wined3d: Only call activate_dimensions() if the texture is used.
[wine] / dlls / wined3d / pixelshader.c
1 /*
2  * shaders implementation
3  *
4  * Copyright 2002-2003 Jason Edmeades
5  * Copyright 2002-2003 Raphael Junqueira
6  * Copyright 2004 Christian Costa
7  * Copyright 2005 Oliver Stieber
8  * Copyright 2006 Ivan Gyurdiev
9  *
10  * This library is free software; you can redistribute it and/or
11  * modify it under the terms of the GNU Lesser General Public
12  * License as published by the Free Software Foundation; either
13  * version 2.1 of the License, or (at your option) any later version.
14  *
15  * This library is distributed in the hope that it will be useful,
16  * but WITHOUT ANY WARRANTY; without even the implied warranty of
17  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
18  * Lesser General Public License for more details.
19  *
20  * You should have received a copy of the GNU Lesser General Public
21  * License along with this library; if not, write to the Free Software
22  * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
23  */
24
25 #include "config.h"
26
27 #include <math.h>
28 #include <stdio.h>
29
30 #include "wined3d_private.h"
31
32 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader);
33
34 #define GLINFO_LOCATION ((IWineD3DDeviceImpl *) This->baseShader.device)->adapter->gl_info
35
36 #if 0 /* Must not be 1 in cvs version */
37 # define PSTRACE(A) TRACE A
38 # define TRACE_VSVECTOR(name) TRACE( #name "=(%f, %f, %f, %f)\n", name.x, name.y, name.z, name.w)
39 #else
40 # define PSTRACE(A)
41 # define TRACE_VSVECTOR(name)
42 #endif
43
44 #define GLNAME_REQUIRE_GLSL  ((const char *)1)
45 /* *******************************************
46    IWineD3DPixelShader IUnknown parts follow
47    ******************************************* */
48 static HRESULT  WINAPI IWineD3DPixelShaderImpl_QueryInterface(IWineD3DPixelShader *iface, REFIID riid, LPVOID *ppobj)
49 {
50     IWineD3DPixelShaderImpl *This = (IWineD3DPixelShaderImpl *)iface;
51     TRACE("(%p)->(%s,%p)\n",This,debugstr_guid(riid),ppobj);
52     if (IsEqualGUID(riid, &IID_IUnknown)
53         || IsEqualGUID(riid, &IID_IWineD3DBase)
54         || IsEqualGUID(riid, &IID_IWineD3DBaseShader)
55         || IsEqualGUID(riid, &IID_IWineD3DPixelShader)) {
56         IUnknown_AddRef(iface);
57         *ppobj = This;
58         return S_OK;
59     }
60     *ppobj = NULL;
61     return E_NOINTERFACE;
62 }
63
64 static ULONG  WINAPI IWineD3DPixelShaderImpl_AddRef(IWineD3DPixelShader *iface) {
65     IWineD3DPixelShaderImpl *This = (IWineD3DPixelShaderImpl *)iface;
66     TRACE("(%p) : AddRef increasing from %d\n", This, This->ref);
67     return InterlockedIncrement(&This->ref);
68 }
69
70 static ULONG  WINAPI IWineD3DPixelShaderImpl_Release(IWineD3DPixelShader *iface) {
71     IWineD3DPixelShaderImpl *This = (IWineD3DPixelShaderImpl *)iface;
72     ULONG ref;
73     TRACE("(%p) : Releasing from %d\n", This, This->ref);
74     ref = InterlockedDecrement(&This->ref);
75     if (ref == 0) {
76         /* SetPixelShader does not AddRef. If the bound pixel shader is destroyed, the pointer in the stateblock remains
77          * unchanged. Drawing again will most likely crash, even on windows. A problem can occur if the application creates
78          * a new pixel shader which resides at the same address. Then SetPixelShader will think it is a NOP change, and won't
79          * dirtify the state.
80          *
81          * Do NOT call GetPixelShader here. This will addRef and cause a recursion. And do NOT set the pixel shader to NULL,
82          * Windows does not do that(Although no test exists since they'd crash randomly)
83          */
84         if(iface == ((IWineD3DDeviceImpl *) This->baseShader.device)->stateBlock->pixelShader) {
85             IWineD3DDeviceImpl_MarkStateDirty((IWineD3DDeviceImpl *) This->baseShader.device, STATE_PIXELSHADER);
86         }
87
88         if (This->baseShader.shader_mode == SHADER_GLSL && This->baseShader.prgId != 0) {
89             struct list *linked_programs = &This->baseShader.linked_programs;
90
91             TRACE("Deleting linked programs\n");
92             if (linked_programs->next) {
93                 struct glsl_shader_prog_link *entry, *entry2;
94                 LIST_FOR_EACH_ENTRY_SAFE(entry, entry2, linked_programs, struct glsl_shader_prog_link, pshader_entry) {
95                     delete_glsl_program_entry(This->baseShader.device, entry);
96                 }
97             }
98
99             TRACE("Deleting shader object %u\n", This->baseShader.prgId);
100             GL_EXTCALL(glDeleteObjectARB(This->baseShader.prgId));
101             checkGLcall("glDeleteObjectARB");
102         }
103         shader_delete_constant_list(&This->baseShader.constantsF);
104         shader_delete_constant_list(&This->baseShader.constantsB);
105         shader_delete_constant_list(&This->baseShader.constantsI);
106         HeapFree(GetProcessHeap(), 0, This);
107     }
108     return ref;
109 }
110
111 /* *******************************************
112    IWineD3DPixelShader IWineD3DPixelShader parts follow
113    ******************************************* */
114
115 static HRESULT  WINAPI IWineD3DPixelShaderImpl_GetParent(IWineD3DPixelShader *iface, IUnknown** parent){
116     IWineD3DPixelShaderImpl *This = (IWineD3DPixelShaderImpl *)iface;
117
118     *parent = This->parent;
119     IUnknown_AddRef(*parent);
120     TRACE("(%p) : returning %p\n", This, *parent);
121     return WINED3D_OK;
122 }
123
124 static HRESULT  WINAPI IWineD3DPixelShaderImpl_GetDevice(IWineD3DPixelShader* iface, IWineD3DDevice **pDevice){
125     IWineD3DPixelShaderImpl *This = (IWineD3DPixelShaderImpl *)iface;
126     IWineD3DDevice_AddRef(This->baseShader.device);
127     *pDevice = This->baseShader.device;
128     TRACE("(%p) returning %p\n", This, *pDevice);
129     return WINED3D_OK;
130 }
131
132
133 static HRESULT  WINAPI IWineD3DPixelShaderImpl_GetFunction(IWineD3DPixelShader* impl, VOID* pData, UINT* pSizeOfData) {
134   IWineD3DPixelShaderImpl *This = (IWineD3DPixelShaderImpl *)impl;
135   TRACE("(%p) : pData(%p), pSizeOfData(%p)\n", This, pData, pSizeOfData);
136
137   if (NULL == pData) {
138     *pSizeOfData = This->baseShader.functionLength;
139     return WINED3D_OK;
140   }
141   if (*pSizeOfData < This->baseShader.functionLength) {
142     /* MSDN claims (for d3d8 at least) that if *pSizeOfData is smaller
143      * than the required size we should write the required size and
144      * return D3DERR_MOREDATA. That's not actually true. */
145     return WINED3DERR_INVALIDCALL;
146   }
147   if (NULL == This->baseShader.function) { /* no function defined */
148     TRACE("(%p) : GetFunction no User Function defined using NULL to %p\n", This, pData);
149     (*(DWORD **) pData) = NULL;
150   } else {
151     if (This->baseShader.functionLength == 0) {
152
153     }
154     TRACE("(%p) : GetFunction copying to %p\n", This, pData);
155     memcpy(pData, This->baseShader.function, This->baseShader.functionLength);
156   }
157   return WINED3D_OK;
158 }
159
160 CONST SHADER_OPCODE IWineD3DPixelShaderImpl_shader_ins[] = {
161     /* Arithmethic */
162     {WINED3DSIO_NOP,  "nop", "NOP", 0, 0, pshader_hw_map2gl, NULL, 0, 0},
163     {WINED3DSIO_MOV,  "mov", "MOV", 1, 2, pshader_hw_map2gl, shader_glsl_mov, 0, 0},
164     {WINED3DSIO_ADD,  "add", "ADD", 1, 3, pshader_hw_map2gl, shader_glsl_arith, 0, 0},
165     {WINED3DSIO_SUB,  "sub", "SUB", 1, 3, pshader_hw_map2gl, shader_glsl_arith, 0, 0},
166     {WINED3DSIO_MAD,  "mad", "MAD", 1, 4, pshader_hw_map2gl, shader_glsl_mad, 0, 0},
167     {WINED3DSIO_MUL,  "mul", "MUL", 1, 3, pshader_hw_map2gl, shader_glsl_arith, 0, 0},
168     {WINED3DSIO_RCP,  "rcp", "RCP",  1, 2, pshader_hw_map2gl, shader_glsl_rcp, 0, 0},
169     {WINED3DSIO_RSQ,  "rsq",  "RSQ", 1, 2, pshader_hw_map2gl, shader_glsl_map2gl, 0, 0},
170     {WINED3DSIO_DP3,  "dp3",  "DP3", 1, 3, pshader_hw_map2gl, shader_glsl_dot, 0, 0},
171     {WINED3DSIO_DP4,  "dp4",  "DP4", 1, 3, pshader_hw_map2gl, shader_glsl_dot, 0, 0},
172     {WINED3DSIO_MIN,  "min",  "MIN", 1, 3, pshader_hw_map2gl, shader_glsl_map2gl, 0, 0},
173     {WINED3DSIO_MAX,  "max",  "MAX", 1, 3, pshader_hw_map2gl, shader_glsl_map2gl, 0, 0},
174     {WINED3DSIO_SLT,  "slt",  "SLT", 1, 3, pshader_hw_map2gl, shader_glsl_compare, 0, 0},
175     {WINED3DSIO_SGE,  "sge",  "SGE", 1, 3, pshader_hw_map2gl, shader_glsl_compare, 0, 0},
176     {WINED3DSIO_ABS,  "abs",  "ABS", 1, 2, pshader_hw_map2gl, shader_glsl_map2gl, 0, 0},
177     {WINED3DSIO_EXP,  "exp",  "EX2", 1, 2, pshader_hw_map2gl, shader_glsl_map2gl, 0, 0},
178     {WINED3DSIO_LOG,  "log",  "LG2", 1, 2, pshader_hw_map2gl, shader_glsl_map2gl, 0, 0},
179     {WINED3DSIO_EXPP, "expp", "EXP", 1, 2, pshader_hw_map2gl, shader_glsl_expp, 0, 0},
180     {WINED3DSIO_LOGP, "logp", "LOG", 1, 2, pshader_hw_map2gl, shader_glsl_map2gl, 0, 0},
181     {WINED3DSIO_DST,  "dst",  "DST", 1, 3, pshader_hw_map2gl, shader_glsl_dst, 0, 0},
182     {WINED3DSIO_LRP,  "lrp",  "LRP", 1, 4, pshader_hw_map2gl, shader_glsl_lrp, 0, 0},
183     {WINED3DSIO_FRC,  "frc",  "FRC", 1, 2, pshader_hw_map2gl, shader_glsl_map2gl, 0, 0},
184     {WINED3DSIO_CND,  "cnd",  NULL, 1, 4, pshader_hw_cnd, shader_glsl_cnd, WINED3DPS_VERSION(1,1), WINED3DPS_VERSION(1,4)},
185     {WINED3DSIO_CMP,  "cmp",  NULL, 1, 4, pshader_hw_cmp, shader_glsl_cmp, WINED3DPS_VERSION(1,2), WINED3DPS_VERSION(3,0)},
186     {WINED3DSIO_POW,  "pow",  "POW", 1, 3, NULL, shader_glsl_pow, 0, 0},
187     {WINED3DSIO_CRS,  "crs",  "XPS", 1, 3, NULL, shader_glsl_cross, 0, 0},
188     /* TODO: xyz normalise can be performed as VS_ARB using one temporary register,
189         DP3 tmp , vec, vec;
190         RSQ tmp, tmp.x;
191         MUL vec.xyz, vec, tmp;
192     but I think this is better because it accounts for w properly.
193         DP3 tmp , vec, vec;
194         RSQ tmp, tmp.x;
195         MUL vec, vec, tmp;
196     */
197     {WINED3DSIO_NRM,      "nrm",      NULL, 1, 2, NULL, shader_glsl_map2gl, 0, 0},
198     {WINED3DSIO_SINCOS,   "sincos",   NULL, 1, 4, NULL, shader_glsl_sincos, WINED3DPS_VERSION(2,0), WINED3DPS_VERSION(2,1)},
199     {WINED3DSIO_SINCOS,   "sincos",   NULL, 1, 2, NULL, shader_glsl_sincos, WINED3DPS_VERSION(3,0), -1},
200     /* TODO: dp2add can be made out of multiple instuctions */
201     {WINED3DSIO_DP2ADD,   "dp2add",   GLNAME_REQUIRE_GLSL,  1, 4, NULL, pshader_glsl_dp2add, WINED3DPS_VERSION(2,0), -1},
202     /* Matrix */
203     {WINED3DSIO_M4x4, "m4x4", "undefined", 1, 3, NULL, shader_glsl_mnxn, 0, 0},
204     {WINED3DSIO_M4x3, "m4x3", "undefined", 1, 3, NULL, shader_glsl_mnxn, 0, 0},
205     {WINED3DSIO_M3x4, "m3x4", "undefined", 1, 3, NULL, shader_glsl_mnxn, 0, 0},
206     {WINED3DSIO_M3x3, "m3x3", "undefined", 1, 3, NULL, shader_glsl_mnxn, 0, 0},
207     {WINED3DSIO_M3x2, "m3x2", "undefined", 1, 3, NULL, shader_glsl_mnxn, 0, 0},
208     /* Register declarations */
209     {WINED3DSIO_DCL,      "dcl",      NULL, 0, 2, NULL, NULL, 0, 0},
210     /* Flow control - requires GLSL or software shaders */
211     {WINED3DSIO_REP ,     "rep",      NULL, 0, 1, NULL, shader_glsl_rep,    WINED3DPS_VERSION(2,1), -1},
212     {WINED3DSIO_ENDREP,   "endrep",   NULL, 0, 0, NULL, shader_glsl_end,    WINED3DPS_VERSION(2,1), -1},
213     {WINED3DSIO_IF,       "if",       NULL, 0, 1, NULL, shader_glsl_if,     WINED3DPS_VERSION(2,1), -1},
214     {WINED3DSIO_IFC,      "ifc",      NULL, 0, 2, NULL, shader_glsl_ifc,    WINED3DPS_VERSION(2,1), -1},
215     {WINED3DSIO_ELSE,     "else",     NULL, 0, 0, NULL, shader_glsl_else,   WINED3DPS_VERSION(2,1), -1},
216     {WINED3DSIO_ENDIF,    "endif",    NULL, 0, 0, NULL, shader_glsl_end,    WINED3DPS_VERSION(2,1), -1},
217     {WINED3DSIO_BREAK,    "break",    NULL, 0, 0, NULL, shader_glsl_break,  WINED3DPS_VERSION(2,1), -1},
218     {WINED3DSIO_BREAKC,   "breakc",   NULL, 0, 2, NULL, shader_glsl_breakc, WINED3DPS_VERSION(2,1), -1},
219     {WINED3DSIO_BREAKP,   "breakp",   GLNAME_REQUIRE_GLSL, 0, 1, NULL, NULL, 0, 0},
220     {WINED3DSIO_CALL,     "call",     NULL, 0, 1, NULL, shader_glsl_call, WINED3DPS_VERSION(2,1), -1},
221     {WINED3DSIO_CALLNZ,   "callnz",   NULL, 0, 2, NULL, shader_glsl_callnz, WINED3DPS_VERSION(2,1), -1},
222     {WINED3DSIO_LOOP,     "loop",     NULL, 0, 2, NULL, shader_glsl_loop,   WINED3DPS_VERSION(3,0), -1},
223     {WINED3DSIO_RET,      "ret",      NULL, 0, 0, NULL, NULL,               WINED3DPS_VERSION(2,1), -1},
224     {WINED3DSIO_ENDLOOP,  "endloop",  NULL, 0, 0, NULL, shader_glsl_end,    WINED3DPS_VERSION(3,0), -1},
225     {WINED3DSIO_LABEL,    "label",    NULL, 0, 1, NULL, shader_glsl_label,  WINED3DPS_VERSION(2,1), -1},
226     /* Constant definitions */
227     {WINED3DSIO_DEF,      "def",      "undefined",         1, 5, NULL, NULL, 0, 0},
228     {WINED3DSIO_DEFB,     "defb",     GLNAME_REQUIRE_GLSL, 1, 2, NULL, NULL, 0, 0},
229     {WINED3DSIO_DEFI,     "defi",     GLNAME_REQUIRE_GLSL, 1, 5, NULL, NULL, 0, 0},
230     /* Texture */
231     {WINED3DSIO_TEXCOORD, "texcoord", "undefined", 1, 1, pshader_hw_texcoord, pshader_glsl_texcoord, 0, WINED3DPS_VERSION(1,3)},
232     {WINED3DSIO_TEXCOORD, "texcrd",   "undefined", 1, 2, pshader_hw_texcoord, pshader_glsl_texcoord, WINED3DPS_VERSION(1,4), WINED3DPS_VERSION(1,4)},
233     {WINED3DSIO_TEXKILL,  "texkill",  "KIL",       1, 1, pshader_hw_map2gl, pshader_glsl_texkill, WINED3DPS_VERSION(1,0), WINED3DPS_VERSION(3,0)},
234     {WINED3DSIO_TEX,      "tex",      "undefined", 1, 1, pshader_hw_tex, pshader_glsl_tex, 0, WINED3DPS_VERSION(1,3)},
235     {WINED3DSIO_TEX,      "texld",    "undefined", 1, 2, pshader_hw_tex, pshader_glsl_tex, WINED3DPS_VERSION(1,4), WINED3DPS_VERSION(1,4)},
236     {WINED3DSIO_TEX,      "texld",    "undefined", 1, 3, pshader_hw_tex, pshader_glsl_tex, WINED3DPS_VERSION(2,0), -1},
237     {WINED3DSIO_TEXBEM,   "texbem",   "undefined", 1, 2, pshader_hw_texbem, pshader_glsl_texbem, 0, WINED3DPS_VERSION(1,3)},
238     {WINED3DSIO_TEXBEML,  "texbeml",  GLNAME_REQUIRE_GLSL, 1, 2, NULL, NULL, WINED3DPS_VERSION(1,0), WINED3DPS_VERSION(1,3)},
239     {WINED3DSIO_TEXREG2AR,"texreg2ar","undefined", 1, 2, pshader_hw_texreg2ar, pshader_glsl_texreg2ar, WINED3DPS_VERSION(1,1), WINED3DPS_VERSION(1,3)},
240     {WINED3DSIO_TEXREG2GB,"texreg2gb","undefined", 1, 2, pshader_hw_texreg2gb, pshader_glsl_texreg2gb, WINED3DPS_VERSION(1,1), WINED3DPS_VERSION(1,3)},
241     {WINED3DSIO_TEXREG2RGB,   "texreg2rgb",   GLNAME_REQUIRE_GLSL, 1, 2, NULL, pshader_glsl_texreg2rgb, WINED3DPS_VERSION(1,2), WINED3DPS_VERSION(1,3)},
242     {WINED3DSIO_TEXM3x2PAD,   "texm3x2pad",   "undefined", 1, 2, pshader_hw_texm3x2pad, pshader_glsl_texm3x2pad, WINED3DPS_VERSION(1,0), WINED3DPS_VERSION(1,3)},
243     {WINED3DSIO_TEXM3x2TEX,   "texm3x2tex",   "undefined", 1, 2, pshader_hw_texm3x2tex, pshader_glsl_texm3x2tex, WINED3DPS_VERSION(1,0), WINED3DPS_VERSION(1,3)},
244     {WINED3DSIO_TEXM3x3PAD,   "texm3x3pad",   "undefined", 1, 2, pshader_hw_texm3x3pad, pshader_glsl_texm3x3pad, WINED3DPS_VERSION(1,0), WINED3DPS_VERSION(1,3)},
245     {WINED3DSIO_TEXM3x3DIFF,  "texm3x3diff",  GLNAME_REQUIRE_GLSL, 1, 2, NULL, NULL, WINED3DPS_VERSION(0,0), WINED3DPS_VERSION(0,0)},
246     {WINED3DSIO_TEXM3x3SPEC,  "texm3x3spec",  "undefined", 1, 3, pshader_hw_texm3x3spec, pshader_glsl_texm3x3spec, WINED3DPS_VERSION(1,0), WINED3DPS_VERSION(1,3)},
247     {WINED3DSIO_TEXM3x3VSPEC, "texm3x3vspec",  "undefined", 1, 2, pshader_hw_texm3x3vspec, pshader_glsl_texm3x3vspec, WINED3DPS_VERSION(1,0), WINED3DPS_VERSION(1,3)},
248     {WINED3DSIO_TEXM3x3TEX,   "texm3x3tex",   "undefined", 1, 2, pshader_hw_texm3x3tex, pshader_glsl_texm3x3tex, WINED3DPS_VERSION(1,0), WINED3DPS_VERSION(1,3)},
249     {WINED3DSIO_TEXDP3TEX,    "texdp3tex",    GLNAME_REQUIRE_GLSL, 1, 2, NULL, pshader_glsl_texdp3tex, WINED3DPS_VERSION(1,2), WINED3DPS_VERSION(1,3)},
250     {WINED3DSIO_TEXM3x2DEPTH, "texm3x2depth", GLNAME_REQUIRE_GLSL, 1, 2, NULL, pshader_glsl_texm3x2depth, WINED3DPS_VERSION(1,3), WINED3DPS_VERSION(1,3)},
251     {WINED3DSIO_TEXDP3,   "texdp3",   GLNAME_REQUIRE_GLSL, 1, 2, NULL, pshader_glsl_texdp3, WINED3DPS_VERSION(1,2), WINED3DPS_VERSION(1,3)},
252     {WINED3DSIO_TEXM3x3,  "texm3x3",  GLNAME_REQUIRE_GLSL, 1, 2, NULL, pshader_glsl_texm3x3, WINED3DPS_VERSION(1,2), WINED3DPS_VERSION(1,3)},
253     {WINED3DSIO_TEXDEPTH, "texdepth", GLNAME_REQUIRE_GLSL, 1, 1, NULL, pshader_glsl_texdepth, WINED3DPS_VERSION(1,4), WINED3DPS_VERSION(1,4)},
254     {WINED3DSIO_BEM,      "bem",      "undefined",         1, 3, pshader_hw_bem, pshader_glsl_bem, WINED3DPS_VERSION(1,4), WINED3DPS_VERSION(1,4)},
255     {WINED3DSIO_DSX,      "dsx",      GLNAME_REQUIRE_GLSL, 1, 2, NULL, NULL, 0, 0},
256     {WINED3DSIO_DSY,      "dsy",      GLNAME_REQUIRE_GLSL, 1, 2, NULL, NULL, 0, 0},
257     {WINED3DSIO_TEXLDD,   "texldd",   GLNAME_REQUIRE_GLSL, 1, 5, NULL, NULL, WINED3DPS_VERSION(2,1), -1},
258     {WINED3DSIO_SETP,     "setp",     GLNAME_REQUIRE_GLSL, 1, 3, NULL, NULL, 0, 0},
259     {WINED3DSIO_TEXLDL,   "texldl",   NULL, 1, 3, NULL, shader_glsl_texldl, WINED3DPS_VERSION(3,0), -1},
260     {WINED3DSIO_PHASE,    "phase",    GLNAME_REQUIRE_GLSL, 0, 0, NULL, NULL, 0, 0},
261     {0,               NULL,       NULL,   0, 0, NULL,            NULL, 0, 0}
262 };
263
264 static void pshader_set_limits(
265       IWineD3DPixelShaderImpl *This) { 
266
267       This->baseShader.limits.attributes = 0;
268       This->baseShader.limits.address = 0;
269       This->baseShader.limits.packed_output = 0;
270
271       switch (This->baseShader.hex_version) {
272           case WINED3DPS_VERSION(1,0):
273           case WINED3DPS_VERSION(1,1):
274           case WINED3DPS_VERSION(1,2):
275           case WINED3DPS_VERSION(1,3): 
276                    This->baseShader.limits.temporary = 2;
277                    This->baseShader.limits.constant_float = 8;
278                    This->baseShader.limits.constant_int = 0;
279                    This->baseShader.limits.constant_bool = 0;
280                    This->baseShader.limits.texcoord = 4;
281                    This->baseShader.limits.sampler = 4;
282                    This->baseShader.limits.packed_input = 0;
283                    This->baseShader.limits.label = 0;
284                    break;
285
286           case WINED3DPS_VERSION(1,4):
287                    This->baseShader.limits.temporary = 6;
288                    This->baseShader.limits.constant_float = 8;
289                    This->baseShader.limits.constant_int = 0;
290                    This->baseShader.limits.constant_bool = 0;
291                    This->baseShader.limits.texcoord = 6;
292                    This->baseShader.limits.sampler = 6;
293                    This->baseShader.limits.packed_input = 0;
294                    This->baseShader.limits.label = 0;
295                    break;
296                
297           /* FIXME: temporaries must match D3DPSHADERCAPS2_0.NumTemps */ 
298           case WINED3DPS_VERSION(2,0):
299                    This->baseShader.limits.temporary = 32;
300                    This->baseShader.limits.constant_float = 32;
301                    This->baseShader.limits.constant_int = 16;
302                    This->baseShader.limits.constant_bool = 16;
303                    This->baseShader.limits.texcoord = 8;
304                    This->baseShader.limits.sampler = 16;
305                    This->baseShader.limits.packed_input = 0;
306                    break;
307
308           case WINED3DPS_VERSION(2,1):
309                    This->baseShader.limits.temporary = 32;
310                    This->baseShader.limits.constant_float = 32;
311                    This->baseShader.limits.constant_int = 16;
312                    This->baseShader.limits.constant_bool = 16;
313                    This->baseShader.limits.texcoord = 8;
314                    This->baseShader.limits.sampler = 16;
315                    This->baseShader.limits.packed_input = 0;
316                    This->baseShader.limits.label = 16;
317                    break;
318
319           case WINED3DPS_VERSION(3,0):
320                    This->baseShader.limits.temporary = 32;
321                    This->baseShader.limits.constant_float = 224;
322                    This->baseShader.limits.constant_int = 16;
323                    This->baseShader.limits.constant_bool = 16;
324                    This->baseShader.limits.texcoord = 0;
325                    This->baseShader.limits.sampler = 16;
326                    This->baseShader.limits.packed_input = 12;
327                    This->baseShader.limits.label = 16; /* FIXME: 2048 */
328                    break;
329
330           default: This->baseShader.limits.temporary = 32;
331                    This->baseShader.limits.constant_float = 32;
332                    This->baseShader.limits.constant_int = 16;
333                    This->baseShader.limits.constant_bool = 16;
334                    This->baseShader.limits.texcoord = 8;
335                    This->baseShader.limits.sampler = 16;
336                    This->baseShader.limits.packed_input = 0;
337                    This->baseShader.limits.label = 0;
338                    FIXME("Unrecognized pixel shader version %#x\n", 
339                        This->baseShader.hex_version);
340       }
341 }
342
343 /** Generate a pixel shader string using either GL_FRAGMENT_PROGRAM_ARB
344     or GLSL and send it to the card */
345 static inline VOID IWineD3DPixelShaderImpl_GenerateShader(
346     IWineD3DPixelShader *iface,
347     shader_reg_maps* reg_maps,
348     CONST DWORD *pFunction) {
349
350     IWineD3DPixelShaderImpl *This = (IWineD3DPixelShaderImpl *)iface;
351     SHADER_BUFFER buffer;
352
353 #if 0 /* FIXME: Use the buffer that is held by the device, this is ok since fixups will be skipped for software shaders
354         it also requires entering a critical section but cuts down the runtime footprint of wined3d and any memory fragmentation that may occur... */
355     if (This->device->fixupVertexBufferSize < SHADER_PGMSIZE) {
356         HeapFree(GetProcessHeap(), 0, This->fixupVertexBuffer);
357         This->fixupVertexBuffer = HeapAlloc(GetProcessHeap() , 0, SHADER_PGMSIZE);
358         This->fixupVertexBufferSize = PGMSIZE;
359         This->fixupVertexBuffer[0] = 0;
360     }
361     buffer.buffer = This->device->fixupVertexBuffer;
362 #else
363     buffer.buffer = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, SHADER_PGMSIZE); 
364 #endif
365     buffer.bsize = 0;
366     buffer.lineNo = 0;
367     buffer.newline = TRUE;
368
369     if (This->baseShader.shader_mode == SHADER_GLSL) {
370
371         /* Create the hw GLSL shader object and assign it as the baseShader.prgId */
372         GLhandleARB shader_obj = GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB));
373
374         if (GL_SUPPORT(ARB_DRAW_BUFFERS)) {
375             shader_addline(&buffer, "#extension GL_ARB_draw_buffers : enable\n");
376         }
377
378         /* Base Declarations */
379         shader_generate_glsl_declarations( (IWineD3DBaseShader*) This, reg_maps, &buffer, &GLINFO_LOCATION);
380
381         /* Pack 3.0 inputs */
382         if (This->baseShader.hex_version >= WINED3DPS_VERSION(3,0))
383             pshader_glsl_input_pack(&buffer, This->semantics_in);
384
385         /* Base Shader Body */
386         shader_generate_main( (IWineD3DBaseShader*) This, &buffer, reg_maps, pFunction);
387
388         /* Pixel shaders < 2.0 place the resulting color in R0 implicitly */
389         if (This->baseShader.hex_version < WINED3DPS_VERSION(2,0)) {
390             /* Some older cards like GeforceFX ones don't support multiple buffers, so also not gl_FragData */
391             if(GL_SUPPORT(ARB_DRAW_BUFFERS))
392                 shader_addline(&buffer, "gl_FragData[0] = R0;\n");
393             else
394                 shader_addline(&buffer, "gl_FragColor = R0;\n");
395         }
396
397         /* Pixel shader < 3.0 do not replace the fog stage.
398          * This implements linear fog computation and blending.
399          * TODO: non linear fog
400          * NOTE: gl_Fog.start and gl_Fog.end don't hold fog start s and end e but
401          * -1/(e-s) and e/(e-s) respectively.
402          */
403         if(This->baseShader.hex_version < WINED3DPS_VERSION(3,0)) {
404             shader_addline(&buffer, "float Fog = clamp(gl_FogFragCoord * gl_Fog.start + gl_Fog.end, 0.0, 1.0);\n");
405             if(GL_SUPPORT(ARB_DRAW_BUFFERS))
406                 shader_addline(&buffer, "gl_FragData[0].xyz = mix(gl_Fog.color.xyz, gl_FragData[0].xyz, Fog);\n");
407             else
408                 shader_addline(&buffer, "gl_FragColor.xyz = mix(gl_Fog.color.xyz, gl_FragColor.xyz, Fog);\n");
409         }
410
411         shader_addline(&buffer, "}\n");
412
413         TRACE("Compiling shader object %u\n", shader_obj);
414         GL_EXTCALL(glShaderSourceARB(shader_obj, 1, (const char**)&buffer.buffer, NULL));
415         GL_EXTCALL(glCompileShaderARB(shader_obj));
416         print_glsl_info_log(&GLINFO_LOCATION, shader_obj);
417
418         /* Store the shader object */
419         This->baseShader.prgId = shader_obj;
420
421     } else if (This->baseShader.shader_mode == SHADER_ARB) {
422         /*  Create the hw ARB shader */
423         shader_addline(&buffer, "!!ARBfp1.0\n");
424
425         shader_addline(&buffer, "TEMP TMP;\n");     /* Used in matrix ops */
426         shader_addline(&buffer, "TEMP TMP2;\n");    /* Used in matrix ops */
427         shader_addline(&buffer, "TEMP TA;\n");      /* Used for modifiers */
428         shader_addline(&buffer, "TEMP TB;\n");      /* Used for modifiers */
429         shader_addline(&buffer, "TEMP TC;\n");      /* Used for modifiers */
430         shader_addline(&buffer, "PARAM coefdiv = { 0.5, 0.25, 0.125, 0.0625 };\n");
431         shader_addline(&buffer, "PARAM coefmul = { 2, 4, 8, 16 };\n");
432         shader_addline(&buffer, "PARAM one = { 1.0, 1.0, 1.0, 1.0 };\n");
433
434         /* Base Declarations */
435         shader_generate_arb_declarations( (IWineD3DBaseShader*) This, reg_maps, &buffer, &GLINFO_LOCATION);
436
437         /* We need two variables for fog blending */
438         shader_addline(&buffer, "TEMP TMP_FOG;\n");
439         if (This->baseShader.hex_version >= WINED3DPS_VERSION(2,0)) {
440             shader_addline(&buffer, "TEMP TMP_COLOR;\n");
441         }
442
443         /* Base Shader Body */
444         shader_generate_main( (IWineD3DBaseShader*) This, &buffer, reg_maps, pFunction);
445
446         /* calculate fog and blend it
447          * NOTE: state.fog.params.y and state.fog.params.z don't hold fog start s and end e but
448          * -1/(e-s) and e/(e-s) respectively.
449          */
450         shader_addline(&buffer, "MAD_SAT TMP_FOG, fragment.fogcoord, state.fog.params.y, state.fog.params.z;\n");
451         if (This->baseShader.hex_version < WINED3DPS_VERSION(2,0)) {
452             shader_addline(&buffer, "LRP result.color.rgb, TMP_FOG.x, R0, state.fog.color;\n");
453             shader_addline(&buffer, "MOV result.color.a, R0.a;\n");
454         } else {
455             shader_addline(&buffer, "LRP result.color.rgb, TMP_FOG.x, TMP_COLOR, state.fog.color;\n");
456             shader_addline(&buffer, "MOV result.color.a, TMP_COLOR.a;\n");
457         }
458
459         shader_addline(&buffer, "END\n"); 
460
461         /* TODO: change to resource.glObjectHandle or something like that */
462         GL_EXTCALL(glGenProgramsARB(1, &This->baseShader.prgId));
463
464         TRACE("Creating a hw pixel shader, prg=%d\n", This->baseShader.prgId);
465         GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, This->baseShader.prgId));
466
467         TRACE("Created hw pixel shader, prg=%d\n", This->baseShader.prgId);
468         /* Create the program and check for errors */
469         GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB,
470             buffer.bsize, buffer.buffer));
471
472         if (glGetError() == GL_INVALID_OPERATION) {
473             GLint errPos;
474             glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &errPos);
475             FIXME("HW PixelShader Error at position %d: %s\n",
476                   errPos, debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
477             This->baseShader.prgId = -1;
478         }
479     }
480
481     This->needsbumpmat = reg_maps->bumpmat;
482
483 #if 1 /* if were using the data buffer of device then we don't need to free it */
484   HeapFree(GetProcessHeap(), 0, buffer.buffer);
485 #endif
486 }
487
488 static HRESULT WINAPI IWineD3DPixelShaderImpl_SetFunction(IWineD3DPixelShader *iface, CONST DWORD *pFunction) {
489
490     IWineD3DPixelShaderImpl *This =(IWineD3DPixelShaderImpl *)iface;
491     IWineD3DDeviceImpl *deviceImpl = (IWineD3DDeviceImpl *) This->baseShader.device;
492
493     TRACE("(%p) : pFunction %p\n", iface, pFunction);
494
495     /* First pass: trace shader */
496     shader_trace_init((IWineD3DBaseShader*) This, pFunction);
497     pshader_set_limits(This);
498
499     /* Initialize immediate constant lists */
500     list_init(&This->baseShader.constantsF);
501     list_init(&This->baseShader.constantsB);
502     list_init(&This->baseShader.constantsI);
503
504     if (WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) > 1) {
505         shader_reg_maps *reg_maps = &This->baseShader.reg_maps;
506         HRESULT hr;
507
508         /* Second pass: figure out which registers are used, what the semantics are, etc.. */
509         memset(reg_maps, 0, sizeof(shader_reg_maps));
510         hr = shader_get_registers_used((IWineD3DBaseShader*) This, reg_maps,
511             This->semantics_in, NULL, pFunction, NULL);
512         if (FAILED(hr)) return hr;
513         /* FIXME: validate reg_maps against OpenGL */
514     }
515
516     This->baseShader.shader_mode = deviceImpl->ps_selected_mode;
517
518     TRACE("(%p) : Copying the function\n", This);
519     if (NULL != pFunction) {
520         void *function;
521
522         function = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, This->baseShader.functionLength);
523         if (!function) return E_OUTOFMEMORY;
524         memcpy(function, pFunction, This->baseShader.functionLength);
525         This->baseShader.function = function;
526     } else {
527         This->baseShader.function = NULL;
528     }
529
530     return WINED3D_OK;
531 }
532
533 static HRESULT WINAPI IWineD3DPixelShaderImpl_CompileShader(IWineD3DPixelShader *iface) {
534
535     IWineD3DPixelShaderImpl *This =(IWineD3DPixelShaderImpl *)iface;
536     IWineD3DDeviceImpl *deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
537     CONST DWORD *function = This->baseShader.function;
538
539     TRACE("(%p) : function %p\n", iface, function);
540
541     /* We're already compiled. */
542     if (This->baseShader.is_compiled) return WINED3D_OK;
543
544     /* We don't need to compile */
545     if (!function) {
546         This->baseShader.is_compiled = TRUE;
547         return WINED3D_OK;
548     }
549
550     if (WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) == 1) {
551         shader_reg_maps *reg_maps = &This->baseShader.reg_maps;
552         HRESULT hr;
553
554         /* Second pass: figure out which registers are used, what the semantics are, etc.. */
555         memset(reg_maps, 0, sizeof(shader_reg_maps));
556         hr = shader_get_registers_used((IWineD3DBaseShader*)This, reg_maps,
557             This->semantics_in, NULL, This->baseShader.function, deviceImpl->stateBlock);
558         if (FAILED(hr)) return hr;
559         /* FIXME: validate reg_maps against OpenGL */
560     }
561
562     /* Generate the HW shader */
563     TRACE("(%p) : Generating hardware program\n", This);
564     IWineD3DPixelShaderImpl_GenerateShader(iface, &This->baseShader.reg_maps, function);
565
566     This->baseShader.is_compiled = TRUE;
567
568     return WINED3D_OK;
569 }
570
571 const IWineD3DPixelShaderVtbl IWineD3DPixelShader_Vtbl =
572 {
573     /*** IUnknown methods ***/
574     IWineD3DPixelShaderImpl_QueryInterface,
575     IWineD3DPixelShaderImpl_AddRef,
576     IWineD3DPixelShaderImpl_Release,
577     /*** IWineD3DBase methods ***/
578     IWineD3DPixelShaderImpl_GetParent,
579     /*** IWineD3DBaseShader methods ***/
580     IWineD3DPixelShaderImpl_SetFunction,
581     IWineD3DPixelShaderImpl_CompileShader,
582     /*** IWineD3DPixelShader methods ***/
583     IWineD3DPixelShaderImpl_GetDevice,
584     IWineD3DPixelShaderImpl_GetFunction
585 };