glsl_shader.c

这是一个开放源代码的与WINNT/WIN2K/WIN2003兼容的操作系统

void shader_glsl_dot(SHADER_OPCODE_ARG* arg) {
    CONST SHADER_OPCODE* curOpcode = arg->opcode;
    SHADER_BUFFER* buffer = arg->buffer;
    glsl_src_param_t src0_param;
    glsl_src_param_t src1_param;
    DWORD dst_write_mask, src_write_mask;
    size_t dst_size = 0;

    dst_write_mask = shader_glsl_append_dst(buffer, arg);
    dst_size = shader_glsl_get_write_mask_size(dst_write_mask);

    /* dp3 works on vec3, dp4 on vec4 */
    if (curOpcode->opcode == WINED3DSIO_DP4) {
        src_write_mask = WINED3DSP_WRITEMASK_ALL;
    } else {
        src_write_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
    }

    shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_write_mask, &src0_param);
    shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], src_write_mask, &src1_param);

    if (dst_size > 1) {
        shader_addline(buffer, "vec%d(dot(%s, %s)));\n", dst_size, src0_param.param_str, src1_param.param_str);
    } else {
        shader_addline(buffer, "dot(%s, %s));\n", src0_param.param_str, src1_param.param_str);
    }
}

/* Note that this instruction has some restrictions. The destination write mask
 * can't contain the w component, and the source swizzles have to be .xyzw */
void shader_glsl_cross(SHADER_OPCODE_ARG *arg) {
    DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
    glsl_src_param_t src0_param;
    glsl_src_param_t src1_param;
    char dst_mask[6];

    shader_glsl_get_write_mask(arg->dst, dst_mask);
    shader_glsl_append_dst(arg->buffer, arg);
    shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
    shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], src_mask, &src1_param);
    shader_addline(arg->buffer, "cross(%s, %s).%s);\n", src0_param.param_str, src1_param.param_str, dst_mask);
}

/* Process the WINED3DSIO_POW instruction in GLSL (dst = |src0|^src1)
 * Src0 and src1 are scalars. Note that D3D uses the absolute of src0, while
 * GLSL uses the value as-is. */
void shader_glsl_pow(SHADER_OPCODE_ARG *arg) {
    SHADER_BUFFER *buffer = arg->buffer;
    glsl_src_param_t src0_param;
    glsl_src_param_t src1_param;
    DWORD dst_write_mask;
    size_t dst_size;

    dst_write_mask = shader_glsl_append_dst(buffer, arg);
    dst_size = shader_glsl_get_write_mask_size(dst_write_mask);

    shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
    shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_0, &src1_param);

    if (dst_size > 1) {
        shader_addline(buffer, "vec%d(pow(abs(%s), %s)));\n", dst_size, src0_param.param_str, src1_param.param_str);
    } else {
        shader_addline(buffer, "pow(abs(%s), %s));\n", src0_param.param_str, src1_param.param_str);
    }
}

/* Map the opcode 1-to-1 to the GL code (arg->dst = instruction(src0, src1, ...) */
void shader_glsl_map2gl(SHADER_OPCODE_ARG* arg) {
    CONST SHADER_OPCODE* curOpcode = arg->opcode;
    SHADER_BUFFER* buffer = arg->buffer;
    glsl_src_param_t src_param;
    const char *instruction;
    char arguments[256];
    DWORD write_mask;
    unsigned i;

    /* Determine the GLSL function to use based on the opcode */
    /* TODO: Possibly make this a table for faster lookups */
    switch (curOpcode->opcode) {
        case WINED3DSIO_MIN: instruction = "min"; break;
        case WINED3DSIO_MAX: instruction = "max"; break;
        case WINED3DSIO_ABS: instruction = "abs"; break;
        case WINED3DSIO_FRC: instruction = "fract"; break;
        case WINED3DSIO_NRM: instruction = "normalize"; break;
        case WINED3DSIO_LOGP:
        case WINED3DSIO_LOG: instruction = "log2"; break;
        case WINED3DSIO_EXP: instruction = "exp2"; break;
        case WINED3DSIO_SGN: instruction = "sign"; break;
        case WINED3DSIO_DSX: instruction = "dFdx"; break;
        case WINED3DSIO_DSY: instruction = "dFdy"; break;
        default: instruction = "";
            FIXME("Opcode %s not yet handled in GLSL\n", curOpcode->name);
            break;
    }

    write_mask = shader_glsl_append_dst(buffer, arg);

    arguments[0] = '\0';
    if (curOpcode->num_params > 0) {
        shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src_param);
        strcat(arguments, src_param.param_str);
        for (i = 2; i < curOpcode->num_params; ++i) {
            strcat(arguments, ", ");
            shader_glsl_add_src_param(arg, arg->src[i-1], arg->src_addr[i-1], write_mask, &src_param);
            strcat(arguments, src_param.param_str);
        }
    }

    shader_addline(buffer, "%s(%s));\n", instruction, arguments);
}

/** Process the WINED3DSIO_EXPP instruction in GLSL:
 * For shader model 1.x, do the following (and honor the writemask, so use a temporary variable):
 *   dst.x = 2^(floor(src))
 *   dst.y = src - floor(src)
 *   dst.z = 2^src   (partial precision is allowed, but optional)
 *   dst.w = 1.0;
 * For 2.0 shaders, just do this (honoring writemask and swizzle):
 *   dst = 2^src;    (partial precision is allowed, but optional)
 */
void shader_glsl_expp(SHADER_OPCODE_ARG* arg) {
    IWineD3DBaseShaderImpl *shader = (IWineD3DBaseShaderImpl *)arg->shader;
    glsl_src_param_t src_param;

    shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src_param);

    if (shader->baseShader.hex_version < WINED3DPS_VERSION(2,0)) {
        char dst_mask[6];

        shader_addline(arg->buffer, "tmp0.x = exp2(floor(%s));\n", src_param.param_str);
        shader_addline(arg->buffer, "tmp0.y = %s - floor(%s);\n", src_param.param_str, src_param.param_str);
        shader_addline(arg->buffer, "tmp0.z = exp2(%s);\n", src_param.param_str);
        shader_addline(arg->buffer, "tmp0.w = 1.0;\n");

        shader_glsl_append_dst(arg->buffer, arg);
        shader_glsl_get_write_mask(arg->dst, dst_mask);
        shader_addline(arg->buffer, "tmp0%s);\n", dst_mask);
    } else {
        DWORD write_mask;
        size_t mask_size;

        write_mask = shader_glsl_append_dst(arg->buffer, arg);
        mask_size = shader_glsl_get_write_mask_size(write_mask);

        if (mask_size > 1) {
            shader_addline(arg->buffer, "vec%d(exp2(%s)));\n", mask_size, src_param.param_str);
        } else {
            shader_addline(arg->buffer, "exp2(%s));\n", src_param.param_str);
        }
    }
}

/** Process the RCP (reciprocal or inverse) opcode in GLSL (dst = 1 / src) */
void shader_glsl_rcp(SHADER_OPCODE_ARG* arg) {
    glsl_src_param_t src_param;
    DWORD write_mask;
    size_t mask_size;

    write_mask = shader_glsl_append_dst(arg->buffer, arg);
    mask_size = shader_glsl_get_write_mask_size(write_mask);
    shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_3, &src_param);

    if (mask_size > 1) {
        shader_addline(arg->buffer, "vec%d(1.0 / %s));\n", mask_size, src_param.param_str);
    } else {
        shader_addline(arg->buffer, "1.0 / %s);\n", src_param.param_str);
    }
}

void shader_glsl_rsq(SHADER_OPCODE_ARG* arg) {
    SHADER_BUFFER* buffer = arg->buffer;
    glsl_src_param_t src_param;
    DWORD write_mask;
    size_t mask_size;

    write_mask = shader_glsl_append_dst(buffer, arg);
    mask_size = shader_glsl_get_write_mask_size(write_mask);

    shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_3, &src_param);

    if (mask_size > 1) {
        shader_addline(buffer, "vec%d(inversesqrt(%s)));\n", mask_size, src_param.param_str);
    } else {
        shader_addline(buffer, "inversesqrt(%s));\n", src_param.param_str);
    }
}

/** Process signed comparison opcodes in GLSL. */
void shader_glsl_compare(SHADER_OPCODE_ARG* arg) {
    glsl_src_param_t src0_param;
    glsl_src_param_t src1_param;
    DWORD write_mask;
    size_t mask_size;

    write_mask = shader_glsl_append_dst(arg->buffer, arg);
    mask_size = shader_glsl_get_write_mask_size(write_mask);
    shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src0_param);
    shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);

    if (mask_size > 1) {
        const char *compare;

        switch(arg->opcode->opcode) {
            case WINED3DSIO_SLT: compare = "lessThan"; break;
            case WINED3DSIO_SGE: compare = "greaterThanEqual"; break;
            default: compare = "";
                FIXME("Can't handle opcode %s\n", arg->opcode->name);
        }

        shader_addline(arg->buffer, "vec%d(%s(%s, %s)));\n", mask_size, compare,
                src0_param.param_str, src1_param.param_str);
    } else {
        const char *compare;

        switch(arg->opcode->opcode) {
            case WINED3DSIO_SLT: compare = "<"; break;
            case WINED3DSIO_SGE: compare = ">="; break;
            default: compare = "";
                FIXME("Can't handle opcode %s\n", arg->opcode->name);
        }

        shader_addline(arg->buffer, "(%s %s %s) ? 1.0 : 0.0);\n",
                src0_param.param_str, compare, src1_param.param_str);
    }
}

/** Process CMP instruction in GLSL (dst = src0 >= 0.0 ? src1 : src2), per channel */
void shader_glsl_cmp(SHADER_OPCODE_ARG* arg) {
    glsl_src_param_t src0_param;
    glsl_src_param_t src1_param;
    glsl_src_param_t src2_param;
    DWORD write_mask, cmp_channel = 0;
    unsigned int i, j;

    /* Cycle through all source0 channels */
    for (i=0; i<4; i++) {
        write_mask = 0;
        /* Find the destination channels which use the current source0 channel */
        for (j=0; j<4; j++) {
            if ( ((arg->src[0] >> (WINED3DSP_SWIZZLE_SHIFT + 2*j)) & 0x3) == i ) {
                write_mask |= WINED3DSP_WRITEMASK_0 << j;
                cmp_channel = WINED3DSP_WRITEMASK_0 << j;
            }
        }
        write_mask = shader_glsl_append_dst_ext(arg->buffer, arg, arg->dst & (~WINED3DSP_SWIZZLE_MASK | write_mask));
        if (!write_mask) continue;

        shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], cmp_channel, &src0_param);
        shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
        shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], write_mask, &src2_param);

        shader_addline(arg->buffer, "%s >= 0.0 ? %s : %s);\n",
                src0_param.param_str, src1_param.param_str, src2_param.param_str);
    }
}

/** Process the CND opcode in GLSL (dst = (src0 > 0.5) ? src1 : src2) */
/* For ps 1.1-1.3, only a single component of src0 is used. For ps 1.4
 * the compare is done per component of src0. */
void shader_glsl_cnd(SHADER_OPCODE_ARG* arg) {
    IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
    glsl_src_param_t src0_param;
    glsl_src_param_t src1_param;
    glsl_src_param_t src2_param;
    DWORD write_mask, cmp_channel = 0;
    unsigned int i, j;

    if (shader->baseShader.hex_version < WINED3DPS_VERSION(1, 4)) {
        write_mask = shader_glsl_append_dst(arg->buffer, arg);
        shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
        shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
        shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], write_mask, &src2_param);
        shader_addline(arg->buffer, "%s > 0.5 ? %s : %s);\n",
                src0_param.param_str, src1_param.param_str, src2_param.param_str);
        return;
    }
    /* Cycle through all source0 channels */
    for (i=0; i<4; i++) {
        write_mask = 0;
        /* Find the destination channels which use the current source0 channel */
        for (j=0; j<4; j++) {
            if ( ((arg->src[0] >> (WINED3DSP_SWIZZLE_SHIFT + 2*j)) & 0x3) == i ) {
                write_mask |= WINED3DSP_WRITEMASK_0 << j;
                cmp_channel = WINED3DSP_WRITEMASK_0 << j;
            }
        }
        write_mask = shader_glsl_append_dst_ext(arg->buffer, arg, arg->dst & (~WINED3DSP_SWIZZLE_MASK | write_mask));
        if (!write_mask) continue;

        shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], cmp_channel, &src0_param);
        shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
        shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], write_mask, &src2_param);

        shader_addline(arg->buffer, "%s > 0.5 ? %s : %s);\n",
                src0_param.param_str, src1_param.param_str, src2_param.param_str);
    }
}

/** GLSL code generation for WINED3DSIO_MAD: Multiply the first 2 opcodes, then add the last */
void shader_glsl_mad(SHADER_OPCODE_ARG* arg) {
    glsl_src_param_t src0_param;
    glsl_src_param_t src1_param;
    glsl_src_param_t src2_param;
    DWORD write_mask;

    write_mask = shader_glsl_append_dst(arg->buffer, arg);
    shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src0_param);
    shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
    shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], write_mask, &src2_param);
    shader_addline(arg->buffer, "(%s * %s) + %s);\n",
            src0_param.param_str, src1_param.param_str, src2_param.param_str);
}

/** Handles transforming all WINED3DSIO_M?x? opcodes for 
    Vertex shaders to GLSL codes */
void shader_glsl_mnxn(SHADER_OPCODE_ARG* arg) {
    int i;
    int nComponents = 0;
    SHADER_OPCODE_ARG tmpArg;
   
    memset(&tmpArg, 0, sizeof(SHADER_OPCODE_ARG));

    /* Set constants for the temporary argument */
    tmpArg.shader      = arg->shader;
    tmpArg.buffer      = arg->buffer;