vertexshader.c

这是一个开放源代码的与WINNT/WIN2K/WIN2003兼容的操作系统

static void vshader_set_input(
    IWineD3DVertexShaderImpl* This,
    unsigned int regnum,
    BYTE usage, BYTE usage_idx) {

    /* Fake usage: set reserved bit, usage, usage_idx */
    DWORD usage_token = (0x1 << 31) |
        (usage << WINED3DSP_DCL_USAGE_SHIFT) | (usage_idx << WINED3DSP_DCL_USAGEINDEX_SHIFT);

    /* Fake register; set reserved bit, regnum, type: input, wmask: all */
    DWORD reg_token = (0x1 << 31) |
        WINED3DSP_WRITEMASK_ALL | (WINED3DSPR_INPUT << WINED3DSP_REGTYPE_SHIFT) | regnum;

    This->semantics_in[regnum].usage = usage_token;
    This->semantics_in[regnum].reg = reg_token;
}

static BOOL match_usage(BYTE usage1, BYTE usage_idx1, BYTE usage2, BYTE usage_idx2) {
    if (usage_idx1 != usage_idx2) return FALSE;
    if (usage1 == usage2) return TRUE;
    if (usage1 == WINED3DDECLUSAGE_POSITION && usage2 == WINED3DDECLUSAGE_POSITIONT) return TRUE;
    if (usage2 == WINED3DDECLUSAGE_POSITION && usage1 == WINED3DDECLUSAGE_POSITIONT) return TRUE;

    return FALSE;
}

BOOL vshader_get_input(
    IWineD3DVertexShader* iface,
    BYTE usage_req, BYTE usage_idx_req,
    unsigned int* regnum) {

    IWineD3DVertexShaderImpl* This = (IWineD3DVertexShaderImpl*) iface;
    int i;

    for (i = 0; i < MAX_ATTRIBS; i++) {
        DWORD usage_token = This->semantics_in[i].usage;
        DWORD usage = (usage_token & WINED3DSP_DCL_USAGE_MASK) >> WINED3DSP_DCL_USAGE_SHIFT;
        DWORD usage_idx = (usage_token & WINED3DSP_DCL_USAGEINDEX_MASK) >> WINED3DSP_DCL_USAGEINDEX_SHIFT;

        if (usage_token && match_usage(usage, usage_idx, usage_req, usage_idx_req)) {
            *regnum = i;
            return TRUE;
        }
    }
    return FALSE;
}

BOOL vshader_input_is_color(
    IWineD3DVertexShader* iface,
    unsigned int regnum) {

    IWineD3DVertexShaderImpl* This = (IWineD3DVertexShaderImpl*) iface;
    IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
    IWineD3DVertexDeclarationImpl *vertexDeclaration = (IWineD3DVertexDeclarationImpl *)deviceImpl->stateBlock->vertexDecl;

    DWORD usage_token = This->semantics_in[regnum].usage;
    DWORD usage = (usage_token & WINED3DSP_DCL_USAGE_MASK) >> WINED3DSP_DCL_USAGE_SHIFT;
    DWORD usage_idx = (usage_token & WINED3DSP_DCL_USAGEINDEX_MASK) >> WINED3DSP_DCL_USAGEINDEX_SHIFT;

    if (vertexDeclaration) {
        int i;
        /* Find the declaration element that matches our register, then check
         * if it has D3DCOLOR as it's type. This works for both d3d8 and d3d9. */
        for (i = 0; i < vertexDeclaration->declarationWNumElements-1; ++i) {
            WINED3DVERTEXELEMENT *element = vertexDeclaration->pDeclarationWine + i;
            if (match_usage(element->Usage, element->UsageIndex, usage, usage_idx)) {
                return element->Type == WINED3DDECLTYPE_D3DCOLOR;
            }
        }
    }

    ERR("Either no vertexdeclaration present, or register not matched. This should never happen.\n");
    return FALSE;
}

/** Generate a vertex shader string using either GL_VERTEX_PROGRAM_ARB
    or GLSL and send it to the card */
static VOID IWineD3DVertexShaderImpl_GenerateShader(
    IWineD3DVertexShader *iface,
    shader_reg_maps* reg_maps,
    CONST DWORD *pFunction) {

    IWineD3DVertexShaderImpl *This = (IWineD3DVertexShaderImpl *)iface;
    SHADER_BUFFER buffer;

#if 0 /* FIXME: Use the buffer that is held by the device, this is ok since fixups will be skipped for software shaders
        it also requires entering a critical section but cuts down the runtime footprint of wined3d and any memory fragmentation that may occur... */
    if (This->device->fixupVertexBufferSize < SHADER_PGMSIZE) {
        HeapFree(GetProcessHeap(), 0, This->fixupVertexBuffer);
        This->fixupVertexBuffer = HeapAlloc(GetProcessHeap() , 0, SHADER_PGMSIZE);
        This->fixupVertexBufferSize = PGMSIZE;
        This->fixupVertexBuffer[0] = 0;
    }
    buffer.buffer = This->device->fixupVertexBuffer;
#else
    buffer.buffer = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, SHADER_PGMSIZE); 
#endif
    buffer.bsize = 0;
    buffer.lineNo = 0;
    buffer.newline = TRUE;

    if (This->baseShader.shader_mode == SHADER_GLSL) {

        /* Create the hw GLSL shader program and assign it as the baseShader.prgId */
        GLhandleARB shader_obj = GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB));

        /* Base Declarations */
        shader_generate_glsl_declarations( (IWineD3DBaseShader*) This, reg_maps, &buffer, &GLINFO_LOCATION);

        /* Base Shader Body */
        shader_generate_main( (IWineD3DBaseShader*) This, &buffer, reg_maps, pFunction);

        /* Unpack 3.0 outputs */
        if (This->baseShader.hex_version >= WINED3DVS_VERSION(3,0))
            vshader_glsl_output_unpack(&buffer, This->semantics_out);

        /* If this shader doesn't use fog copy the z coord to the fog coord so that we can use table fog */
        if (!reg_maps->fog)
            shader_addline(&buffer, "gl_FogFragCoord = gl_Position.z;\n");
        
        /* Write the final position.
         *
         * OpenGL coordinates specify the center of the pixel while d3d coords specify
         * the corner. The offsets are stored in z and w in the 2nd row of the projection
         * matrix to avoid wasting a free shader constant. Add them to the w and z coord
         * of the 2nd row
         */
        shader_addline(&buffer, "gl_Position.x = gl_Position.x + posFixup[2];\n");
        shader_addline(&buffer, "gl_Position.y = gl_Position.y + posFixup[3];\n");
        /* Account for any inverted textures (render to texture case) by reversing the y coordinate
         *  (this is handled in drawPrim() when it sets the MODELVIEW and PROJECTION matrices)
         */
        shader_addline(&buffer, "gl_Position.y = gl_Position.y * posFixup[1];\n");

        shader_addline(&buffer, "}\n");

        TRACE("Compiling shader object %u\n", shader_obj);
        GL_EXTCALL(glShaderSourceARB(shader_obj, 1, (const char**)&buffer.buffer, NULL));
        GL_EXTCALL(glCompileShaderARB(shader_obj));
        print_glsl_info_log(&GLINFO_LOCATION, shader_obj);

        /* Store the shader object */
        This->baseShader.prgId = shader_obj;

    } else if (This->baseShader.shader_mode == SHADER_ARB) {

        /*  Create the hw ARB shader */
        shader_addline(&buffer, "!!ARBvp1.0\n");

        /* Mesa supports only 95 constants */
        if (GL_VEND(MESA) || GL_VEND(WINE))
            This->baseShader.limits.constant_float = 
                min(95, This->baseShader.limits.constant_float);

        /* Base Declarations */
        shader_generate_arb_declarations( (IWineD3DBaseShader*) This, reg_maps, &buffer, &GLINFO_LOCATION);

        /* We need a constant to fixup the final position */
        shader_addline(&buffer, "PARAM posFixup = program.env[%d];\n", ARB_SHADER_PRIVCONST_POS);

        /* Base Shader Body */
        shader_generate_main( (IWineD3DBaseShader*) This, &buffer, reg_maps, pFunction);

        /* If this shader doesn't use fog copy the z coord to the fog coord so that we can use table fog */
        if (!reg_maps->fog)
            shader_addline(&buffer, "MOV result.fogcoord, TMP_OUT.z;\n");

        /* Write the final position.
         *
         * OpenGL coordinates specify the center of the pixel while d3d coords specify
         * the corner. The offsets are stored in the 2nd row of the projection matrix,
         * the x offset in z and the y offset in w. Add them to the resulting position
         */
        shader_addline(&buffer, "ADD TMP_OUT.x, TMP_OUT.x, posFixup.z;\n");
        shader_addline(&buffer, "ADD TMP_OUT.y, TMP_OUT.y, posFixup.w;\n");
        /* Account for any inverted textures (render to texture case) by reversing the y coordinate
         *  (this is handled in drawPrim() when it sets the MODELVIEW and PROJECTION matrices)
         */
        shader_addline(&buffer, "MUL TMP_OUT.y, TMP_OUT.y, posFixup.y;\n");

        shader_addline(&buffer, "MOV result.position, TMP_OUT;\n");
        
        shader_addline(&buffer, "END\n"); 

        /* TODO: change to resource.glObjectHandle or something like that */
        GL_EXTCALL(glGenProgramsARB(1, &This->baseShader.prgId));

        TRACE("Creating a hw vertex shader, prg=%d\n", This->baseShader.prgId);
        GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, This->baseShader.prgId));

        TRACE("Created hw vertex shader, prg=%d\n", This->baseShader.prgId);
        /* Create the program and check for errors */
        GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB,
            buffer.bsize, buffer.buffer));

        if (glGetError() == GL_INVALID_OPERATION) {
            GLint errPos;
            glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &errPos);
            FIXME("HW VertexShader Error at position %d: %s\n",
                  errPos, debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
            This->baseShader.prgId = -1;
        }
    }

#if 1 /* if were using the data buffer of device then we don't need to free it */
  HeapFree(GetProcessHeap(), 0, buffer.buffer);
#endif
}

/* *******************************************
   IWineD3DVertexShader IUnknown parts follow
   ******************************************* */
static HRESULT WINAPI IWineD3DVertexShaderImpl_QueryInterface(IWineD3DVertexShader *iface, REFIID riid, LPVOID *ppobj)
{
    IWineD3DVertexShaderImpl *This = (IWineD3DVertexShaderImpl *)iface;
    TRACE("(%p)->(%s,%p)\n",This,debugstr_guid(riid),ppobj);
    if (IsEqualGUID(riid, &IID_IUnknown) 
        || IsEqualGUID(riid, &IID_IWineD3DBase)
        || IsEqualGUID(riid, &IID_IWineD3DBaseShader)
        || IsEqualGUID(riid, &IID_IWineD3DVertexShader)) {
        IUnknown_AddRef(iface);
        *ppobj = This;