 return ka
->vshader
== kb
->vshader
&& ka
->pshader
== kb
->pshader
;
3372 static HRESULT
shader_glsl_alloc(IWineD3DDevice
*iface
) {
3373 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3374 This
->shader_priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct shader_glsl_priv
));
3375 This
->glsl_program_lookup
= hash_table_create(&glsl_program_key_hash
, &glsl_program_key_compare
);
3379 static void shader_glsl_free(IWineD3DDevice
*iface
) {
3380 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3381 HeapFree(GetProcessHeap(), 0, This
->shader_priv
);
3382 This
->shader_priv
= NULL
;
3385 static BOOL
shader_glsl_dirty_const(IWineD3DDevice
*iface
) {
3386 /* TODO: GL_EXT_bindable_uniform can be used to share constants across shaders */
3390 static void shader_glsl_generate_pshader(IWineD3DPixelShader
*iface
, SHADER_BUFFER
*buffer
) {
3391 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)iface
;
3392 shader_reg_maps
* reg_maps
= &This
->baseShader
.reg_maps
;
3393 CONST DWORD
*function
= This
->baseShader
.function
;
3394 const char *fragcolor
;
3395 WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->adapter
->gl_info
;
3397 /* Create the hw GLSL shader object and assign it as the baseShader.prgId */
3398 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
3400 if (GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
3401 shader_addline(buffer
, "#extension GL_ARB_draw_buffers : enable\n");
3403 if (GL_SUPPORT(ARB_TEXTURE_RECTANGLE
)) {
3404 /* The spec says that it doesn't have to be explicitly enabled, but the nvidia
3405 * drivers write a warning if we don't do so
3407 shader_addline(buffer
, "#extension GL_ARB_texture_rectangle : enable\n");
3410 /* Base Declarations */
3411 shader_generate_glsl_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
);
3413 /* Pack 3.0 inputs */
3414 if (This
->baseShader
.hex_version
>= WINED3DPS_VERSION(3,0)) {
3416 if(((IWineD3DDeviceImpl
*) This
->baseShader
.device
)->strided_streams
.u
.s
.position_transformed
) {
3417 This
->vertexprocessing
= pretransformed
;
3418 pshader_glsl_input_pack(buffer
, This
->semantics_in
, iface
);
3419 } else if(!use_vs((IWineD3DDeviceImpl
*) This
->baseShader
.device
)) {
3420 This
->vertexprocessing
= fixedfunction
;
3421 pshader_glsl_input_pack(buffer
, This
->semantics_in
, iface
);
3423 This
->vertexprocessing
= vertexshader
;
3427 /* Base Shader Body */
3428 shader_generate_main( (IWineD3DBaseShader
*) This
, buffer
, reg_maps
, function
);
3430 /* Pixel shaders < 2.0 place the resulting color in R0 implicitly */
3431 if (This
->baseShader
.hex_version
< WINED3DPS_VERSION(2,0)) {
3432 /* Some older cards like GeforceFX ones don't support multiple buffers, so also not gl_FragData */
3433 if(GL_SUPPORT(ARB_DRAW_BUFFERS
))
3434 shader_addline(buffer
, "gl_FragData[0] = R0;\n");
3436 shader_addline(buffer
, "gl_FragColor = R0;\n");
3439 if(GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
3440 fragcolor
= "gl_FragData[0]";
3442 fragcolor
= "gl_FragColor";
3444 if(This
->srgb_enabled
) {
3445 shader_addline(buffer
, "tmp0.xyz = pow(%s.xyz, vec3(%f, %f, %f)) * vec3(%f, %f, %f) - vec3(%f, %f, %f);\n",
3446 fragcolor
, srgb_pow
, srgb_pow
, srgb_pow
, srgb_mul_high
, srgb_mul_high
, srgb_mul_high
,
3447 srgb_sub_high
, srgb_sub_high
, srgb_sub_high
);
3448 shader_addline(buffer
, "tmp1.xyz = %s.xyz * srgb_mul_low.xyz;\n", fragcolor
);
3449 shader_addline(buffer
, "%s.x = %s.x < srgb_comparison.x ? tmp1.x : tmp0.x;\n", fragcolor
, fragcolor
);
3450 shader_addline(buffer
, "%s.y = %s.y < srgb_comparison.y ? tmp1.y : tmp0.y;\n", fragcolor
, fragcolor
);
3451 shader_addline(buffer
, "%s.z = %s.z < srgb_comparison.z ? tmp1.z : tmp0.z;\n", fragcolor
, fragcolor
);
3452 shader_addline(buffer
, "%s = clamp(%s, 0.0, 1.0);\n", fragcolor
, fragcolor
);
3454 /* Pixel shader < 3.0 do not replace the fog stage.
3455 * This implements linear fog computation and blending.
3456 * TODO: non linear fog
3457 * NOTE: gl_Fog.start and gl_Fog.end don't hold fog start s and end e but
3458 * -1/(e-s) and e/(e-s) respectively.
3460 if(This
->baseShader
.hex_version
< WINED3DPS_VERSION(3,0)) {
3461 shader_addline(buffer
, "float Fog = clamp(gl_FogFragCoord * gl_Fog.start + gl_Fog.end, 0.0, 1.0);\n");
3462 shader_addline(buffer
, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor
, fragcolor
);
3465 shader_addline(buffer
, "}\n");
3467 TRACE("Compiling shader object %u\n", shader_obj
);
3468 GL_EXTCALL(glShaderSourceARB(shader_obj
, 1, (const char**)&buffer
->buffer
, NULL
));
3469 GL_EXTCALL(glCompileShaderARB(shader_obj
));
3470 print_glsl_info_log(&GLINFO_LOCATION
, shader_obj
);
3472 /* Store the shader object */
3473 This
->baseShader
.prgId
= shader_obj
;
3476 static void shader_glsl_generate_vshader(IWineD3DVertexShader
*iface
, SHADER_BUFFER
*buffer
) {
3477 IWineD3DVertexShaderImpl
*This
= (IWineD3DVertexShaderImpl
*)iface
;
3478 shader_reg_maps
* reg_maps
= &This
->baseShader
.reg_maps
;
3479 CONST DWORD
*function
= This
->baseShader
.function
;
3480 WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->adapter
->gl_info
;
3482 /* Create the hw GLSL shader program and assign it as the baseShader.prgId */
3483 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3485 /* Base Declarations */
3486 shader_generate_glsl_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
);
3488 /* Base Shader Body */
3489 shader_generate_main( (IWineD3DBaseShader
*) This
, buffer
, reg_maps
, function
);
3491 /* Unpack 3.0 outputs */
3492 if (This
->baseShader
.hex_version
>= WINED3DVS_VERSION(3,0)) {
3493 shader_addline(buffer
, "order_ps_input(OUT);\n");
3495 shader_addline(buffer
, "order_ps_input();\n");
3498 /* If this shader doesn't use fog copy the z coord to the fog coord so that we can use table fog */
3500 shader_addline(buffer
, "gl_FogFragCoord = gl_Position.z;\n");
3502 /* Write the final position.
3504 * OpenGL coordinates specify the center of the pixel while d3d coords specify
3505 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
3506 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
3507 * contains 1.0 to allow a mad.
3509 shader_addline(buffer
, "gl_Position.y = gl_Position.y * posFixup.y;\n");
3510 shader_addline(buffer
, "gl_Position.xy += posFixup.zw * gl_Position.ww;\n");
3512 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
3514 * Basically we want (in homogeneous coordinates) z = z * 2 - 1. However, shaders are run
3515 * before the homogeneous divide, so we have to take the w into account: z = ((z / w) * 2 - 1) * w,
3516 * which is the same as z = z / 2 - w.
3518 shader_addline(buffer
, "gl_Position.z = gl_Position.z * 2.0 - gl_Position.w;\n");
3520 shader_addline(buffer
, "}\n");
3522 TRACE("Compiling shader object %u\n", shader_obj
);
3523 GL_EXTCALL(glShaderSourceARB(shader_obj
, 1, (const char**)&buffer
->buffer
, NULL
));
3524 GL_EXTCALL(glCompileShaderARB(shader_obj
));
3525 print_glsl_info_log(&GLINFO_LOCATION
, shader_obj
);
3527 /* Store the shader object */
3528 This
->baseShader
.prgId
= shader_obj
;
3531 static void shader_glsl_get_caps(WINED3DDEVTYPE devtype
, WineD3D_GL_Info
*gl_info
, struct shader_caps
*pCaps
) {
3532 /* We don't have a GLSL fixed function pipeline yet, so let the none backend set its caps,
3533 * then overwrite the shader specific ones
3535 none_shader_backend
.shader_get_caps(devtype
, gl_info
, pCaps
);
3537 /* Nvidia Geforce6/7 or Ati R4xx/R5xx cards with GLSL support, support VS 3.0 but older Nvidia/Ati
3538 * models with GLSL support only support 2.0. In case of nvidia we can detect VS 2.0 support using
3539 * vs_nv_version which is based on NV_vertex_program.
3540 * For Ati cards there's no way using glsl (it abstracts the lowlevel info away) and also not
3541 * using ARB_vertex_program. It is safe to assume that when a card supports pixel shader 2.0 it
3542 * supports vertex shader 2.0 too and the way around. We can detect ps2.0 using the maximum number
3543 * of native instructions, so use that here. For more info see the pixel shader versioning code below.
3545 if((GLINFO_LOCATION
.vs_nv_version
== VS_VERSION_20
) || (GLINFO_LOCATION
.ps_arb_max_instructions
<= 512))
3546 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(2,0);
3548 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(3,0);
3549 TRACE_(d3d_caps
)("Hardware vertex shader version %d.%d enabled (GLSL)\n", (pCaps
->VertexShaderVersion
>> 8) & 0xff, pCaps
->VertexShaderVersion
& 0xff);
3550 pCaps
->MaxVertexShaderConst
= GL_LIMITS(vshader_constantsF
);
3552 /* Older DX9-class videocards (GeforceFX / Radeon >9500/X*00) only support pixel shader 2.0/2.0a/2.0b.
3553 * In OpenGL the extensions related to GLSL abstract lowlevel GL info away which is needed
3554 * to distinguish between 2.0 and 3.0 (and 2.0a/2.0b). In case of Nvidia we use their fragment
3555 * program extensions. On other hardware including ATI GL_ARB_fragment_program offers the info
3556 * in max native instructions. Intel and others also offer the info in this extension but they
3557 * don't support GLSL (at least on Windows).
3559 * PS2.0 requires at least 96 instructions, 2.0a/2.0b go up to 512. Assume that if the number
3560 * of instructions is 512 or less we have to do with ps2.0 hardware.
3561 * NOTE: ps3.0 hardware requires 512 or more instructions but ati and nvidia offer 'enough' (1024 vs 4096) on their most basic ps3.0 hardware.
3563 if((GLINFO_LOCATION
.ps_nv_version
== PS_VERSION_20
) || (GLINFO_LOCATION
.ps_arb_max_instructions
<= 512))
3564 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(2,0);
3566 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(3,0);
3568 /* FIXME: The following line is card dependent. -8.0 to 8.0 is the
3569 * Direct3D minimum requirement.
3571 * Both GL_ARB_fragment_program and GLSL require a "maximum representable magnitude"
3572 * of colors to be 2^10, and 2^32 for other floats. Should we use 1024 here?
3574 * The problem is that the refrast clamps temporary results in the shader to
3575 * [-MaxValue;+MaxValue]. If the card's max value is bigger than the one we advertize here,
3576 * then applications may miss the clamping behavior. On the other hand, if it is smaller,
3577 * the shader will generate incorrect results too. Unfortunately, GL deliberately doesn't
3578 * offer a way to query this.
3580 pCaps
->PixelShader1xMaxValue
= 8.0;
3581 TRACE_(d3d_caps
)("Hardware pixel shader version %d.%d enabled (GLSL)\n", (pCaps
->PixelShaderVersion
>> 8) & 0xff, pCaps
->PixelShaderVersion
& 0xff);
3584 static void shader_glsl_load_init(void) {}
3586 const shader_backend_t glsl_shader_backend
= {
3587 &shader_glsl_select
,
3588 &shader_glsl_select_depth_blt
,
3589 &shader_glsl_destroy_depth_blt
,
3590 &shader_glsl_load_constants
,
3591 &shader_glsl_cleanup
,
3592 &shader_glsl_color_correction
,
3593 &shader_glsl_destroy
,
3596 &shader_glsl_dirty_const
,
3597 &shader_glsl_generate_pshader
,
3598 &shader_glsl_generate_vshader
,
3599 &shader_glsl_get_caps
,
3600 &shader_glsl_load_init
,