glsl_shader.c

Go to the documentation of this file.

/*
 * GLSL pixel and vertex shader implementation
 *
 * Copyright 2006 Jason Green
 * Copyright 2006-2007 Henri Verbeet
 * Copyright 2007-2009, 2013 Stefan Dösinger for CodeWeavers
 * Copyright 2009-2011 Henri Verbeet for CodeWeavers
 *
 * This library is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * This library is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
 * License along with this library; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
 */
 
/*
 * D3D shader asm has swizzles on source parameters, and write masks for
 * destination parameters. GLSL uses swizzles for both. The result of this is
 * that for example "mov dst.xw, src.zyxw" becomes "dst.xw = src.zw" in GLSL.
 * Ie, to generate a proper GLSL source swizzle, we need to take the D3D write
 * mask for the destination parameter into account.
 */
 
#include "config.h"
#include "wine/port.h"
 
#include <limits.h>
#include <stdio.h>
#ifdef HAVE_FLOAT_H
# include <float.h>
#endif
 
#include "wined3d_private.h"
 
WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader);
WINE_DECLARE_DEBUG_CHANNEL(d3d);
WINE_DECLARE_DEBUG_CHANNEL(winediag);
 
#define WINED3D_GLSL_SAMPLE_PROJECTED   0x01
#define WINED3D_GLSL_SAMPLE_LOD         0x02
#define WINED3D_GLSL_SAMPLE_GRAD        0x04
#define WINED3D_GLSL_SAMPLE_LOAD        0x08
#define WINED3D_GLSL_SAMPLE_OFFSET      0x10
 
static const struct
{
    unsigned int coord_size;
    unsigned int resinfo_size;
    const char *type_part;
}
resource_type_info[] =
{
    {0, 0, ""},        /* WINED3D_SHADER_RESOURCE_NONE */
    {1, 1, "Buffer"},  /* WINED3D_SHADER_RESOURCE_BUFFER */
    {1, 1, "1D"},      /* WINED3D_SHADER_RESOURCE_TEXTURE_1D */
    {2, 2, "2D"},      /* WINED3D_SHADER_RESOURCE_TEXTURE_2D */
    {2, 2, ""},        /* WINED3D_SHADER_RESOURCE_TEXTURE_2DMS */
    {3, 3, "3D"},      /* WINED3D_SHADER_RESOURCE_TEXTURE_3D */
    {3, 2, "Cube"},    /* WINED3D_SHADER_RESOURCE_TEXTURE_CUBE */
    {2, 2, ""},        /* WINED3D_SHADER_RESOURCE_TEXTURE_1DARRAY */
    {3, 3, "2DArray"}, /* WINED3D_SHADER_RESOURCE_TEXTURE_2DARRAY */
    {3, 3, ""},        /* WINED3D_SHADER_RESOURCE_TEXTURE_2DMSARRAY */
    {4, 3, ""},        /* WINED3D_SHADER_RESOURCE_TEXTURE_CUBEARRAY */
};
 
struct glsl_dst_param
{
    char reg_name[150];
    char mask_str[6];
};
 
struct glsl_src_param
{
    char reg_name[150];
    char param_str[200];
};
 
struct glsl_sample_function
{
    struct wined3d_string_buffer *name;
    unsigned int coord_mask;
    unsigned int deriv_mask;
    enum wined3d_data_type data_type;
    BOOL output_single_component;
    unsigned int offset_size;
    enum wined3d_shader_resource_type emulate_lod;
};
 
enum heap_node_op
{
    HEAP_NODE_TRAVERSE_LEFT,
    HEAP_NODE_TRAVERSE_RIGHT,
    HEAP_NODE_POP,
};
 
struct constant_entry
{
    unsigned int idx;
    unsigned int version;
};
 
struct constant_heap
{
    struct constant_entry *entries;
    BOOL *contained;
    unsigned int *positions;
    unsigned int size;
};
 
/* GLSL shader private data */
struct shader_glsl_priv
{
    struct wined3d_string_buffer shader_buffer;
    struct wined3d_string_buffer_list string_buffers;
    struct wine_rb_tree program_lookup;
    struct constant_heap vconst_heap;
    struct constant_heap pconst_heap;
    unsigned char *stack;
    UINT next_constant_version;
 
    const struct wined3d_vertex_pipe_ops *vertex_pipe;
    const struct fragment_pipeline *fragment_pipe;
    struct wine_rb_tree ffp_vertex_shaders;
    struct wine_rb_tree ffp_fragment_shaders;
    BOOL ffp_proj_control;
    BOOL legacy_lighting;
};
 
struct glsl_vs_program
{
    struct list shader_entry;
    GLuint id;
    GLenum vertex_color_clamp;
    GLint uniform_f_locations[WINED3D_MAX_VS_CONSTS_F];
    GLint uniform_i_locations[WINED3D_MAX_CONSTS_I];
    GLint uniform_b_locations[WINED3D_MAX_CONSTS_B];
    GLint pos_fixup_location;
 
    GLint modelview_matrix_location[MAX_VERTEX_INDEX_BLENDS];
    GLint normal_matrix_location[MAX_VERTEX_INDEX_BLENDS];
    GLint projection_matrix_location;
    GLint texture_matrix_location[MAX_TEXTURES];
    GLint material_ambient_location;
    GLint material_diffuse_location;
    GLint material_specular_location;
    GLint material_emissive_location;
    GLint material_shininess_location;
    GLint light_ambient_location;
    struct
    {
        GLint diffuse;
        GLint specular;
        GLint ambient;
        GLint position;
        GLint direction;
        GLint range;
        GLint falloff;
        GLint c_att;
        GLint l_att;
        GLint q_att;
        GLint cos_htheta;
        GLint cos_hphi;
    } light_location[MAX_ACTIVE_LIGHTS];
    GLint pointsize_location;
    GLint pointsize_min_location;
    GLint pointsize_max_location;
    GLint pointsize_c_att_location;
    GLint pointsize_l_att_location;
    GLint pointsize_q_att_location;
    GLint clip_planes_location;
};
 
struct glsl_hs_program
{
    struct list shader_entry;
    GLuint id;
};
 
struct glsl_ds_program
{
    struct list shader_entry;
    GLuint id;
 
    GLint pos_fixup_location;
};
 
struct glsl_gs_program
{
    struct list shader_entry;
    GLuint id;
 
    GLint pos_fixup_location;
};
 
struct glsl_ps_program
{
    struct list shader_entry;
    GLuint id;
    GLint uniform_f_locations[WINED3D_MAX_PS_CONSTS_F];
    GLint uniform_i_locations[WINED3D_MAX_CONSTS_I];
    GLint uniform_b_locations[WINED3D_MAX_CONSTS_B];
    GLint bumpenv_mat_location[MAX_TEXTURES];
    GLint bumpenv_lum_scale_location[MAX_TEXTURES];
    GLint bumpenv_lum_offset_location[MAX_TEXTURES];
    GLint tss_constant_location[MAX_TEXTURES];
    GLint tex_factor_location;
    GLint specular_enable_location;
    GLint fog_color_location;
    GLint fog_density_location;
    GLint fog_end_location;
    GLint fog_scale_location;
    GLint alpha_test_ref_location;
    GLint ycorrection_location;
    GLint np2_fixup_location;
    GLint color_key_location;
    const struct ps_np2fixup_info *np2_fixup_info;
};
 
struct glsl_cs_program
{
    struct list shader_entry;
    GLuint id;
};
 
/* Struct to maintain data about a linked GLSL program */
struct glsl_shader_prog_link
{
    struct wine_rb_entry program_lookup_entry;
    struct glsl_vs_program vs;
    struct glsl_hs_program hs;
    struct glsl_ds_program ds;
    struct glsl_gs_program gs;
    struct glsl_ps_program ps;
    struct glsl_cs_program cs;
    GLuint id;
    DWORD constant_update_mask;
    unsigned int constant_version;
    DWORD shader_controlled_clip_distances : 1;
    DWORD clip_distance_mask : 8; /* MAX_CLIP_DISTANCES, 8 */
    DWORD padding : 23;
};
 
struct glsl_program_key
{
    GLuint vs_id;
    GLuint hs_id;
    GLuint ds_id;
    GLuint gs_id;
    GLuint ps_id;
    GLuint cs_id;
};
 
struct shader_glsl_ctx_priv {
    const struct vs_compile_args    *cur_vs_args;
    const struct ds_compile_args    *cur_ds_args;
    const struct ps_compile_args    *cur_ps_args;
    struct ps_np2fixup_info         *cur_np2fixup_info;
    struct wined3d_string_buffer_list *string_buffers;
};
 
struct glsl_context_data
{
    struct glsl_shader_prog_link *glsl_program;
    GLenum vertex_color_clamp;
    BOOL rasterization_disabled;
};
 
struct glsl_ps_compiled_shader
{
    struct ps_compile_args          args;
    struct ps_np2fixup_info         np2fixup;
    GLuint                          id;
};
 
struct glsl_vs_compiled_shader
{
    struct vs_compile_args          args;
    GLuint                          id;
};
 
struct glsl_hs_compiled_shader
{
    GLuint id;
};
 
struct glsl_ds_compiled_shader
{
    struct ds_compile_args args;
    GLuint id;
};
 
struct glsl_gs_compiled_shader
{
    struct gs_compile_args args;
    GLuint id;
};
 
struct glsl_cs_compiled_shader
{
    GLuint id;
};
 
struct glsl_shader_private
{
    union
    {
        struct glsl_vs_compiled_shader *vs;
        struct glsl_hs_compiled_shader *hs;
        struct glsl_ds_compiled_shader *ds;
        struct glsl_gs_compiled_shader *gs;
        struct glsl_ps_compiled_shader *ps;
        struct glsl_cs_compiled_shader *cs;
    } gl_shaders;
    unsigned int num_gl_shaders, shader_array_size;
};
 
struct glsl_ffp_vertex_shader
{
    struct wined3d_ffp_vs_desc desc;
    GLuint id;
    struct list linked_programs;
};
 
struct glsl_ffp_fragment_shader
{
    struct ffp_frag_desc entry;
    GLuint id;
    struct list linked_programs;
};
 
struct glsl_ffp_destroy_ctx
{
    struct shader_glsl_priv *priv;
    const struct wined3d_gl_info *gl_info;
};
 
static void shader_glsl_generate_shader_epilogue(const struct wined3d_shader_context *ctx);
 
static const char *debug_gl_shader_type(GLenum type)
{
    switch (type)
    {
#define WINED3D_TO_STR(u) case u: return #u
        WINED3D_TO_STR(GL_VERTEX_SHADER);
        WINED3D_TO_STR(GL_TESS_CONTROL_SHADER);
        WINED3D_TO_STR(GL_TESS_EVALUATION_SHADER);
        WINED3D_TO_STR(GL_GEOMETRY_SHADER);
        WINED3D_TO_STR(GL_FRAGMENT_SHADER);
        WINED3D_TO_STR(GL_COMPUTE_SHADER);
#undef WINED3D_TO_STR
        default:
            return wine_dbg_sprintf("UNKNOWN(%#x)", type);
    }
}
 
static const char *shader_glsl_get_prefix(enum wined3d_shader_type type)
{
    switch (type)
    {
        case WINED3D_SHADER_TYPE_VERTEX:
            return "vs";
 
        case WINED3D_SHADER_TYPE_HULL:
            return "hs";
 
        case WINED3D_SHADER_TYPE_DOMAIN:
            return "ds";
 
        case WINED3D_SHADER_TYPE_GEOMETRY:
            return "gs";
 
        case WINED3D_SHADER_TYPE_PIXEL:
            return "ps";
 
        case WINED3D_SHADER_TYPE_COMPUTE:
            return "cs";
 
        default:
            FIXME("Unhandled shader type %#x.\n", type);
            return "unknown";
    }
}
 
static unsigned int shader_glsl_get_version(const struct wined3d_gl_info *gl_info)
{
    if (gl_info->glsl_version >= MAKEDWORD_VERSION(4, 40))
        return 440;
    else if (gl_info->glsl_version >= MAKEDWORD_VERSION(1, 50))
        return 150;
    else if (gl_info->glsl_version >= MAKEDWORD_VERSION(1, 30))
        return 130;
    else
        return 120;
}
 
static void shader_glsl_add_version_declaration(struct wined3d_string_buffer *buffer,
        const struct wined3d_gl_info *gl_info)
{
    shader_addline(buffer, "#version %u\n", shader_glsl_get_version(gl_info));
}
 
static void shader_glsl_append_imm_vec4(struct wined3d_string_buffer *buffer, const float *values)
{
    char str[4][17];
 
    wined3d_ftoa(values[0], str[0]);
    wined3d_ftoa(values[1], str[1]);
    wined3d_ftoa(values[2], str[2]);
    wined3d_ftoa(values[3], str[3]);
    shader_addline(buffer, "vec4(%s, %s, %s, %s)", str[0], str[1], str[2], str[3]);
}
 
static void shader_glsl_append_imm_ivec(struct wined3d_string_buffer *buffer,
        const int *values, unsigned int size)
{
    int i;
 
    if (!size || size > 4)
    {
        ERR("Invalid vector size %u.\n", size);
        return;
    }
 
    if (size > 1)
        shader_addline(buffer, "ivec%u(", size);
 
    for (i = 0; i < size; ++i)
        shader_addline(buffer, i ? ", %#x" : "%#x", values[i]);
 
    if (size > 1)
        shader_addline(buffer, ")");
}
 
static const char *get_info_log_line(const char **ptr)
{
    const char *p, *q;
 
    p = *ptr;
    if (!(q = strstr(p, "\n")))
    {
        if (!*p) return NULL;
        *ptr += strlen(p);
        return p;
    }
    *ptr = q + 1;
 
    return p;
}
 
/* Context activation is done by the caller. */
void print_glsl_info_log(const struct wined3d_gl_info *gl_info, GLuint id, BOOL program)
{
    int length = 0;
    char *log;
 
    if (!WARN_ON(d3d_shader) && !FIXME_ON(d3d_shader))
        return;
 
    if (program)
        GL_EXTCALL(glGetProgramiv(id, GL_INFO_LOG_LENGTH, &length));
    else
        GL_EXTCALL(glGetShaderiv(id, GL_INFO_LOG_LENGTH, &length));
 
    /* A size of 1 is just a null-terminated string, so the log should be bigger than
     * that if there are errors. */
    if (length > 1)
    {
        const char *ptr, *line;
 
        log = heap_alloc(length);
        /* The info log is supposed to be zero-terminated, but at least some
         * versions of fglrx don't terminate the string properly. The reported
         * length does include the terminator, so explicitly set it to zero
         * here. */
        log[length - 1] = 0;
        if (program)
            GL_EXTCALL(glGetProgramInfoLog(id, length, NULL, log));
        else
            GL_EXTCALL(glGetShaderInfoLog(id, length, NULL, log));
 
        ptr = log;
        if (gl_info->quirks & WINED3D_QUIRK_INFO_LOG_SPAM)
        {
            WARN("Info log received from GLSL shader #%u:\n", id);
            while ((line = get_info_log_line(&ptr))) WARN("    %.*s", (int)(ptr - line), line);
        }
        else
        {
            FIXME("Info log received from GLSL shader #%u:\n", id);
            while ((line = get_info_log_line(&ptr))) FIXME("    %.*s", (int)(ptr - line), line);
        }
        heap_free(log);
    }
}
 
/* Context activation is done by the caller. */
static void shader_glsl_compile(const struct wined3d_gl_info *gl_info, GLuint shader, const char *src)
{
    const char *ptr, *line;
 
    TRACE("Compiling shader object %u.\n", shader);
 
    if (TRACE_ON(d3d_shader))
    {
        ptr = src;
        while ((line = get_info_log_line(&ptr))) TRACE_(d3d_shader)("    %.*s", (int)(ptr - line), line);
    }
 
    GL_EXTCALL(glShaderSource(shader, 1, &src, NULL));
    checkGLcall("glShaderSource");
    GL_EXTCALL(glCompileShader(shader));
    checkGLcall("glCompileShader");
    print_glsl_info_log(gl_info, shader, FALSE);
}
 
/* Context activation is done by the caller. */
static void shader_glsl_dump_program_source(const struct wined3d_gl_info *gl_info, GLuint program)
{
    GLint i, shader_count, source_size = -1;
    GLuint *shaders;
    char *source = NULL;
 
    GL_EXTCALL(glGetProgramiv(program, GL_ATTACHED_SHADERS, &shader_count));
    if (!(shaders = heap_calloc(shader_count, sizeof(*shaders))))
    {
        ERR("Failed to allocate shader array memory.\n");
        return;
    }
 
    GL_EXTCALL(glGetAttachedShaders(program, shader_count, NULL, shaders));
    for (i = 0; i < shader_count; ++i)
    {
        const char *ptr, *line;
        GLint tmp;
 
        GL_EXTCALL(glGetShaderiv(shaders[i], GL_SHADER_SOURCE_LENGTH, &tmp));
 
        if (source_size < tmp)
        {
            heap_free(source);
 
            if (!(source = heap_alloc_zero(tmp)))
            {
                ERR("Failed to allocate %d bytes for shader source.\n", tmp);
                heap_free(shaders);
                return;
            }
            source_size = tmp;
        }
 
        FIXME("Shader %u:\n", shaders[i]);
        GL_EXTCALL(glGetShaderiv(shaders[i], GL_SHADER_TYPE, &tmp));
        FIXME("    GL_SHADER_TYPE: %s.\n", debug_gl_shader_type(tmp));
        GL_EXTCALL(glGetShaderiv(shaders[i], GL_COMPILE_STATUS, &tmp));
        FIXME("    GL_COMPILE_STATUS: %d.\n", tmp);
        FIXME("\n");
 
        ptr = source;
        GL_EXTCALL(glGetShaderSource(shaders[i], source_size, NULL, source));
        while ((line = get_info_log_line(&ptr))) FIXME("    %.*s", (int)(ptr - line), line);
        FIXME("\n");
    }
 
    heap_free(source);
    heap_free(shaders);
}
 
/* Context activation is done by the caller. */
void shader_glsl_validate_link(const struct wined3d_gl_info *gl_info, GLuint program)
{
    GLint tmp;
 
    if (!TRACE_ON(d3d_shader) && !FIXME_ON(d3d_shader))
        return;
 
    GL_EXTCALL(glGetProgramiv(program, GL_LINK_STATUS, &tmp));
    if (!tmp)
    {
        FIXME("Program %u link status invalid.\n", program);
        shader_glsl_dump_program_source(gl_info, program);
    }
 
    print_glsl_info_log(gl_info, program, TRUE);
}
 
static BOOL shader_glsl_use_layout_qualifier(const struct wined3d_gl_info *gl_info)
{
    /* Layout qualifiers were introduced in GLSL 1.40. The Nvidia Legacy GPU
     * driver (series 340.xx) doesn't parse layout qualifiers in older GLSL
     * versions. */
    return shader_glsl_get_version(gl_info) >= 140;
}
 
static BOOL shader_glsl_use_layout_binding_qualifier(const struct wined3d_gl_info *gl_info)
{
    return gl_info->supported[ARB_SHADING_LANGUAGE_420PACK] && shader_glsl_use_layout_qualifier(gl_info);
}
 
static void shader_glsl_init_uniform_block_bindings(const struct wined3d_gl_info *gl_info,
        struct shader_glsl_priv *priv, GLuint program_id,
        const struct wined3d_shader_reg_maps *reg_maps)
{
    const char *prefix = shader_glsl_get_prefix(reg_maps->shader_version.type);
    struct wined3d_string_buffer *name;
    unsigned int i, base, count;
    GLuint block_idx;
 
    if (shader_glsl_use_layout_binding_qualifier(gl_info))
        return;
 
    name = string_buffer_get(&priv->string_buffers);
    wined3d_gl_limits_get_uniform_block_range(&gl_info->limits, reg_maps->shader_version.type, &base, &count);
    for (i = 0; i < count; ++i)
    {
        if (!reg_maps->cb_sizes[i])
            continue;
 
        string_buffer_sprintf(name, "block_%s_cb%u", prefix, i);
        block_idx = GL_EXTCALL(glGetUniformBlockIndex(program_id, name->buffer));
        GL_EXTCALL(glUniformBlockBinding(program_id, block_idx, base + i));
    }
    checkGLcall("glUniformBlockBinding");
    string_buffer_release(&priv->string_buffers, name);
}
 
/* Context activation is done by the caller. */
static void shader_glsl_load_samplers_range(const struct wined3d_gl_info *gl_info,
        struct shader_glsl_priv *priv, GLuint program_id, const char *prefix,
        unsigned int base, unsigned int count, const DWORD *tex_unit_map)
{
    struct wined3d_string_buffer *sampler_name = string_buffer_get(&priv->string_buffers);
    unsigned int i, mapped_unit;
    GLint name_loc;
 
    for (i = 0; i < count; ++i)
    {
        string_buffer_sprintf(sampler_name, "%s_sampler%u", prefix, i);
        name_loc = GL_EXTCALL(glGetUniformLocation(program_id, sampler_name->buffer));
        if (name_loc == -1)
            continue;
 
        mapped_unit = tex_unit_map ? tex_unit_map[base + i] : base + i;
        if (mapped_unit == WINED3D_UNMAPPED_STAGE || mapped_unit >= gl_info->limits.combined_samplers)
        {
            ERR("Trying to load sampler %s on unsupported unit %u.\n", sampler_name->buffer, mapped_unit);
            continue;
        }
 
        TRACE("Loading sampler %s on unit %u.\n", sampler_name->buffer, mapped_unit);
        GL_EXTCALL(glUniform1i(name_loc, mapped_unit));
    }
    checkGLcall("Load sampler bindings");
    string_buffer_release(&priv->string_buffers, sampler_name);
}
 
static unsigned int shader_glsl_map_tex_unit(const struct wined3d_context *context,
        const struct wined3d_shader_version *shader_version, unsigned int sampler_idx)
{
    const DWORD *tex_unit_map;
    unsigned int base, count;
 
    tex_unit_map = context_get_tex_unit_mapping(context, shader_version, &base, &count);
    if (sampler_idx >= count)
        return WINED3D_UNMAPPED_STAGE;
    if (!tex_unit_map)
        return base + sampler_idx;
    return tex_unit_map[base + sampler_idx];
}
 
static void shader_glsl_append_sampler_binding_qualifier(struct wined3d_string_buffer *buffer,
        const struct wined3d_context *context, const struct wined3d_shader_version *shader_version,
        unsigned int sampler_idx)
{
    unsigned int mapped_unit = shader_glsl_map_tex_unit(context, shader_version, sampler_idx);
    if (mapped_unit != WINED3D_UNMAPPED_STAGE)
        shader_addline(buffer, "layout(binding = %u)\n", mapped_unit);
    else
        ERR("Unmapped sampler %u.\n", sampler_idx);
}
 
/* Context activation is done by the caller. */
static void shader_glsl_load_samplers(const struct wined3d_context *context,
        struct shader_glsl_priv *priv, GLuint program_id, const struct wined3d_shader_reg_maps *reg_maps)
{
    const struct wined3d_gl_info *gl_info = context->gl_info;
    const struct wined3d_shader_version *shader_version;
    const DWORD *tex_unit_map;
    unsigned int base, count;
    const char *prefix;
 
    if (shader_glsl_use_layout_binding_qualifier(gl_info))
        return;
 
    shader_version = reg_maps ? &reg_maps->shader_version : NULL;
    prefix = shader_glsl_get_prefix(shader_version ? shader_version->type : WINED3D_SHADER_TYPE_PIXEL);
    tex_unit_map = context_get_tex_unit_mapping(context, shader_version, &base, &count);
    shader_glsl_load_samplers_range(gl_info, priv, program_id, prefix, base, count, tex_unit_map);
}
 
static void shader_glsl_load_icb(const struct wined3d_gl_info *gl_info, struct shader_glsl_priv *priv,
        GLuint program_id, const struct wined3d_shader_reg_maps *reg_maps)
{
    const struct wined3d_shader_immediate_constant_buffer *icb = reg_maps->icb;
 
    if (icb)
    {
        struct wined3d_string_buffer *icb_name = string_buffer_get(&priv->string_buffers);
        const char *prefix = shader_glsl_get_prefix(reg_maps->shader_version.type);
        GLint icb_location;
 
        string_buffer_sprintf(icb_name, "%s_icb", prefix);
        icb_location = GL_EXTCALL(glGetUniformLocation(program_id, icb_name->buffer));
        GL_EXTCALL(glUniform4fv(icb_location, icb->vec4_count, (const GLfloat *)icb->data));
        checkGLcall("Load immediate constant buffer");
 
        string_buffer_release(&priv->string_buffers, icb_name);
    }
}
 
/* Context activation is done by the caller. */
static void shader_glsl_load_images(const struct wined3d_gl_info *gl_info, struct shader_glsl_priv *priv,
        GLuint program_id, const struct wined3d_shader_reg_maps *reg_maps)
{
    const char *prefix = shader_glsl_get_prefix(reg_maps->shader_version.type);
    struct wined3d_string_buffer *name;
    GLint location;
    unsigned int i;
 
    if (shader_glsl_use_layout_binding_qualifier(gl_info))
        return;
 
    name = string_buffer_get(&priv->string_buffers);
    for (i = 0; i < MAX_UNORDERED_ACCESS_VIEWS; ++i)
    {
        if (!reg_maps->uav_resource_info[i].type)
            continue;
 
        string_buffer_sprintf(name, "%s_image%u", prefix, i);
        location = GL_EXTCALL(glGetUniformLocation(program_id, name->buffer));
        if (location == -1)
            continue;
 
        TRACE("Loading image %s on unit %u.\n", name->buffer, i);
        GL_EXTCALL(glUniform1i(location, i));
    }
    checkGLcall("Load image bindings");
    string_buffer_release(&priv->string_buffers, name);
}
 
/* Context activation is done by the caller. */
static void shader_glsl_load_program_resources(const struct wined3d_context *context,
        struct shader_glsl_priv *priv, GLuint program_id, const struct wined3d_shader *shader)
{
    const struct wined3d_shader_reg_maps *reg_maps = &shader->reg_maps;
 
    shader_glsl_init_uniform_block_bindings(context->gl_info, priv, program_id, reg_maps);
    shader_glsl_load_icb(context->gl_info, priv, program_id, reg_maps);
    /* Texture unit mapping is set up to be the same each time the shader
     * program is used so we can hardcode the sampler uniform values. */
    shader_glsl_load_samplers(context, priv, program_id, reg_maps);
}
 
static void append_transform_feedback_varying(const char **varyings, unsigned int *varying_count,
        char **strings, unsigned int *strings_length, struct wined3d_string_buffer *buffer)
{
    if (varyings && *strings)
    {
        char *ptr = *strings;
 
        varyings[*varying_count] = ptr;
 
        memcpy(ptr, buffer->buffer, buffer->content_size + 1);
        ptr += buffer->content_size + 1;
 
        *strings = ptr;
    }
 
    *strings_length += buffer->content_size + 1;
    ++(*varying_count);
}
 
static void append_transform_feedback_skip_components(const char **varyings,
        unsigned int *varying_count, char **strings, unsigned int *strings_length,
        struct wined3d_string_buffer *buffer, unsigned int component_count)
{
    unsigned int j;
 
    for (j = 0; j < component_count / 4; ++j)
    {
        string_buffer_sprintf(buffer, "gl_SkipComponents4");
        append_transform_feedback_varying(varyings, varying_count, strings, strings_length, buffer);
    }
    if (component_count % 4)
    {
        string_buffer_sprintf(buffer, "gl_SkipComponents%u", component_count % 4);
        append_transform_feedback_varying(varyings, varying_count, strings, strings_length, buffer);
    }
}
 
static void shader_glsl_generate_transform_feedback_varyings(const struct wined3d_stream_output_desc *so_desc,
        struct wined3d_string_buffer *buffer, const char **varyings, unsigned int *varying_count,
        char *strings, unsigned int *strings_length, GLenum buffer_mode)
{
    unsigned int i, buffer_idx, count, length, highest_output_slot, stride;
 
    count = length = 0;
    highest_output_slot = 0;
    for (buffer_idx = 0; buffer_idx < WINED3D_MAX_STREAM_OUTPUT_BUFFERS; ++buffer_idx)
    {
        stride = 0;
 
        for (i = 0; i < so_desc->element_count; ++i)
        {
            const struct wined3d_stream_output_element *e = &so_desc->elements[i];
 
            highest_output_slot = max(highest_output_slot, e->output_slot);
            if (e->output_slot != buffer_idx)
                continue;
 
            if (e->stream_idx)
            {
                FIXME("Unhandled stream %u.\n", e->stream_idx);
                continue;
            }
 
            stride += e->component_count;
 
            if (e->register_idx == WINED3D_STREAM_OUTPUT_GAP)
            {
                append_transform_feedback_skip_components(varyings, &count,
                        &strings, &length, buffer, e->component_count);
                continue;
            }
 
            if (e->component_idx || e->component_count != 4)
            {
                if (so_desc->rasterizer_stream_idx != WINED3D_NO_RASTERIZER_STREAM)
                {
                    FIXME("Unsupported component range %u-%u.\n", e->component_idx, e->component_count);
                    append_transform_feedback_skip_components(varyings, &count,
                            &strings, &length, buffer, e->component_count);
                    continue;
                }
 
                string_buffer_sprintf(buffer, "shader_in_out.reg%u_%u_%u",
                        e->register_idx, e->component_idx, e->component_idx + e->component_count - 1);
                append_transform_feedback_varying(varyings, &count, &strings, &length, buffer);
            }
            else
            {
                string_buffer_sprintf(buffer, "shader_in_out.reg%u", e->register_idx);
                append_transform_feedback_varying(varyings, &count, &strings, &length, buffer);
            }
        }
 
        if (buffer_idx < so_desc->buffer_stride_count
                && stride < so_desc->buffer_strides[buffer_idx] / 4)
        {
            unsigned int component_count = so_desc->buffer_strides[buffer_idx] / 4 - stride;
            append_transform_feedback_skip_components(varyings, &count,
                    &strings, &length, buffer, component_count);
        }
 
        if (highest_output_slot <= buffer_idx)
            break;
 
        if (buffer_mode == GL_INTERLEAVED_ATTRIBS)
        {
            string_buffer_sprintf(buffer, "gl_NextBuffer");
            append_transform_feedback_varying(varyings, &count, &strings, &length, buffer);
        }
    }
 
    if (varying_count)
        *varying_count = count;
    if (strings_length)
        *strings_length = length;
}
 
static void shader_glsl_init_transform_feedback(const struct wined3d_context *context,
        struct shader_glsl_priv *priv, GLuint program_id, const struct wined3d_shader *shader)
{
    const struct wined3d_stream_output_desc *so_desc = &shader->u.gs.so_desc;
    const struct wined3d_gl_info *gl_info = context->gl_info;
    struct wined3d_string_buffer *buffer;
    unsigned int i, count, length;
    const char **varyings;
    char *strings;
    GLenum mode;
 
    if (!so_desc->element_count)
        return;
 
    if (gl_info->supported[ARB_TRANSFORM_FEEDBACK3])
    {
        mode = GL_INTERLEAVED_ATTRIBS;
    }
    else
    {
        unsigned int element_count[WINED3D_MAX_STREAM_OUTPUT_BUFFERS] = {0};
 
        for (i = 0; i < so_desc->element_count; ++i)
        {
            if (so_desc->elements[i].register_idx == WINED3D_STREAM_OUTPUT_GAP)
            {
                FIXME("ARB_transform_feedback3 is needed for stream output gaps.\n");
                return;
            }
            ++element_count[so_desc->elements[i].output_slot];
        }
 
        if (element_count[0] == so_desc->element_count)
        {
            mode = GL_INTERLEAVED_ATTRIBS;
        }
        else
        {
            mode = GL_SEPARATE_ATTRIBS;
            for (i = 0; i < ARRAY_SIZE(element_count); ++i)
            {
                if (element_count[i] != 1)
                    break;
            }
            for (; i < ARRAY_SIZE(element_count); ++i)
            {
                if (element_count[i])
                {
                    FIXME("Only single element per buffer is allowed in separate mode.\n");
                    return;
                }
            }
        }
    }
 
    buffer = string_buffer_get(&priv->string_buffers);
 
    shader_glsl_generate_transform_feedback_varyings(so_desc, buffer, NULL, &count, NULL, &length, mode);
 
    if (!(varyings = heap_calloc(count, sizeof(*varyings))))
    {
        ERR("Out of memory.\n");
        string_buffer_release(&priv->string_buffers, buffer);
        return;
    }
    if (!(strings = heap_calloc(length, sizeof(*strings))))
    {
        ERR("Out of memory.\n");
        heap_free(varyings);
        string_buffer_release(&priv->string_buffers, buffer);
        return;
    }
 
    shader_glsl_generate_transform_feedback_varyings(so_desc, buffer, varyings, NULL, strings, NULL, mode);
    GL_EXTCALL(glTransformFeedbackVaryings(program_id, count, varyings, mode));
    checkGLcall("glTransformFeedbackVaryings");
 
    heap_free(varyings);
    heap_free(strings);
    string_buffer_release(&priv->string_buffers, buffer);
}
 
/* Context activation is done by the caller. */
static inline void walk_constant_heap(const struct wined3d_gl_info *gl_info, const struct wined3d_vec4 *constants,
        const GLint *constant_locations, const struct constant_heap *heap, unsigned char *stack, DWORD version)
{
    unsigned int start = ~0U, end = 0;
    int stack_idx = 0;
    unsigned int heap_idx = 1;
    unsigned int idx;
 
    if (heap->entries[heap_idx].version <= version) return;
 
    idx = heap->entries[heap_idx].idx;
    if (constant_locations[idx] != -1)
        start = end = idx;
    stack[stack_idx] = HEAP_NODE_TRAVERSE_LEFT;
 
    while (stack_idx >= 0)
    {
        /* Note that we fall through to the next case statement. */
        switch(stack[stack_idx])
        {
            case HEAP_NODE_TRAVERSE_LEFT:
            {
                unsigned int left_idx = heap_idx << 1;
                if (left_idx < heap->size && heap->entries[left_idx].version > version)
                {
                    heap_idx = left_idx;
                    idx = heap->entries[heap_idx].idx;
                    if (constant_locations[idx] != -1)
                    {
                        if (start > idx)
                            start = idx;
                        if (end < idx)
                            end = idx;
                    }
 
                    stack[stack_idx++] = HEAP_NODE_TRAVERSE_RIGHT;
                    stack[stack_idx] = HEAP_NODE_TRAVERSE_LEFT;
                    break;
                }
            }
 
            case HEAP_NODE_TRAVERSE_RIGHT:
            {
                unsigned int right_idx = (heap_idx << 1) + 1;
                if (right_idx < heap->size && heap->entries[right_idx].version > version)
                {
                    heap_idx = right_idx;
                    idx = heap->entries[heap_idx].idx;
                    if (constant_locations[idx] != -1)
                    {
                        if (start > idx)
                            start = idx;
                        if (end < idx)
                            end = idx;
                    }
 
                    stack[stack_idx++] = HEAP_NODE_POP;
                    stack[stack_idx] = HEAP_NODE_TRAVERSE_LEFT;
                    break;
                }
            }
 
            case HEAP_NODE_POP:
                heap_idx >>= 1;
                --stack_idx;
                break;
        }
    }
    if (start <= end)
        GL_EXTCALL(glUniform4fv(constant_locations[start], end - start + 1, &constants[start].x));
    checkGLcall("walk_constant_heap()");
}
 
/* Context activation is done by the caller. */
static inline void apply_clamped_constant(const struct wined3d_gl_info *gl_info,
        GLint location, const struct wined3d_vec4 *data)
{
    GLfloat clamped_constant[4];
 
    if (location == -1) return;
 
    clamped_constant[0] = data->x < -1.0f ? -1.0f : data->x > 1.0f ? 1.0f : data->x;
    clamped_constant[1] = data->y < -1.0f ? -1.0f : data->y > 1.0f ? 1.0f : data->y;
    clamped_constant[2] = data->z < -1.0f ? -1.0f : data->z > 1.0f ? 1.0f : data->z;
    clamped_constant[3] = data->w < -1.0f ? -1.0f : data->w > 1.0f ? 1.0f : data->w;
 
    GL_EXTCALL(glUniform4fv(location, 1, clamped_constant));
}
 
/* Context activation is done by the caller. */
static inline void walk_constant_heap_clamped(const struct wined3d_gl_info *gl_info,
        const struct wined3d_vec4 *constants, const GLint *constant_locations,
        const struct constant_heap *heap, unsigned char *stack, DWORD version)
{
    int stack_idx = 0;
    unsigned int heap_idx = 1;
    unsigned int idx;
 
    if (heap->entries[heap_idx].version <= version) return;
 
    idx = heap->entries[heap_idx].idx;
    apply_clamped_constant(gl_info, constant_locations[idx], &constants[idx]);
    stack[stack_idx] = HEAP_NODE_TRAVERSE_LEFT;
 
    while (stack_idx >= 0)
    {
        /* Note that we fall through to the next case statement. */
        switch(stack[stack_idx])
        {
            case HEAP_NODE_TRAVERSE_LEFT:
            {
                unsigned int left_idx = heap_idx << 1;
                if (left_idx < heap->size && heap->entries[left_idx].version > version)
                {
                    heap_idx = left_idx;
                    idx = heap->entries[heap_idx].idx;
                    apply_clamped_constant(gl_info, constant_locations[idx], &constants[idx]);
 
                    stack[stack_idx++] = HEAP_NODE_TRAVERSE_RIGHT;
                    stack[stack_idx] = HEAP_NODE_TRAVERSE_LEFT;
                    break;
                }
            }
 
            case HEAP_NODE_TRAVERSE_RIGHT:
            {
                unsigned int right_idx = (heap_idx << 1) + 1;
                if (right_idx < heap->size && heap->entries[right_idx].version > version)
                {
                    heap_idx = right_idx;
                    idx = heap->entries[heap_idx].idx;
                    apply_clamped_constant(gl_info, constant_locations[idx], &constants[idx]);
 
                    stack[stack_idx++] = HEAP_NODE_POP;
                    stack[stack_idx] = HEAP_NODE_TRAVERSE_LEFT;
                    break;
                }
            }
 
            case HEAP_NODE_POP:
                heap_idx >>= 1;
                --stack_idx;
                break;
        }
    }
    checkGLcall("walk_constant_heap_clamped()");
}
 
/* Context activation is done by the caller. */
static void shader_glsl_load_constants_f(const struct wined3d_shader *shader, const struct wined3d_gl_info *gl_info,
        const struct wined3d_vec4 *constants, const GLint *constant_locations, const struct constant_heap *heap,
        unsigned char *stack, unsigned int version)
{
    const struct wined3d_shader_lconst *lconst;
 
    /* 1.X pshaders have the constants clamped to [-1;1] implicitly. */
    if (shader->reg_maps.shader_version.major == 1
            && shader->reg_maps.shader_version.type == WINED3D_SHADER_TYPE_PIXEL)
        walk_constant_heap_clamped(gl_info, constants, constant_locations, heap, stack, version);
    else
        walk_constant_heap(gl_info, constants, constant_locations, heap, stack, version);
 
    if (!shader->load_local_constsF)
    {
        TRACE("No need to load local float constants for this shader.\n");
        return;
    }
 
    /* Immediate constants are clamped to [-1;1] at shader creation time if needed */
    LIST_FOR_EACH_ENTRY(lconst, &shader->constantsF, struct wined3d_shader_lconst, entry)
    {
        GL_EXTCALL(glUniform4fv(constant_locations[lconst->idx], 1, (const GLfloat *)lconst->value));
    }
    checkGLcall("glUniform4fv()");
}
 
/* Context activation is done by the caller. */
static void shader_glsl_load_constants_i(const struct wined3d_shader *shader, const struct wined3d_gl_info *gl_info,
        const struct wined3d_ivec4 *constants, const GLint locations[WINED3D_MAX_CONSTS_I], WORD constants_set)
{
    unsigned int i;
    struct list* ptr;
 
    for (i = 0; constants_set; constants_set >>= 1, ++i)
    {
        if (!(constants_set & 1)) continue;
 
        /* We found this uniform name in the program - go ahead and send the data */
        GL_EXTCALL(glUniform4iv(locations[i], 1, &constants[i].x));
    }
 
    /* Load immediate constants */
    ptr = list_head(&shader->constantsI);
    while (ptr)
    {
        const struct wined3d_shader_lconst *lconst = LIST_ENTRY(ptr, const struct wined3d_shader_lconst, entry);
        unsigned int idx = lconst->idx;
        const GLint *values = (const GLint *)lconst->value;
 
        /* We found this uniform name in the program - go ahead and send the data */
        GL_EXTCALL(glUniform4iv(locations[idx], 1, values));
        ptr = list_next(&shader->constantsI, ptr);
    }
    checkGLcall("glUniform4iv()");
}
 
/* Context activation is done by the caller. */
static void shader_glsl_load_constantsB(const struct wined3d_shader *shader, const struct wined3d_gl_info *gl_info,
        const GLint locations[WINED3D_MAX_CONSTS_B], const BOOL *constants, WORD constants_set)
{
    unsigned int i;
    struct list* ptr;
 
    for (i = 0; constants_set; constants_set >>= 1, ++i)
    {
        if (!(constants_set & 1)) continue;
 
        GL_EXTCALL(glUniform1iv(locations[i], 1, &constants[i]));
    }
 
    /* Load immediate constants */
    ptr = list_head(&shader->constantsB);
    while (ptr)
    {
        const struct wined3d_shader_lconst *lconst = LIST_ENTRY(ptr, const struct wined3d_shader_lconst, entry);
        unsigned int idx = lconst->idx;
        const GLint *values = (const GLint *)lconst->value;
 
        GL_EXTCALL(glUniform1iv(locations[idx], 1, values));
        ptr = list_next(&shader->constantsB, ptr);
    }
    checkGLcall("glUniform1iv()");
}
 
static void reset_program_constant_version(struct wine_rb_entry *entry, void *context)
{
    WINE_RB_ENTRY_VALUE(entry, struct glsl_shader_prog_link, program_lookup_entry)->constant_version = 0;
}
 
/* Context activation is done by the caller (state handler). */
static void shader_glsl_load_np2fixup_constants(const struct glsl_ps_program *ps,
        const struct wined3d_gl_info *gl_info, const struct wined3d_state *state)
{
    struct
    {
        float sx, sy;
    }
    np2fixup_constants[MAX_FRAGMENT_SAMPLERS];
    UINT fixup = ps->np2_fixup_info->active;
    UINT i;
 
    for (i = 0; fixup; fixup >>= 1, ++i)
    {
        const struct wined3d_texture *tex = state->textures[i];
        unsigned char idx = ps->np2_fixup_info->idx[i];
 
        if (!tex)
        {
            ERR("Nonexistent texture is flagged for NP2 texcoord fixup.\n");
            continue;
        }
 
        np2fixup_constants[idx].sx = tex->pow2_matrix[0];
        np2fixup_constants[idx].sy = tex->pow2_matrix[5];
    }
 
    GL_EXTCALL(glUniform4fv(ps->np2_fixup_location, ps->np2_fixup_info->num_consts, &np2fixup_constants[0].sx));
}
 
static void shader_glsl_ffp_vertex_texmatrix_uniform(const struct wined3d_context *context,
        const struct wined3d_state *state, unsigned int tex, struct glsl_shader_prog_link *prog)
{
    const struct wined3d_gl_info *gl_info = context->gl_info;
    struct wined3d_matrix mat;
 
    if (tex >= MAX_TEXTURES)
        return;
    if (prog->vs.texture_matrix_location[tex] == -1)
        return;
 
    get_texture_matrix(context, state, tex, &mat);
    GL_EXTCALL(glUniformMatrix4fv(prog->vs.texture_matrix_location[tex], 1, FALSE, &mat._11));
    checkGLcall("glUniformMatrix4fv");
}
 
static void shader_glsl_ffp_vertex_material_uniform(const struct wined3d_context *context,
        const struct wined3d_state *state, struct glsl_shader_prog_link *prog)
{
    const struct wined3d_gl_info *gl_info = context->gl_info;
 
    if (state->render_states[WINED3D_RS_SPECULARENABLE])
    {
        GL_EXTCALL(glUniform4fv(prog->vs.material_specular_location, 1, &state->material.specular.r));
        GL_EXTCALL(glUniform1f(prog->vs.material_shininess_location, state->material.power));
    }
    else
    {
        static const float black[] = {0.0f, 0.0f, 0.0f, 0.0f};
 
        GL_EXTCALL(glUniform4fv(prog->vs.material_specular_location, 1, black));
    }
    GL_EXTCALL(glUniform4fv(prog->vs.material_ambient_location, 1, &state->material.ambient.r));
    GL_EXTCALL(glUniform4fv(prog->vs.material_diffuse_location, 1, &state->material.diffuse.r));
    GL_EXTCALL(glUniform4fv(prog->vs.material_emissive_location, 1, &state->material.emissive.r));
    checkGLcall("setting FFP material uniforms");
}
 
static void shader_glsl_ffp_vertex_lightambient_uniform(const struct wined3d_context *context,
        const struct wined3d_state *state, struct glsl_shader_prog_link *prog)
{
    const struct wined3d_gl_info *gl_info = context->gl_info;
    struct wined3d_color color;
 
    wined3d_color_from_d3dcolor(&color, state->render_states[WINED3D_RS_AMBIENT]);
    GL_EXTCALL(glUniform3fv(prog->vs.light_ambient_location, 1, &color.r));
    checkGLcall("glUniform3fv");
}
 
static void multiply_vector_matrix(struct wined3d_vec4 *dest, const struct wined3d_vec4 *src1,
        const struct wined3d_matrix *src2)
{
    struct wined3d_vec4 temp;
 
    temp.x = (src1->x * src2->_11) + (src1->y * src2->_21) + (src1->z * src2->_31) + (src1->w * src2->_41);
    temp.y = (src1->x * src2->_12) + (src1->y * src2->_22) + (src1->z * src2->_32) + (src1->w * src2->_42);
    temp.z = (src1->x * src2->_13) + (src1->y * src2->_23) + (src1->z * src2->_33) + (src1->w * src2->_43);
    temp.w = (src1->x * src2->_14) + (src1->y * src2->_24) + (src1->z * src2->_34) + (src1->w * src2->_44);
 
    *dest = temp;
}
 
static void shader_glsl_ffp_vertex_light_uniform(const struct wined3d_context *context,
        const struct wined3d_state *state, unsigned int light, const struct wined3d_light_info *light_info,
        struct glsl_shader_prog_link *prog)
{
    const struct wined3d_matrix *view = &state->transforms[WINED3D_TS_VIEW];
    const struct wined3d_gl_info *gl_info = context->gl_info;
    struct wined3d_vec4 vec4;
 
    GL_EXTCALL(glUniform4fv(prog->vs.light_location[light].diffuse, 1, &light_info->OriginalParms.diffuse.r));
    GL_EXTCALL(glUniform4fv(prog->vs.light_location[light].specular, 1, &light_info->OriginalParms.specular.r));
    GL_EXTCALL(glUniform4fv(prog->vs.light_location[light].ambient, 1, &light_info->OriginalParms.ambient.r));
 
    switch (light_info->OriginalParms.type)
    {
        case WINED3D_LIGHT_POINT:
            multiply_vector_matrix(&vec4, &light_info->position, view);
            GL_EXTCALL(glUniform4fv(prog->vs.light_location[light].position, 1, &vec4.x));
            GL_EXTCALL(glUniform1f(prog->vs.light_location[light].range, light_info->OriginalParms.range));
            GL_EXTCALL(glUniform1f(prog->vs.light_location[light].c_att, light_info->OriginalParms.attenuation0));
            GL_EXTCALL(glUniform1f(prog->vs.light_location[light].l_att, light_info->OriginalParms.attenuation1));
            GL_EXTCALL(glUniform1f(prog->vs.light_location[light].q_att, light_info->OriginalParms.attenuation2));
            break;
 
        case WINED3D_LIGHT_SPOT:
            multiply_vector_matrix(&vec4, &light_info->position, view);
            GL_EXTCALL(glUniform4fv(prog->vs.light_location[light].position, 1, &vec4.x));
 
            multiply_vector_matrix(&vec4, &light_info->direction, view);
            GL_EXTCALL(glUniform3fv(prog->vs.light_location[light].direction, 1, &vec4.x));
 
            GL_EXTCALL(glUniform1f(prog->vs.light_location[light].range, light_info->OriginalParms.range));
            GL_EXTCALL(glUniform1f(prog->vs.light_location[light].falloff, light_info->OriginalParms.falloff));
            GL_EXTCALL(glUniform1f(prog->vs.light_location[light].c_att, light_info->OriginalParms.attenuation0));
            GL_EXTCALL(glUniform1f(prog->vs.light_location[light].l_att, light_info->OriginalParms.attenuation1));
            GL_EXTCALL(glUniform1f(prog->vs.light_location[light].q_att, light_info->OriginalParms.attenuation2));
            GL_EXTCALL(glUniform1f(prog->vs.light_location[light].cos_htheta, cosf(light_info->OriginalParms.theta / 2.0f)));
            GL_EXTCALL(glUniform1f(prog->vs.light_location[light].cos_hphi, cosf(light_info->OriginalParms.phi / 2.0f)));
            break;
 
        case WINED3D_LIGHT_DIRECTIONAL:
            multiply_vector_matrix(&vec4, &light_info->direction, view);
            GL_EXTCALL(glUniform3fv(prog->vs.light_location[light].direction, 1, &vec4.x));
            break;
 
        case WINED3D_LIGHT_PARALLELPOINT:
            multiply_vector_matrix(&vec4, &light_info->position, view);
            GL_EXTCALL(glUniform4fv(prog->vs.light_location[light].position, 1, &vec4.x));
            break;
 
        default:
            FIXME("Unrecognized light type %#x.\n", light_info->OriginalParms.type);
    }
    checkGLcall("setting FFP lights uniforms");
}
 
static void shader_glsl_pointsize_uniform(const struct wined3d_context *context,
        const struct wined3d_state *state, struct glsl_shader_prog_link *prog)
{
    const struct wined3d_gl_info *gl_info = context->gl_info;
    float min, max;
    float size, att[3];
 
    get_pointsize_minmax(context, state, &min, &max);
 
    GL_EXTCALL(glUniform1f(prog->vs.pointsize_min_location, min));
    checkGLcall("glUniform1f");
    GL_EXTCALL(glUniform1f(prog->vs.pointsize_max_location, max));
    checkGLcall("glUniform1f");
 
    get_pointsize(context, state, &size, att);
 
    GL_EXTCALL(glUniform1f(prog->vs.pointsize_location, size));
    checkGLcall("glUniform1f");
    GL_EXTCALL(glUniform1f(prog->vs.pointsize_c_att_location, att[0]));
    checkGLcall("glUniform1f");
    GL_EXTCALL(glUniform1f(prog->vs.pointsize_l_att_location, att[1]));
    checkGLcall("glUniform1f");
    GL_EXTCALL(glUniform1f(prog->vs.pointsize_q_att_location, att[2]));
    checkGLcall("glUniform1f");
}
 
static void shader_glsl_load_fog_uniform(const struct wined3d_context *context,
        const struct wined3d_state *state, struct glsl_shader_prog_link *prog)
{
    const struct wined3d_gl_info *gl_info = context->gl_info;
    struct wined3d_color color;
    float start, end, scale;
    union
    {
        DWORD d;
        float f;
    } tmpvalue;
 
    wined3d_color_from_d3dcolor(&color, state->render_states[WINED3D_RS_FOGCOLOR]);
    GL_EXTCALL(glUniform4fv(prog->ps.fog_color_location, 1, &color.r));
    tmpvalue.d = state->render_states[WINED3D_RS_FOGDENSITY];
    GL_EXTCALL(glUniform1f(prog->ps.fog_density_location, tmpvalue.f));
    get_fog_start_end(context, state, &start, &end);
    scale = 1.0f / (end - start);
    GL_EXTCALL(glUniform1f(prog->ps.fog_end_location, end));
    GL_EXTCALL(glUniform1f(prog->ps.fog_scale_location, scale));
    checkGLcall("fog emulation uniforms");
}
 
static void shader_glsl_clip_plane_uniform(const struct wined3d_context *context,
        const struct wined3d_state *state, unsigned int index, struct glsl_shader_prog_link *prog)
{
    const struct wined3d_gl_info *gl_info = context->gl_info;
    struct wined3d_matrix matrix;
    struct wined3d_vec4 plane;
 
    plane = state->clip_planes[index];
 
    /* Clip planes are affected by the view transform in d3d for FFP draws. */
    if (!use_vs(state))
    {
        invert_matrix(&matrix, &state->transforms[WINED3D_TS_VIEW]);
        transpose_matrix(&matrix, &matrix);
        multiply_vector_matrix(&plane, &plane, &matrix);
    }
 
    GL_EXTCALL(glUniform4fv(prog->vs.clip_planes_location + index, 1, &plane.x));
}
 
/* Context activation is done by the caller (state handler). */
static void shader_glsl_load_color_key_constant(const struct glsl_ps_program *ps,
        const struct wined3d_gl_info *gl_info, const struct wined3d_state *state)
{
    struct wined3d_color float_key[2];
    const struct wined3d_texture *texture = state->textures[0];
 
    wined3d_format_get_float_color_key(texture->resource.format, &texture->async.src_blt_color_key, float_key);
    GL_EXTCALL(glUniform4fv(ps->color_key_location, 2, &float_key[0].r));
}
 
/* Context activation is done by the caller. */
static void get_normal_matrix(struct wined3d_context *context, struct wined3d_matrix *mat, float *normal)
{
    int i, j;
 
    if (context->d3d_info->wined3d_creation_flags & WINED3D_LEGACY_FFP_LIGHTING)
        invert_matrix_3d(mat, mat);
    else
        invert_matrix(mat, mat);
    /* Tests show that singular modelview matrices are used unchanged as normal
     * matrices on D3D3 and older. There seems to be no clearly consistent
     * behavior on newer D3D versions so always follow older ddraw behavior. */
    for (i = 0; i < 3; ++i)
        for (j = 0; j < 3; ++j)
            normal[i * 3 + j] = (&mat->_11)[j * 4 + i];
}
 
/* Context activation is done by the caller (state handler). */
static void shader_glsl_load_constants(void *shader_priv, struct wined3d_context *context,
        const struct wined3d_state *state)
{
    const struct glsl_context_data *ctx_data = context->shader_backend_data;
    const struct wined3d_shader *vshader = state->shader[WINED3D_SHADER_TYPE_VERTEX];
    const struct wined3d_shader *pshader = state->shader[WINED3D_SHADER_TYPE_PIXEL];
    const struct wined3d_gl_info *gl_info = context->gl_info;
    struct shader_glsl_priv *priv = shader_priv;
    float position_fixup[4];
    float normal[3 * 3];
    DWORD update_mask;
 
    struct glsl_shader_prog_link *prog = ctx_data->glsl_program;
    UINT constant_version;
    int i;
 
    if (!prog) {
        /* No GLSL program set - nothing to do. */
        return;
    }
    constant_version = prog->constant_version;
    update_mask = context->constant_update_mask & prog->constant_update_mask;
 
    if (update_mask & WINED3D_SHADER_CONST_VS_F)
        shader_glsl_load_constants_f(vshader, gl_info, state->vs_consts_f,
                prog->vs.uniform_f_locations, &priv->vconst_heap, priv->stack, constant_version);
 
    if (update_mask & WINED3D_SHADER_CONST_VS_I)
        shader_glsl_load_constants_i(vshader, gl_info, state->vs_consts_i,
                prog->vs.uniform_i_locations, vshader->reg_maps.integer_constants);
 
    if (update_mask & WINED3D_SHADER_CONST_VS_B)
        shader_glsl_load_constantsB(vshader, gl_info, prog->vs.uniform_b_locations, state->vs_consts_b,
                vshader->reg_maps.boolean_constants);
 
    if (update_mask & WINED3D_SHADER_CONST_VS_CLIP_PLANES)
    {
        for (i = 0; i < gl_info->limits.user_clip_distances; ++i)
            shader_glsl_clip_plane_uniform(context, state, i, prog);
    }
 
    if (update_mask & WINED3D_SHADER_CONST_VS_POINTSIZE)
        shader_glsl_pointsize_uniform(context, state, prog);
 
    if (update_mask & WINED3D_SHADER_CONST_POS_FIXUP)
    {
        shader_get_position_fixup(context, state, position_fixup);
        if (state->shader[WINED3D_SHADER_TYPE_GEOMETRY])
            GL_EXTCALL(glUniform4fv(prog->gs.pos_fixup_location, 1, position_fixup));
        else if (state->shader[WINED3D_SHADER_TYPE_DOMAIN])
            GL_EXTCALL(glUniform4fv(prog->ds.pos_fixup_location, 1, position_fixup));
        else
            GL_EXTCALL(glUniform4fv(prog->vs.pos_fixup_location, 1, position_fixup));
        checkGLcall("glUniform4fv");
    }
 
    if (update_mask & WINED3D_SHADER_CONST_FFP_MODELVIEW)
    {
        struct wined3d_matrix mat;
 
        get_modelview_matrix(context, state, 0, &mat);
        GL_EXTCALL(glUniformMatrix4fv(prog->vs.modelview_matrix_location[0], 1, FALSE, &mat._11));
        checkGLcall("glUniformMatrix4fv");
 
        get_normal_matrix(context, &mat, normal);
        GL_EXTCALL(glUniformMatrix3fv(prog->vs.normal_matrix_location[0], 1, FALSE, normal));
        checkGLcall("glUniformMatrix3fv");
    }
 
    if (update_mask & WINED3D_SHADER_CONST_FFP_VERTEXBLEND)
    {
        struct wined3d_matrix mat;
 
        for (i = 1; i < MAX_VERTEX_INDEX_BLENDS; ++i)
        {
            if (prog->vs.modelview_matrix_location[i] == -1)
                break;
            if (!(update_mask & WINED3D_SHADER_CONST_FFP_VERTEXBLEND_INDEX(i)))
                continue;
 
            get_modelview_matrix(context, state, i, &mat);
            GL_EXTCALL(glUniformMatrix4fv(prog->vs.modelview_matrix_location[i], 1, FALSE, &mat._11));
            checkGLcall("glUniformMatrix4fv");
 
            get_normal_matrix(context, &mat, normal);
            GL_EXTCALL(glUniformMatrix3fv(prog->vs.normal_matrix_location[i], 1, FALSE, normal));
            checkGLcall("glUniformMatrix3fv");
        }
    }
 
    if (update_mask & WINED3D_SHADER_CONST_FFP_PROJ)
    {
        struct wined3d_matrix projection;
 
        get_projection_matrix(context, state, &projection);
        GL_EXTCALL(glUniformMatrix4fv(prog->vs.projection_matrix_location, 1, FALSE, &projection._11));
        checkGLcall("glUniformMatrix4fv");
    }
 
    if (update_mask & WINED3D_SHADER_CONST_FFP_TEXMATRIX)
    {
        for (i = 0; i < MAX_TEXTURES; ++i)
            shader_glsl_ffp_vertex_texmatrix_uniform(context, state, i, prog);
    }
 
    if (update_mask & WINED3D_SHADER_CONST_FFP_MATERIAL)
        shader_glsl_ffp_vertex_material_uniform(context, state, prog);
 
    if (update_mask & WINED3D_SHADER_CONST_FFP_LIGHTS)
    {
        unsigned int point_idx, spot_idx, directional_idx, parallel_point_idx;
        DWORD point_count = 0;
        DWORD spot_count = 0;
        DWORD directional_count = 0;
        DWORD parallel_point_count = 0;
 
        for (i = 0; i < MAX_ACTIVE_LIGHTS; ++i)
        {
            if (!state->lights[i])
                continue;
 
            switch (state->lights[i]->OriginalParms.type)
            {
                case WINED3D_LIGHT_POINT:
                    ++point_count;
                    break;
                case WINED3D_LIGHT_SPOT:
                    ++spot_count;
                    break;
                case WINED3D_LIGHT_DIRECTIONAL:
                    ++directional_count;
                    break;
                case WINED3D_LIGHT_PARALLELPOINT:
                    ++parallel_point_count;
                    break;
                default:
                    FIXME("Unhandled light type %#x.\n", state->lights[i]->OriginalParms.type);
                    break;
            }
        }
        point_idx = 0;
        spot_idx = point_idx + point_count;
        directional_idx = spot_idx + spot_count;
        parallel_point_idx = directional_idx + directional_count;
 
        shader_glsl_ffp_vertex_lightambient_uniform(context, state, prog);
        for (i = 0; i < MAX_ACTIVE_LIGHTS; ++i)
        {
            const struct wined3d_light_info *light_info = state->lights[i];
            unsigned int idx;
 
            if (!light_info)
                continue;
 
            switch (light_info->OriginalParms.type)
            {
                case WINED3D_LIGHT_POINT:
                    idx = point_idx++;
                    break;
                case WINED3D_LIGHT_SPOT:
                    idx = spot_idx++;
                    break;
                case WINED3D_LIGHT_DIRECTIONAL:
                    idx = directional_idx++;
                    break;
                case WINED3D_LIGHT_PARALLELPOINT:
                    idx = parallel_point_idx++;
                    break;
                default:
                    FIXME("Unhandled light type %#x.\n", light_info->OriginalParms.type);
                    continue;
            }
            shader_glsl_ffp_vertex_light_uniform(context, state, idx, light_info, prog);
        }
    }
 
    if (update_mask & WINED3D_SHADER_CONST_PS_F)
        shader_glsl_load_constants_f(pshader, gl_info, state->ps_consts_f,
                prog->ps.uniform_f_locations, &priv->pconst_heap, priv->stack, constant_version);
 
    if (update_mask & WINED3D_SHADER_CONST_PS_I)
        shader_glsl_load_constants_i(pshader, gl_info, state->ps_consts_i,
                prog->ps.uniform_i_locations, pshader->reg_maps.integer_constants);
 
    if (update_mask & WINED3D_SHADER_CONST_PS_B)
        shader_glsl_load_constantsB(pshader, gl_info, prog->ps.uniform_b_locations, state->ps_consts_b,
                pshader->reg_maps.boolean_constants);
 
    if (update_mask & WINED3D_SHADER_CONST_PS_BUMP_ENV)
    {
        for (i = 0; i < MAX_TEXTURES; ++i)
        {
            if (prog->ps.bumpenv_mat_location[i] == -1)
                continue;
 
            GL_EXTCALL(glUniformMatrix2fv(prog->ps.bumpenv_mat_location[i], 1, 0,
                    (const GLfloat *)&state->texture_states[i][WINED3D_TSS_BUMPENV_MAT00]));
 
            if (prog->ps.bumpenv_lum_scale_location[i] != -1)
            {
                GL_EXTCALL(glUniform1fv(prog->ps.bumpenv_lum_scale_location[i], 1,
                        (const GLfloat *)&state->texture_states[i][WINED3D_TSS_BUMPENV_LSCALE]));
                GL_EXTCALL(glUniform1fv(prog->ps.bumpenv_lum_offset_location[i], 1,
                        (const GLfloat *)&state->texture_states[i][WINED3D_TSS_BUMPENV_LOFFSET]));
            }
        }
 
        checkGLcall("bump env uniforms");
    }
 
    if (update_mask & WINED3D_SHADER_CONST_PS_Y_CORR)
    {
        const struct wined3d_vec4 correction_params =
        {
            /* Position is relative to the framebuffer, not the viewport. */
            context->render_offscreen ? 0.0f : (float)state->fb->render_targets[0]->height,
            context->render_offscreen ? 1.0f : -1.0f,
            0.0f,
            0.0f,
        };
 
        GL_EXTCALL(glUniform4fv(prog->ps.ycorrection_location, 1, &correction_params.x));
    }
 
    if (update_mask & WINED3D_SHADER_CONST_PS_NP2_FIXUP)
        shader_glsl_load_np2fixup_constants(&prog->ps, gl_info, state);
    if (update_mask & WINED3D_SHADER_CONST_FFP_COLOR_KEY)
        shader_glsl_load_color_key_constant(&prog->ps, gl_info, state);
 
    if (update_mask & WINED3D_SHADER_CONST_FFP_PS)
    {
        struct wined3d_color color;
 
        if (prog->ps.tex_factor_location != -1)
        {
            wined3d_color_from_d3dcolor(&color, state->render_states[WINED3D_RS_TEXTUREFACTOR]);
            GL_EXTCALL(glUniform4fv(prog->ps.tex_factor_location, 1, &color.r));
        }
 
        if (state->render_states[WINED3D_RS_SPECULARENABLE])
            GL_EXTCALL(glUniform4f(prog->ps.specular_enable_location, 1.0f, 1.0f, 1.0f, 0.0f));
        else
            GL_EXTCALL(glUniform4f(prog->ps.specular_enable_location, 0.0f, 0.0f, 0.0f, 0.0f));
 
        for (i = 0; i < MAX_TEXTURES; ++i)
        {
            if (prog->ps.tss_constant_location[i] == -1)
                continue;
 
            wined3d_color_from_d3dcolor(&color, state->texture_states[i][WINED3D_TSS_CONSTANT]);
            GL_EXTCALL(glUniform4fv(prog->ps.tss_constant_location[i], 1, &color.r));
        }
 
        checkGLcall("fixed function uniforms");
    }
 
    if (update_mask & WINED3D_SHADER_CONST_PS_FOG)
        shader_glsl_load_fog_uniform(context, state, prog);
 
    if (update_mask & WINED3D_SHADER_CONST_PS_ALPHA_TEST)
    {
        float ref = state->render_states[WINED3D_RS_ALPHAREF] / 255.0f;
 
        GL_EXTCALL(glUniform1f(prog->ps.alpha_test_ref_location, ref));
        checkGLcall("alpha test emulation uniform");
    }
 
    if (priv->next_constant_version == UINT_MAX)
    {
        TRACE("Max constant version reached, resetting to 0.\n");
        wine_rb_for_each_entry(&priv->program_lookup, reset_program_constant_version, NULL);
        priv->next_constant_version = 1;
    }
    else
    {
        prog->constant_version = priv->next_constant_version++;
    }
}
 
static void update_heap_entry(struct constant_heap *heap, unsigned int idx, DWORD new_version)
{
    struct constant_entry *entries = heap->entries;
    unsigned int *positions = heap->positions;
    unsigned int heap_idx, parent_idx;
 
    if (!heap->contained[idx])
    {
        heap_idx = heap->size++;
        heap->contained[idx] = TRUE;
    }
    else
    {
        heap_idx = positions[idx];
    }
 
    while (heap_idx > 1)
    {
        parent_idx = heap_idx >> 1;
 
        if (new_version <= entries[parent_idx].version) break;
 
        entries[heap_idx] = entries[parent_idx];
        positions[entries[parent_idx].idx] = heap_idx;
        heap_idx = parent_idx;
    }
 
    entries[heap_idx].version = new_version;
    entries[heap_idx].idx = idx;
    positions[idx] = heap_idx;
}
 
static void shader_glsl_update_float_vertex_constants(struct wined3d_device *device, UINT start, UINT count)
{
    struct shader_glsl_priv *priv = device->shader_priv;
    struct constant_heap *heap = &priv->vconst_heap;
    UINT i;
 
    for (i = start; i < count + start; ++i)
    {
        update_heap_entry(heap, i, priv->next_constant_version);
    }
}
 
static void shader_glsl_update_float_pixel_constants(struct wined3d_device *device, UINT start, UINT count)
{
    struct shader_glsl_priv *priv = device->shader_priv;
    struct constant_heap *heap = &priv->pconst_heap;
    UINT i;
 
    for (i = start; i < count + start; ++i)
    {
        update_heap_entry(heap, i, priv->next_constant_version);
    }
}
 
static unsigned int vec4_varyings(DWORD shader_major, const struct wined3d_gl_info *gl_info)
{
    unsigned int ret = gl_info->limits.glsl_varyings / 4;
    /* 4.0 shaders do not write clip coords because d3d10 does not support user clipplanes */
    if(shader_major > 3) return ret;
 
    /* 3.0 shaders may need an extra varying for the clip coord on some cards(mostly dx10 ones) */
    if (gl_info->quirks & WINED3D_QUIRK_GLSL_CLIP_VARYING) ret -= 1;
    return ret;
}
 
static BOOL needs_legacy_glsl_syntax(const struct wined3d_gl_info *gl_info)
{
    return gl_info->glsl_version < MAKEDWORD_VERSION(1, 30);
}
 
static BOOL shader_glsl_use_explicit_attrib_location(const struct wined3d_gl_info *gl_info)
{
    return gl_info->supported[ARB_EXPLICIT_ATTRIB_LOCATION]
            && shader_glsl_use_layout_qualifier(gl_info) && !needs_legacy_glsl_syntax(gl_info);
}
 
static BOOL shader_glsl_use_interface_blocks(const struct wined3d_gl_info *gl_info)
{
    return shader_glsl_get_version(gl_info) >= 150;
}
 
static const char *get_attribute_keyword(const struct wined3d_gl_info *gl_info)
{
    return needs_legacy_glsl_syntax(gl_info) ? "attribute" : "in";
}
 
static void PRINTF_ATTR(4, 5) declare_in_varying(const struct wined3d_gl_info *gl_info,
        struct wined3d_string_buffer *buffer, BOOL flat, const char *format, ...)
{
    va_list args;
    int ret;
 
    shader_addline(buffer, "%s%s ", flat ? "flat " : "",
            needs_legacy_glsl_syntax(gl_info) ? "varying" : "in");
    for (;;)
    {
        va_start(args, format);
        ret = shader_vaddline(buffer, format, args);
        va_end(args);
        if (!ret)
            return;
        if (!string_buffer_resize(buffer, ret))
            return;
    }
}
 
static void PRINTF_ATTR(4, 5) declare_out_varying(const struct wined3d_gl_info *gl_info,
        struct wined3d_string_buffer *buffer, BOOL flat, const char *format, ...)
{
    va_list args;
    int ret;
 
    shader_addline(buffer, "%s%s ", flat ? "flat " : "",
            needs_legacy_glsl_syntax(gl_info) ? "varying" : "out");
    for (;;)
    {
        va_start(args, format);
        ret = shader_vaddline(buffer, format, args);
        va_end(args);
        if (!ret)
            return;
        if (!string_buffer_resize(buffer, ret))
            return;
    }
}
 
static const char *shader_glsl_shader_input_name(const struct wined3d_gl_info *gl_info)
{
    return shader_glsl_use_interface_blocks(gl_info) ? "shader_in.reg" : "ps_link";
}
 
static const char *shader_glsl_shader_output_name(const struct wined3d_gl_info *gl_info)
{
    return shader_glsl_use_interface_blocks(gl_info) ? "shader_out.reg" : "ps_link";
}
 
static const char *shader_glsl_interpolation_qualifiers(enum wined3d_shader_interpolation_mode mode)
{
    switch (mode)
    {
        case WINED3DSIM_CONSTANT:
            return "flat";
        case WINED3DSIM_LINEAR_NOPERSPECTIVE:
            return "noperspective";
        default:
            FIXME("Unhandled interpolation mode %#x.\n", mode);
        case WINED3DSIM_NONE:
        case WINED3DSIM_LINEAR:
            return "";
    }
}
 
static enum wined3d_shader_interpolation_mode wined3d_extract_interpolation_mode(
        const DWORD *packed_interpolation_mode, unsigned int register_idx)
{
    return wined3d_extract_bits(packed_interpolation_mode,
            register_idx * WINED3D_PACKED_INTERPOLATION_BIT_COUNT, WINED3D_PACKED_INTERPOLATION_BIT_COUNT);
}
 
static void shader_glsl_declare_shader_inputs(const struct wined3d_gl_info *gl_info,
        struct wined3d_string_buffer *buffer, unsigned int element_count,
        const DWORD *interpolation_mode, BOOL unroll)
{
    enum wined3d_shader_interpolation_mode mode;
    unsigned int i;
 
    if (shader_glsl_use_interface_blocks(gl_info))
    {
        if (unroll)
        {
            shader_addline(buffer, "in shader_in_out {\n");
            for (i = 0; i < element_count; ++i)
            {
                mode = wined3d_extract_interpolation_mode(interpolation_mode, i);
                shader_addline(buffer, "%s vec4 reg%u;\n", shader_glsl_interpolation_qualifiers(mode), i);
            }
            shader_addline(buffer, "} shader_in;\n");
        }
        else
        {
            shader_addline(buffer, "in shader_in_out { vec4 reg[%u]; } shader_in;\n", element_count);
        }
    }
    else
    {
        declare_in_varying(gl_info, buffer, FALSE, "vec4 ps_link[%u];\n", element_count);
    }
}
 
static void shader_glsl_declare_shader_outputs(const struct wined3d_gl_info *gl_info,
        struct wined3d_string_buffer *buffer, unsigned int element_count, BOOL rasterizer_setup,
        const DWORD *interpolation_mode)
{
    enum wined3d_shader_interpolation_mode mode;
    unsigned int i;
 
    if (shader_glsl_use_interface_blocks(gl_info))
    {
        if (rasterizer_setup)
        {
            shader_addline(buffer, "out shader_in_out {\n");
            for (i = 0; i < element_count; ++i)
            {
                const char *interpolation_qualifiers = "";
                if (needs_interpolation_qualifiers_for_shader_outputs(gl_info))
                {
                    mode = wined3d_extract_interpolation_mode(interpolation_mode, i);
                    interpolation_qualifiers = shader_glsl_interpolation_qualifiers(mode);
                }
                shader_addline(buffer, "%s vec4 reg%u;\n", interpolation_qualifiers, i);
            }
            shader_addline(buffer, "} shader_out;\n");
        }
        else
        {
            shader_addline(buffer, "out shader_in_out { vec4 reg[%u]; } shader_out;\n", element_count);
        }
    }
    else
    {
        declare_out_varying(gl_info, buffer, FALSE, "vec4 ps_link[%u];\n", element_count);
    }
}
 
static const char *glsl_primitive_type_from_d3d(enum wined3d_primitive_type primitive_type)
{
    switch (primitive_type)
    {
        case WINED3D_PT_POINTLIST:
            return "points";
 
        case WINED3D_PT_LINELIST:
            return "lines";
 
        case WINED3D_PT_LINESTRIP:
            return "line_strip";
 
        case WINED3D_PT_TRIANGLELIST:
            return "triangles";
 
        case WINED3D_PT_TRIANGLESTRIP:
            return "triangle_strip";
 
        case WINED3D_PT_LINELIST_ADJ:
            return "lines_adjacency";
 
        case WINED3D_PT_TRIANGLELIST_ADJ:
            return "triangles_adjacency";
 
        default:
            FIXME("Unhandled primitive type %s.\n", debug_d3dprimitivetype(primitive_type));
            return "";
    }
}
 
static BOOL glsl_is_color_reg_read(const struct wined3d_shader *shader, unsigned int idx)
{
    const struct wined3d_shader_signature *input_signature = &shader->input_signature;
    const struct wined3d_shader_reg_maps *reg_maps = &shader->reg_maps;
    DWORD input_reg_used = shader->u.ps.input_reg_used;
    unsigned int i;
 
    if (reg_maps->shader_version.major < 3)
        return input_reg_used & (1u << idx);
 
    for (i = 0; i < input_signature->element_count; ++i)
    {
        const struct wined3d_shader_signature_element *input = &input_signature->elements[i];
 
        if (!(reg_maps->input_registers & (1u << input->register_idx)))
            continue;
 
        if (shader_match_semantic(input->semantic_name, WINED3D_DECL_USAGE_COLOR)
                && input->semantic_idx == idx)
            return input_reg_used & (1u << input->register_idx);
    }
    return FALSE;
}
 
static BOOL glsl_is_shadow_sampler(const struct wined3d_shader *shader,
        const struct ps_compile_args *ps_args, unsigned int resource_idx, unsigned int sampler_idx)
{
    const struct wined3d_shader_version *version = &shader->reg_maps.shader_version;
 
    if (version->major >= 4)
        return shader->reg_maps.sampler_comparison_mode & (1u << sampler_idx);
    else
        return version->type == WINED3D_SHADER_TYPE_PIXEL && (ps_args->shadow & (1u << resource_idx));
}
 
static void shader_glsl_declare_typed_vertex_attribute(struct wined3d_string_buffer *buffer,
        const struct wined3d_gl_info *gl_info, const char *vector_type, const char *scalar_type,
        unsigned int index)
{
    shader_addline(buffer, "%s %s4 vs_in_%s%u;\n",
            get_attribute_keyword(gl_info), vector_type, scalar_type, index);
    shader_addline(buffer, "vec4 vs_in%u = %sBitsToFloat(vs_in_%s%u);\n",
            index, scalar_type, scalar_type, index);
}
 
static void shader_glsl_declare_generic_vertex_attribute(struct wined3d_string_buffer *buffer,
        const struct wined3d_gl_info *gl_info, const struct wined3d_shader_signature_element *e)
{
    unsigned int index = e->register_idx;
 
    if (e->sysval_semantic == WINED3D_SV_VERTEX_ID)
    {
        shader_addline(buffer, "vec4 vs_in%u = vec4(intBitsToFloat(gl_VertexID), 0.0, 0.0, 0.0);\n",
                index);
        return;
    }
    if (e->sysval_semantic == WINED3D_SV_INSTANCE_ID)
    {
        shader_addline(buffer, "vec4 vs_in%u = vec4(intBitsToFloat(gl_InstanceID), 0.0, 0.0, 0.0);\n",
                index);
        return;
    }
    if (e->sysval_semantic && e->sysval_semantic != WINED3D_SV_POSITION)
        FIXME("Unhandled sysval semantic %#x.\n", e->sysval_semantic);
 
    if (shader_glsl_use_explicit_attrib_location(gl_info))
        shader_addline(buffer, "layout(location = %u) ", index);
 
    switch (e->component_type)
    {
        case WINED3D_TYPE_UINT:
            shader_glsl_declare_typed_vertex_attribute(buffer, gl_info, "uvec", "uint", index);
            break;
        case WINED3D_TYPE_INT:
            shader_glsl_declare_typed_vertex_attribute(buffer, gl_info, "ivec", "int", index);
            break;
 
        default:
            FIXME("Unhandled type %#x.\n", e->component_type);
            /* Fall through. */
        case WINED3D_TYPE_UNKNOWN:
        case WINED3D_TYPE_FLOAT:
            shader_addline(buffer, "%s vec4 vs_in%u;\n", get_attribute_keyword(gl_info), index);
            break;
    }
}
 
static void shader_generate_glsl_declarations(const struct wined3d_context *context,
        struct wined3d_string_buffer *buffer, const struct wined3d_shader *shader,
        const struct wined3d_shader_reg_maps *reg_maps, const struct shader_glsl_ctx_priv *ctx_priv)
{
    const struct wined3d_shader_version *version = &reg_maps->shader_version;
    const struct vs_compile_args *vs_args = ctx_priv->cur_vs_args;
    const struct ps_compile_args *ps_args = ctx_priv->cur_ps_args;
    const struct wined3d_gl_info *gl_info = context->gl_info;
    const struct wined3d_shader_indexable_temp *idx_temp_reg;
    unsigned int uniform_block_base, uniform_block_count;
    const struct wined3d_shader_lconst *lconst;
    const char *prefix;
    unsigned int i;
    DWORD map;
 
    prefix = shader_glsl_get_prefix(version->type);
 
    /* Prototype the subroutines */
    for (i = 0, map = reg_maps->labels; map; map >>= 1, ++i)
    {
        if (map & 1) shader_addline(buffer, "void subroutine%u();\n", i);
    }
 
    /* Declare the constants (aka uniforms) */
    if (shader->limits->constant_float > 0)
    {
        unsigned max_constantsF;
 
        /* Unless the shader uses indirect addressing, always declare the
         * maximum array size and ignore that we need some uniforms privately.
         * E.g. if GL supports 256 uniforms, and we need 2 for the pos fixup
         * and immediate values, still declare VC[256]. If the shader needs
         * more uniforms than we have it won't work in any case. If it uses
         * less, the compiler will figure out which uniforms are really used
         * and strip them out. This allows a shader to use c255 on a dx9 card,
         * as long as it doesn't also use all the other constants.
         *
         * If the shader uses indirect addressing the compiler must assume
         * that all declared uniforms are used. In this case, declare only the
         * amount that we're assured to have.
         *
         * Thus we run into problems in these two cases:
         * 1) The shader really uses more uniforms than supported.
         * 2) The shader uses indirect addressing, less constants than
         *    supported, but uses a constant index > #supported consts. */
        if (version->type == WINED3D_SHADER_TYPE_PIXEL)
        {
            /* No indirect addressing here. */
            max_constantsF = gl_info->limits.glsl_ps_float_constants;
        }
        else
        {
            if (reg_maps->usesrelconstF)
            {
                /* Subtract the other potential uniforms from the max
                 * available (bools, ints, and 1 row of projection matrix).
                 * Subtract another uniform for immediate values, which have
                 * to be loaded via uniform by the driver as well. The shader
                 * code only uses 0.5, 2.0, 1.0, 128 and -128 in vertex
                 * shader code, so one vec4 should be enough. (Unfortunately
                 * the Nvidia driver doesn't store 128 and -128 in one float).
                 *
                 * Writing gl_ClipVertex requires one uniform for each
                 * clipplane as well. */
                max_constantsF = gl_info->limits.glsl_vs_float_constants - 3;
                if (vs_args->clip_enabled)
                    max_constantsF -= gl_info->limits.user_clip_distances;
                max_constantsF -= wined3d_popcount(reg_maps->integer_constants);
                /* Strictly speaking a bool only uses one scalar, but the nvidia(Linux) compiler doesn't pack them properly,
                 * so each scalar requires a full vec4. We could work around this by packing the booleans ourselves, but
                 * for now take this into account when calculating the number of available constants
                 */
                max_constantsF -= wined3d_popcount(reg_maps->boolean_constants);
                /* Set by driver quirks in directx.c */
                max_constantsF -= gl_info->reserved_glsl_constants;
 
                if (max_constantsF < shader->limits->constant_float)
                {
                    static unsigned int once;
 
                    if (!once++)
                        ERR_(winediag)("The hardware does not support enough uniform components to run this shader,"
                                " it may not render correctly.\n");
                    else
                        WARN("The hardware does not support enough uniform components to run this shader.\n");
                }
            }
            else
            {
                max_constantsF = gl_info->limits.glsl_vs_float_constants;
            }
        }
        max_constantsF = min(shader->limits->constant_float, max_constantsF);
        shader_addline(buffer, "uniform vec4 %s_c[%u];\n", prefix, max_constantsF);
    }
 
    /* Always declare the full set of constants, the compiler can remove the
     * unused ones because d3d doesn't (yet) support indirect int and bool
     * constant addressing. This avoids problems if the app uses e.g. i0 and i9. */
    if (shader->limits->constant_int > 0 && reg_maps->integer_constants)
        shader_addline(buffer, "uniform ivec4 %s_i[%u];\n", prefix, shader->limits->constant_int);
 
    if (shader->limits->constant_bool > 0 && reg_maps->boolean_constants)
        shader_addline(buffer, "uniform bool %s_b[%u];\n", prefix, shader->limits->constant_bool);
 
    /* Declare immediate constant buffer */
    if (reg_maps->icb)
        shader_addline(buffer, "uniform vec4 %s_icb[%u];\n", prefix, reg_maps->icb->vec4_count);
 
    /* Declare constant buffers */
    wined3d_gl_limits_get_uniform_block_range(&gl_info->limits, version->type,
            &uniform_block_base, &uniform_block_count);
    for (i = 0; i < min(uniform_block_count, WINED3D_MAX_CBS); ++i)
    {
        if (reg_maps->cb_sizes[i])
        {
            shader_addline(buffer, "layout(std140");
            if (shader_glsl_use_layout_binding_qualifier(gl_info))
                shader_addline(buffer, ", binding = %u", uniform_block_base + i);
            shader_addline(buffer, ") uniform block_%s_cb%u { vec4 %s_cb%u[%u]; };\n",
                    prefix, i, prefix, i, reg_maps->cb_sizes[i]);
        }
    }
 
    /* Declare texture samplers */
    for (i = 0; i < reg_maps->sampler_map.count; ++i)
    {
        struct wined3d_shader_sampler_map_entry *entry;
        const char *sampler_type_prefix, *sampler_type;
        BOOL shadow_sampler, tex_rect;
 
        entry = &reg_maps->sampler_map.entries[i];
 
        if (entry->resource_idx >= ARRAY_SIZE(reg_maps->resource_info))
        {
            ERR("Invalid resource index %u.\n", entry->resource_idx);
            continue;
        }
 
        switch (reg_maps->resource_info[entry->resource_idx].data_type)
        {
            case WINED3D_DATA_FLOAT:
            case WINED3D_DATA_UNORM:
            case WINED3D_DATA_SNORM:
                sampler_type_prefix = "";
                break;
 
            case WINED3D_DATA_INT:
                sampler_type_prefix = "i";
                break;
 
            case WINED3D_DATA_UINT:
                sampler_type_prefix = "u";
                break;
 
            default:
                sampler_type_prefix = "";
                ERR("Unhandled resource data type %#x.\n", reg_maps->resource_info[i].data_type);
                break;
        }
 
        shadow_sampler = glsl_is_shadow_sampler(shader, ps_args, entry->resource_idx, entry->sampler_idx);
        switch (reg_maps->resource_info[entry->resource_idx].type)
        {
            case WINED3D_SHADER_RESOURCE_BUFFER:
                sampler_type = "samplerBuffer";
                break;
 
            case WINED3D_SHADER_RESOURCE_TEXTURE_1D:
                if (shadow_sampler)
                    sampler_type = "sampler1DShadow";
                else
                    sampler_type = "sampler1D";
                break;
 
            case WINED3D_SHADER_RESOURCE_TEXTURE_2D:
                tex_rect = version->type == WINED3D_SHADER_TYPE_PIXEL
                        && (ps_args->np2_fixup & (1u << entry->resource_idx))
                        && gl_info->supported[ARB_TEXTURE_RECTANGLE];
                if (shadow_sampler)
                {
                    if (tex_rect)
                        sampler_type = "sampler2DRectShadow";
                    else
                        sampler_type = "sampler2DShadow";
                }
                else
                {
                    if (tex_rect)
                        sampler_type = "sampler2DRect";
                    else
                        sampler_type = "sampler2D";
                }
                break;
 
            case WINED3D_SHADER_RESOURCE_TEXTURE_3D:
                if (shadow_sampler)
                    FIXME("Unsupported 3D shadow sampler.\n");
                sampler_type = "sampler3D";
                break;
 
            case WINED3D_SHADER_RESOURCE_TEXTURE_CUBE:
                if (shadow_sampler)
                    sampler_type = "samplerCubeShadow";
                else
                    sampler_type = "samplerCube";
                break;
 
            case WINED3D_SHADER_RESOURCE_TEXTURE_1DARRAY:
                if (shadow_sampler)
                    sampler_type = "sampler1DArrayShadow";
                else
                    sampler_type = "sampler1DArray";
                break;
 
            case WINED3D_SHADER_RESOURCE_TEXTURE_2DARRAY:
                if (shadow_sampler)
                    sampler_type = "sampler2DArrayShadow";
                else
                    sampler_type = "sampler2DArray";
                break;
 
            case WINED3D_SHADER_RESOURCE_TEXTURE_CUBEARRAY:
                if (shadow_sampler)
                    sampler_type = "samplerCubeArrayShadow";
                else
                    sampler_type = "samplerCubeArray";
                break;
 
            case WINED3D_SHADER_RESOURCE_TEXTURE_2DMS:
                sampler_type = "sampler2DMS";
                break;
 
            case WINED3D_SHADER_RESOURCE_TEXTURE_2DMSARRAY:
                sampler_type = "sampler2DMSArray";
                break;
 
            default:
                sampler_type = "unsupported_sampler";
                FIXME("Unhandled resource type %#x.\n", reg_maps->resource_info[entry->resource_idx].type);
                break;
        }
 
        if (shader_glsl_use_layout_binding_qualifier(gl_info))
            shader_glsl_append_sampler_binding_qualifier(buffer, context, version, entry->bind_idx);
        shader_addline(buffer, "uniform %s%s %s_sampler%u;\n",
                sampler_type_prefix, sampler_type, prefix, entry->bind_idx);
    }
 
    /* Declare images */
    for (i = 0; i < ARRAY_SIZE(reg_maps->uav_resource_info); ++i)
    {
        const char *image_type_prefix, *image_type, *read_format;
 
        if (!reg_maps->uav_resource_info[i].type)
            continue;
 
        switch (reg_maps->uav_resource_info[i].data_type)
        {
            case WINED3D_DATA_FLOAT:
            case WINED3D_DATA_UNORM:
            case WINED3D_DATA_SNORM:
                image_type_prefix = "";
                read_format = "r32f";
                break;
 
            case WINED3D_DATA_INT:
                image_type_prefix = "i";
                read_format = "r32i";
                break;
 
            case WINED3D_DATA_UINT:
                image_type_prefix = "u";
                read_format = "r32ui";
                break;
 
            default:
                image_type_prefix = "";
                read_format = "";
                ERR("Unhandled resource data type %#x.\n", reg_maps->uav_resource_info[i].data_type);
                break;
        }
 
        switch (reg_maps->uav_resource_info[i].type)
        {
            case WINED3D_SHADER_RESOURCE_BUFFER:
                image_type = "imageBuffer";
                break;
 
            case WINED3D_SHADER_RESOURCE_TEXTURE_2D:
                image_type = "image2D";
                break;
 
            case WINED3D_SHADER_RESOURCE_TEXTURE_3D:
                image_type = "image3D";
                break;
 
            case WINED3D_SHADER_RESOURCE_TEXTURE_2DARRAY:
                image_type = "image2DArray";
                break;
 
            default:
                image_type = "unsupported_image";
                FIXME("Unhandled resource type %#x.\n", reg_maps->uav_resource_info[i].type);
                break;
        }
 
        if (shader_glsl_use_layout_binding_qualifier(gl_info))
            shader_addline(buffer, "layout(binding = %u)\n", i);
        if (reg_maps->uav_read_mask & (1u << i))
            shader_addline(buffer, "layout(%s) uniform %s%s %s_image%u;\n",
                    read_format, image_type_prefix, image_type, prefix, i);
        else
            shader_addline(buffer, "writeonly uniform %s%s %s_image%u;\n",
                    image_type_prefix, image_type, prefix, i);
 
        if (reg_maps->uav_counter_mask & (1u << i))
            shader_addline(buffer, "layout(binding = %u) uniform atomic_uint %s_counter%u;\n",
                    i, prefix, i);
    }
 
    /* Declare address variables */
    for (i = 0, map = reg_maps->address; map; map >>= 1, ++i)
    {
        if (map & 1) shader_addline(buffer, "ivec4 A%u;\n", i);
    }
 
    /* Declare output register temporaries */
    if (shader->limits->packed_output)
        shader_addline(buffer, "vec4 %s_out[%u];\n", prefix, shader->limits->packed_output);
 
    /* Declare temporary variables */
    if (reg_maps->temporary_count)
    {
        for (i = 0; i < reg_maps->temporary_count; ++i)
            shader_addline(buffer, "vec4 R%u;\n", i);
    }
    else if (version->major < 4)
    {
        for (i = 0, map = reg_maps->temporary; map; map >>= 1, ++i)
        {
            if (map & 1)
                shader_addline(buffer, "vec4 R%u;\n", i);
        }
    }
 
    /* Declare indexable temporary variables */
    LIST_FOR_EACH_ENTRY(idx_temp_reg, &reg_maps->indexable_temps, struct wined3d_shader_indexable_temp, entry)
    {
        if (idx_temp_reg->component_count != 4)
            FIXME("Ignoring component count %u.\n", idx_temp_reg->component_count);
        shader_addline(buffer, "vec4 X%u[%u];\n", idx_temp_reg->register_idx, idx_temp_reg->register_size);
    }
 
    /* Declare loop registers aLx */
    if (version->major < 4)
    {
        for (i = 0; i < reg_maps->loop_depth; ++i)
        {
            shader_addline(buffer, "int aL%u;\n", i);
            shader_addline(buffer, "int tmpInt%u;\n", i);
        }
    }
 
    /* Temporary variables for matrix operations */
    shader_addline(buffer, "vec4 tmp0;\n");
    shader_addline(buffer, "vec4 tmp1;\n");
 
    if (!shader->load_local_constsF)
    {
        LIST_FOR_EACH_ENTRY(lconst, &shader->constantsF, struct wined3d_shader_lconst, entry)
        {
            shader_addline(buffer, "const vec4 %s_lc%u = ", prefix, lconst->idx);
            shader_glsl_append_imm_vec4(buffer, (const float *)lconst->value);
            shader_addline(buffer, ";\n");
        }
    }
}
 
/* Prototypes */
static void shader_glsl_add_src_param(const struct wined3d_shader_instruction *ins,
        const struct wined3d_shader_src_param *wined3d_src, DWORD mask, struct glsl_src_param *glsl_src);
 
static const char * const shift_glsl_tab[] = {
    "",           /*  0 (none) */
    "2.0 * ",     /*  1 (x2)   */
    "4.0 * ",     /*  2 (x4)   */
    "8.0 * ",     /*  3 (x8)   */
    "16.0 * ",    /*  4 (x16)  */
    "32.0 * ",    /*  5 (x32)  */
    "",           /*  6 (x64)  */
    "",           /*  7 (x128) */
    "",           /*  8 (d256) */
    "",           /*  9 (d128) */
    "",           /* 10 (d64)  */
    "",           /* 11 (d32)  */
    "0.0625 * ",  /* 12 (d16)  */
    "0.125 * ",   /* 13 (d8)   */
    "0.25 * ",    /* 14 (d4)   */
    "0.5 * "      /* 15 (d2)   */
};
 
/* Generate a GLSL parameter that does the input modifier computation and return the input register/mask to use */
static void shader_glsl_gen_modifier(enum wined3d_shader_src_modifier src_modifier,
        const char *in_reg, const char *in_regswizzle, char *out_str)
{
    switch (src_modifier)
    {
    case WINED3DSPSM_DZ: /* Need to handle this in the instructions itself (texld & texcrd). */
    case WINED3DSPSM_DW:
    case WINED3DSPSM_NONE:
        sprintf(out_str, "%s%s", in_reg, in_regswizzle);
        break;
    case WINED3DSPSM_NEG:
        sprintf(out_str, "-%s%s", in_reg, in_regswizzle);
        break;
    case WINED3DSPSM_NOT:
        sprintf(out_str, "!%s%s", in_reg, in_regswizzle);
        break;
    case WINED3DSPSM_BIAS:
        sprintf(out_str, "(%s%s - vec4(0.5)%s)", in_reg, in_regswizzle, in_regswizzle);
        break;
    case WINED3DSPSM_BIASNEG:
        sprintf(out_str, "-(%s%s - vec4(0.5)%s)", in_reg, in_regswizzle, in_regswizzle);
        break;
    case WINED3DSPSM_SIGN:
        sprintf(out_str, "(2.0 * (%s%s - 0.5))", in_reg, in_regswizzle);
        break;
    case WINED3DSPSM_SIGNNEG:
        sprintf(out_str, "-(2.0 * (%s%s - 0.5))", in_reg, in_regswizzle);
        break;
    case WINED3DSPSM_COMP:
        sprintf(out_str, "(1.0 - %s%s)", in_reg, in_regswizzle);
        break;
    case WINED3DSPSM_X2:
        sprintf(out_str, "(2.0 * %s%s)", in_reg, in_regswizzle);
        break;
    case WINED3DSPSM_X2NEG:
        sprintf(out_str, "-(2.0 * %s%s)", in_reg, in_regswizzle);
        break;
    case WINED3DSPSM_ABS:
        sprintf(out_str, "abs(%s%s)", in_reg, in_regswizzle);
        break;
    case WINED3DSPSM_ABSNEG:
        sprintf(out_str, "-abs(%s%s)", in_reg, in_regswizzle);
        break;
    default:
        FIXME("Unhandled modifier %u\n", src_modifier);
        sprintf(out_str, "%s%s", in_reg, in_regswizzle);
    }
}
 
static void shader_glsl_fixup_scalar_register_variable(char *register_name,
        const char *glsl_variable, const struct wined3d_gl_info *gl_info)
{
    /* The ARB_shading_language_420pack extension allows swizzle operations on
     * scalars. */
    if (gl_info->supported[ARB_SHADING_LANGUAGE_420PACK])
        sprintf(register_name, "%s", glsl_variable);
    else
        sprintf(register_name, "ivec2(%s, 0)", glsl_variable);
}
 
static void shader_glsl_get_register_name(const struct wined3d_shader_register *reg,
        enum wined3d_data_type data_type, char *register_name, BOOL *is_color,
        const struct wined3d_shader_instruction *ins)
{
    /* oPos, oFog and oPts in D3D */
    static const char * const hwrastout_reg_names[] = {"vs_out[10]", "vs_out[11].x", "vs_out[11].y"};
 
    const struct wined3d_shader *shader = ins->ctx->shader;
    const struct wined3d_shader_reg_maps *reg_maps = ins->ctx->reg_maps;
    const struct wined3d_shader_version *version = &reg_maps->shader_version;
    const struct wined3d_gl_info *gl_info = ins->ctx->gl_info;
    const char *prefix = shader_glsl_get_prefix(version->type);
    struct glsl_src_param rel_param0, rel_param1;
    char imm_str[4][17];
 
    if (reg->idx[0].offset != ~0U && reg->idx[0].rel_addr)
        shader_glsl_add_src_param(ins, reg->idx[0].rel_addr, WINED3DSP_WRITEMASK_0, &rel_param0);
    if (reg->idx[1].offset != ~0U && reg->idx[1].rel_addr)
        shader_glsl_add_src_param(ins, reg->idx[1].rel_addr, WINED3DSP_WRITEMASK_0, &rel_param1);
    *is_color = FALSE;
 
    switch (reg->type)
    {
        case WINED3DSPR_TEMP:
            sprintf(register_name, "R%u", reg->idx[0].offset);
            break;
 
        case WINED3DSPR_INPUT:
        case WINED3DSPR_INCONTROLPOINT:
            if (version->type == WINED3D_SHADER_TYPE_VERTEX)
            {
                struct shader_glsl_ctx_priv *priv = ins->ctx->backend_data;
 
                if (reg->idx[0].rel_addr)
                    FIXME("VS3 input registers relative addressing.\n");
                if (priv->cur_vs_args->swizzle_map & (1u << reg->idx[0].offset))
                    *is_color = TRUE;
                if (reg->idx[0].rel_addr)
                {
                    sprintf(register_name, "%s_in[%s + %u]",
                            prefix, rel_param0.param_str, reg->idx[0].offset);
                }
                else
                {
                    sprintf(register_name, "%s_in%u", prefix, reg->idx[0].offset);
                }
                break;
            }
 
            if (version->type == WINED3D_SHADER_TYPE_HULL
                    || version->type == WINED3D_SHADER_TYPE_DOMAIN
                    || version->type == WINED3D_SHADER_TYPE_GEOMETRY)
            {
                if (reg->idx[0].rel_addr)
                {
                    if (reg->idx[1].rel_addr)
                        sprintf(register_name, "shader_in[%s + %u].reg[%s + %u]",
                                rel_param0.param_str, reg->idx[0].offset,
                                rel_param1.param_str, reg->idx[1].offset);
                    else
                        sprintf(register_name, "shader_in[%s + %u].reg[%u]",
                                rel_param0.param_str, reg->idx[0].offset,
                                reg->idx[1].offset);
                }
                else if (reg->idx[1].rel_addr)
                    sprintf(register_name, "shader_in[%u].reg[%s + %u]", reg->idx[0].offset,
                            rel_param1.param_str, reg->idx[1].offset);
                else
                    sprintf(register_name, "shader_in[%u].reg[%u]",
                            reg->idx[0].offset, reg->idx[1].offset);
                break;
            }
 
            /* pixel shaders >= 3.0 */
            if (version->major >= 3)
            {
                DWORD idx = shader->u.ps.input_reg_map[reg->idx[0].offset];
                unsigned int in_count = vec4_varyings(version->major, gl_info);
 
                if (reg->idx[0].rel_addr)
                {
                    /* Removing a + 0 would be an obvious optimization, but
                     * OS X doesn't see the NOP operation there. */
                    if (idx)
                    {
                        if (needs_legacy_glsl_syntax(gl_info)
                                && shader->u.ps.declared_in_count > in_count)
                        {
                            sprintf(register_name,
                                    "((%s + %u) > %u ? (%s + %u) > %u ? gl_SecondaryColor : gl_Color : %s_in[%s + %u])",
                                    rel_param0.param_str, idx, in_count - 1, rel_param0.param_str, idx, in_count,
                                    prefix, rel_param0.param_str, idx);
                        }
                        else
                        {
                            sprintf(register_name, "%s_in[%s + %u]", prefix, rel_param0.param_str, idx);
                        }
                    }
                    else
                    {
                        if (needs_legacy_glsl_syntax(gl_info)
                                && shader->u.ps.declared_in_count > in_count)
                        {
                            sprintf(register_name, "((%s) > %u ? (%s) > %u ? gl_SecondaryColor : gl_Color : %s_in[%s])",
                                    rel_param0.param_str, in_count - 1, rel_param0.param_str, in_count,
                                    prefix, rel_param0.param_str);
                        }
                        else
                        {
                            sprintf(register_name, "%s_in[%s]", prefix, rel_param0.param_str);
                        }
                    }
                }
                else
                {
                    if (idx == in_count) sprintf(register_name, "gl_Color");
                    else if (idx == in_count + 1) sprintf(register_name, "gl_SecondaryColor");
                    else sprintf(register_name, "%s_in[%u]", prefix, idx);
                }
            }
            else
            {
                if (!reg->idx[0].offset)
                    strcpy(register_name, "ffp_varying_diffuse");
                else
                    strcpy(register_name, "ffp_varying_specular");
                break;
            }
            break;
 
        case WINED3DSPR_CONST:
            {
                /* Relative addressing */
                if (reg->idx[0].rel_addr)
                {
                    if (wined3d_settings.check_float_constants)
                        sprintf(register_name, "(%s + %u >= 0 && %s + %u < %u ? %s_c[%s + %u] : vec4(0.0))",
                                rel_param0.param_str, reg->idx[0].offset,
                                rel_param0.param_str, reg->idx[0].offset, shader->limits->constant_float,
                                prefix, rel_param0.param_str, reg->idx[0].offset);
                    else if (reg->idx[0].offset)
                        sprintf(register_name, "%s_c[%s + %u]", prefix, rel_param0.param_str, reg->idx[0].offset);
                    else
                        sprintf(register_name, "%s_c[%s]", prefix, rel_param0.param_str);
                }
                else
                {
                    if (shader_constant_is_local(shader, reg->idx[0].offset))
                        sprintf(register_name, "%s_lc%u", prefix, reg->idx[0].offset);
                    else
                        sprintf(register_name, "%s_c[%u]", prefix, reg->idx[0].offset);
                }
            }
            break;
 
        case WINED3DSPR_CONSTINT:
            sprintf(register_name, "%s_i[%u]", prefix, reg->idx[0].offset);
            break;
 
        case WINED3DSPR_CONSTBOOL:
            sprintf(register_name, "%s_b[%u]", prefix, reg->idx[0].offset);
            break;
 
        case WINED3DSPR_TEXTURE: /* case WINED3DSPR_ADDR: */
            if (version->type == WINED3D_SHADER_TYPE_PIXEL)
                sprintf(register_name, "T%u", reg->idx[0].offset);
            else
                sprintf(register_name, "A%u", reg->idx[0].offset);
            break;
 
        case WINED3DSPR_LOOP:
            sprintf(register_name, "aL%u", ins->ctx->state->current_loop_reg - 1);
            break;
 
        case WINED3DSPR_SAMPLER:
            sprintf(register_name, "%s_sampler%u", prefix, reg->idx[0].offset);
            break;
 
        case WINED3DSPR_COLOROUT:
            /* FIXME: should check dual_buffers when dual blending is enabled */
            if (reg->idx[0].offset >= gl_info->limits.buffers)
                WARN("Write to render target %u, only %d supported.\n",
                        reg->idx[0].offset, gl_info->limits.buffers);
 
            sprintf(register_name, needs_legacy_glsl_syntax(gl_info) ? "gl_FragData[%u]" : "ps_out%u",
                    reg->idx[0].offset);
            break;
 
        case WINED3DSPR_RASTOUT:
            sprintf(register_name, "%s", hwrastout_reg_names[reg->idx[0].offset]);
            break;
 
        case WINED3DSPR_DEPTHOUT:
        case WINED3DSPR_DEPTHOUTGE:
        case WINED3DSPR_DEPTHOUTLE:
            sprintf(register_name, "gl_FragDepth");
            break;
 
        case WINED3DSPR_ATTROUT:
            if (!reg->idx[0].offset)
                sprintf(register_name, "%s_out[8]", prefix);
            else
                sprintf(register_name, "%s_out[9]", prefix);
            break;
 
        case WINED3DSPR_TEXCRDOUT:
            /* Vertex shaders >= 3.0: WINED3DSPR_OUTPUT */
            if (reg->idx[0].rel_addr)
                sprintf(register_name, "%s_out[%s + %u]",
                        prefix, rel_param0.param_str, reg->idx[0].offset);
            else
                sprintf(register_name, "%s_out[%u]", prefix, reg->idx[0].offset);
            break;
 
        case WINED3DSPR_MISCTYPE:
            if (!reg->idx[0].offset)
            {
                /* vPos */
                sprintf(register_name, "vpos");
            }
            else if (reg->idx[0].offset == 1)
            {
                /* Note that gl_FrontFacing is a bool, while vFace is
                 * a float for which the sign determines front/back */
                sprintf(register_name, "(gl_FrontFacing ? 1.0 : -1.0)");
            }
            else
            {
                FIXME("Unhandled misctype register %u.\n", reg->idx[0].offset);
                sprintf(register_name, "unrecognized_register");
            }
            break;
 
        case WINED3DSPR_IMMCONST:
            switch (reg->immconst_type)
            {
                case WINED3D_IMMCONST_SCALAR:
                    switch (data_type)
                    {
                        case WINED3D_DATA_FLOAT:
                            if (gl_info->supported[ARB_SHADER_BIT_ENCODING])
                                sprintf(register_name, "uintBitsToFloat(%#xu)", reg->u.immconst_data[0]);
                            else
                                wined3d_ftoa(*(const float *)reg->u.immconst_data, register_name);
                            break;
                        case WINED3D_DATA_INT:
                            sprintf(register_name, "%#x", reg->u.immconst_data[0]);
                            break;
                        case WINED3D_DATA_RESOURCE:
                        case WINED3D_DATA_SAMPLER:
                        case WINED3D_DATA_UINT:
                            sprintf(register_name, "%#xu", reg->u.immconst_data[0]);
                            break;
                        default:
                            sprintf(register_name, "<unhandled data type %#x>", data_type);
                            break;
                    }
                    break;
 
                case WINED3D_IMMCONST_VEC4:
                    switch (data_type)
                    {
                        case WINED3D_DATA_FLOAT:
                            if (gl_info->supported[ARB_SHADER_BIT_ENCODING])
                            {
                                sprintf(register_name, "uintBitsToFloat(uvec4(%#xu, %#xu, %#xu, %#xu))",
                                        reg->u.immconst_data[0], reg->u.immconst_data[1],
                                        reg->u.immconst_data[2], reg->u.immconst_data[3]);
                            }
                            else
                            {
                                wined3d_ftoa(*(const float *)&reg->u.immconst_data[0], imm_str[0]);
                                wined3d_ftoa(*(const float *)&reg->u.immconst_data[1], imm_str[1]);
                                wined3d_ftoa(*(const float *)&reg->u.immconst_data[2], imm_str[2]);
                                wined3d_ftoa(*(const float *)&reg->u.immconst_data[3], imm_str[3]);
                                sprintf(register_name, "vec4(%s, %s, %s, %s)",
                                        imm_str[0], imm_str[1], imm_str[2], imm_str[3]);
                            }
                            break;
                        case WINED3D_DATA_INT:
                            sprintf(register_name, "ivec4(%#x, %#x, %#x, %#x)",
                                    reg->u.immconst_data[0], reg->u.immconst_data[1],
                                    reg->u.immconst_data[2], reg->u.immconst_data[3]);
                            break;
                        case WINED3D_DATA_RESOURCE:
                        case WINED3D_DATA_SAMPLER:
                        case WINED3D_DATA_UINT:
                            sprintf(register_name, "uvec4(%#xu, %#xu, %#xu, %#xu)",
                                    reg->u.immconst_data[0], reg->u.immconst_data[1],
                                    reg->u.immconst_data[2], reg->u.immconst_data[3]);
                            break;
                        default:
                            sprintf(register_name, "<unhandled data type %#x>", data_type);
                            break;
                    }
                    break;
 
                default:
                    FIXME("Unhandled immconst type %#x\n", reg->immconst_type);
                    sprintf(register_name, "<unhandled_immconst_type %#x>", reg->immconst_type);
            }
            break;
 
        case WINED3DSPR_CONSTBUFFER:
            if (reg->idx[1].rel_addr)
                sprintf(register_name, "%s_cb%u[%s + %u]",
                        prefix, reg->idx[0].offset, rel_param1.param_str, reg->idx[1].offset);
            else
                sprintf(register_name, "%s_cb%u[%u]", prefix, reg->idx[0].offset, reg->idx[1].offset);
            break;
 
        case WINED3DSPR_IMMCONSTBUFFER:
            if (reg->idx[0].rel_addr)
                sprintf(register_name, "%s_icb[%s + %u]", prefix, rel_param0.param_str, reg->idx[0].offset);
            else
                sprintf(register_name, "%s_icb[%u]", prefix, reg->idx[0].offset);
            break;
 
        case WINED3DSPR_PRIMID:
            if (version->type == WINED3D_SHADER_TYPE_GEOMETRY)
                sprintf(register_name, "gl_PrimitiveIDIn");
            else
                sprintf(register_name, "gl_PrimitiveID");
            break;
 
        case WINED3DSPR_IDXTEMP:
            if (reg->idx[1].rel_addr)
                sprintf(register_name, "X%u[%s + %u]", reg->idx[0].offset, rel_param1.param_str, reg->idx[1].offset);
            else
                sprintf(register_name, "X%u[%u]", reg->idx[0].offset, reg->idx[1].offset);
            break;
 
        case WINED3DSPR_LOCALTHREADINDEX:
            shader_glsl_fixup_scalar_register_variable(register_name,
                    "int(gl_LocalInvocationIndex)", gl_info);
            break;
 
        case WINED3DSPR_GSINSTID:
        case WINED3DSPR_OUTPOINTID:
            shader_glsl_fixup_scalar_register_variable(register_name,
                    "gl_InvocationID", gl_info);
            break;
 
        case WINED3DSPR_THREADID:
            sprintf(register_name, "ivec3(gl_GlobalInvocationID)");
            break;
 
        case WINED3DSPR_THREADGROUPID:
            sprintf(register_name, "ivec3(gl_WorkGroupID)");
            break;
 
        case WINED3DSPR_LOCALTHREADID:
            sprintf(register_name, "ivec3(gl_LocalInvocationID)");
            break;
 
        case WINED3DSPR_FORKINSTID:
        case WINED3DSPR_JOININSTID:
            shader_glsl_fixup_scalar_register_variable(register_name,
                    "phase_instance_id", gl_info);
            break;
 
        case WINED3DSPR_TESSCOORD:
            sprintf(register_name, "gl_TessCoord");
            break;
 
        case WINED3DSPR_OUTCONTROLPOINT:
            if (reg->idx[0].rel_addr)
            {
                if (reg->idx[1].rel_addr)
                    sprintf(register_name, "shader_out[%s + %u].reg[%s + %u]",
                            rel_param0.param_str, reg->idx[0].offset,
                            rel_param1.param_str, reg->idx[1].offset);
                else
                    sprintf(register_name, "shader_out[%s + %u].reg[%u]",
                            rel_param0.param_str, reg->idx[0].offset,
                            reg->idx[1].offset);
            }
            else if (reg->idx[1].rel_addr)
            {
                sprintf(register_name, "shader_out[%u].reg[%s + %u]",
                        reg->idx[0].offset, rel_param1.param_str,
                        reg->idx[1].offset);
            }
            else
            {
                sprintf(register_name, "shader_out[%u].reg[%u]",
                        reg->idx[0].offset, reg->idx[1].offset);
            }
            break;
 
        case WINED3DSPR_PATCHCONST:
            if (version->type == WINED3D_SHADER_TYPE_HULL)
                sprintf(register_name, "hs_out[%u]", reg->idx[0].offset);
            else
                sprintf(register_name, "vpc[%u]", reg->idx[0].offset);
            break;
 
        default:
            FIXME("Unhandled register type %#x.\n", reg->type);
            sprintf(register_name, "unrecognized_register");
            break;
    }
}
 
static void shader_glsl_write_mask_to_str(DWORD write_mask, char *str)
{
    *str++ = '.';
    if (write_mask & WINED3DSP_WRITEMASK_0) *str++ = 'x';
    if (write_mask & WINED3DSP_WRITEMASK_1) *str++ = 'y';
    if (write_mask & WINED3DSP_WRITEMASK_2) *str++ = 'z';
    if (write_mask & WINED3DSP_WRITEMASK_3) *str++ = 'w';
    *str = '\0';
}
 
/* Get the GLSL write mask for the destination register */
static DWORD shader_glsl_get_write_mask(const struct wined3d_shader_dst_param *param, char *write_mask)
{
    DWORD mask = param->write_mask;
 
    if (shader_is_scalar(&param->reg))
    {
        mask = WINED3DSP_WRITEMASK_0;
        *write_mask = '\0';
    }
    else
    {
        shader_glsl_write_mask_to_str(mask, write_mask);
    }
 
    return mask;
}
 
static unsigned int shader_glsl_get_write_mask_size(DWORD write_mask)
{
    unsigned int size = 0;
 
    if (write_mask & WINED3DSP_WRITEMASK_0) ++size;
    if (write_mask & WINED3DSP_WRITEMASK_1) ++size;
    if (write_mask & WINED3DSP_WRITEMASK_2) ++size;
    if (write_mask & WINED3DSP_WRITEMASK_3) ++size;
 
    return size;
}
 
static unsigned int shader_glsl_swizzle_get_component(DWORD swizzle,
        unsigned int component_idx)
{
    /* swizzle bits fields: wwzzyyxx */
    return (swizzle >> (2 * component_idx)) & 0x3;
}
 
static void shader_glsl_swizzle_to_str(DWORD swizzle, BOOL fixup, DWORD mask, char *str)
{
    /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
     * but addressed as "rgba". To fix this we need to swap the register's x
     * and z components. */
    const char *swizzle_chars = fixup ? "zyxw" : "xyzw";
    unsigned int i;
 
    *str++ = '.';
    for (i = 0; i < 4; ++i)
    {
        if (mask & (WINED3DSP_WRITEMASK_0 << i))
            *str++ = swizzle_chars[shader_glsl_swizzle_get_component(swizzle, i)];
    }
    *str = '\0';
}
 
static void shader_glsl_get_swizzle(const struct wined3d_shader_src_param *param,
        BOOL fixup, DWORD mask, char *swizzle_str)
{
    if (shader_is_scalar(&param->reg))
        *swizzle_str = '\0';
    else
        shader_glsl_swizzle_to_str(param->swizzle, fixup, mask, swizzle_str);
}
 
static void shader_glsl_sprintf_cast(struct wined3d_string_buffer *dst_param, const char *src_param,
        enum wined3d_data_type dst_data_type, enum wined3d_data_type src_data_type)
{
    if (dst_data_type == src_data_type)
    {
        string_buffer_sprintf(dst_param, "%s", src_param);
        return;
    }
 
    if (src_data_type == WINED3D_DATA_FLOAT)
    {
        switch (dst_data_type)
        {
            case WINED3D_DATA_INT:
                string_buffer_sprintf(dst_param, "floatBitsToInt(%s)", src_param);
                return;
            case WINED3D_DATA_RESOURCE:
            case WINED3D_DATA_SAMPLER:
            case WINED3D_DATA_UINT:
                string_buffer_sprintf(dst_param, "floatBitsToUint(%s)", src_param);
                return;
            default:
                break;
        }
    }
 
    if (src_data_type == WINED3D_DATA_UINT && dst_data_type == WINED3D_DATA_FLOAT)
    {
        string_buffer_sprintf(dst_param, "uintBitsToFloat(%s)", src_param);
        return;
    }
 
    if (src_data_type == WINED3D_DATA_INT && dst_data_type == WINED3D_DATA_FLOAT)
    {
        string_buffer_sprintf(dst_param, "intBitsToFloat(%s)", src_param);
        return;
    }
 
    FIXME("Unhandled cast from %#x to %#x.\n", src_data_type, dst_data_type);
    string_buffer_sprintf(dst_param, "%s", src_param);
}
 
/* From a given parameter token, generate the corresponding GLSL string.
 * Also, return the actual register name and swizzle in case the
 * caller needs this information as well. */
static void shader_glsl_add_src_param_ext(const struct wined3d_shader_instruction *ins,
        const struct wined3d_shader_src_param *wined3d_src, DWORD mask, struct glsl_src_param *glsl_src,
        enum wined3d_data_type data_type)
{
    struct shader_glsl_ctx_priv *priv = ins->ctx->backend_data;
    struct wined3d_string_buffer *reg_name = string_buffer_get(priv->string_buffers);
    enum wined3d_data_type param_data_type;
    BOOL is_color = FALSE;
    char swizzle_str[6];
 
    glsl_src->reg_name[0] = '\0';
    glsl_src->param_str[0] = '\0';
    swizzle_str[0] = '\0';
 
    shader_glsl_get_register_name(&wined3d_src->reg, data_type, glsl_src->reg_name, &is_color, ins);
    shader_glsl_get_swizzle(wined3d_src, is_color, mask, swizzle_str);
 
    switch (wined3d_src->reg.type)
    {
        case WINED3DSPR_IMMCONST:
            param_data_type = data_type;
            break;
        case WINED3DSPR_FORKINSTID:
        case WINED3DSPR_GSINSTID:
        case WINED3DSPR_JOININSTID:
        case WINED3DSPR_LOCALTHREADID:
        case WINED3DSPR_LOCALTHREADINDEX:
        case WINED3DSPR_OUTPOINTID:
        case WINED3DSPR_PRIMID:
        case WINED3DSPR_THREADGROUPID:
        case WINED3DSPR_THREADID:
            param_data_type = WINED3D_DATA_INT;
            break;
        default:
            param_data_type = WINED3D_DATA_FLOAT;
            break;
    }
 
    shader_glsl_sprintf_cast(reg_name, glsl_src->reg_name, data_type, param_data_type);
    shader_glsl_gen_modifier(wined3d_src->modifiers, reg_name->buffer, swizzle_str, glsl_src->param_str);
 
    string_buffer_release(priv->string_buffers, reg_name);
}
 
static void shader_glsl_add_src_param(const struct wined3d_shader_instruction *ins,
        const struct wined3d_shader_src_param *wined3d_src, DWORD mask, struct glsl_src_param *glsl_src)
{
    shader_glsl_add_src_param_ext(ins, wined3d_src, mask, glsl_src, wined3d_src->reg.data_type);
}
 
/* From a given parameter token, generate the corresponding GLSL string.
 * Also, return the actual register name and swizzle in case the
 * caller needs this information as well. */
static DWORD shader_glsl_add_dst_param(const struct wined3d_shader_instruction *ins,
        const struct wined3d_shader_dst_param *wined3d_dst, struct glsl_dst_param *glsl_dst)
{
    BOOL is_color = FALSE;
 
    glsl_dst->mask_str[0] = '\0';
    glsl_dst->reg_name[0] = '\0';
 
    shader_glsl_get_register_name(&wined3d_dst->reg, wined3d_dst->reg.data_type,
            glsl_dst->reg_name, &is_color, ins);
    return shader_glsl_get_write_mask(wined3d_dst, glsl_dst->mask_str);
}
 
/* Append the destination part of the instruction to the buffer, return the effective write mask */
static DWORD shader_glsl_append_dst_ext(struct wined3d_string_buffer *buffer,
        const struct wined3d_shader_instruction *ins, const struct wined3d_shader_dst_param *dst,
        enum wined3d_data_type data_type)
{
    struct glsl_dst_param glsl_dst;
    DWORD mask;
 
    if ((mask = shader_glsl_add_dst_param(ins, dst, &glsl_dst)))
    {
        switch (data_type)
        {
            case WINED3D_DATA_FLOAT:
                shader_addline(buffer, "%s%s = %s(",
                        glsl_dst.reg_name, glsl_dst.mask_str, shift_glsl_tab[dst->shift]);
                break;
            case WINED3D_DATA_INT:
                shader_addline(buffer, "%s%s = %sintBitsToFloat(",
                        glsl_dst.reg_name, glsl_dst.mask_str, shift_glsl_tab[dst->shift]);
                break;
            case WINED3D_DATA_RESOURCE:
            case WINED3D_DATA_SAMPLER:
            case WINED3D_DATA_UINT:
                shader_addline(buffer, "%s%s = %suintBitsToFloat(",
                        glsl_dst.reg_name, glsl_dst.mask_str, shift_glsl_tab[dst->shift]);
                break;
            default:
                FIXME("Unhandled data type %#x.\n", data_type);
                shader_addline(buffer, "%s%s = %s(",
                        glsl_dst.reg_name, glsl_dst.mask_str, shift_glsl_tab[dst->shift]);
                break;
        }
    }
 
    return mask;
}
 
/* Append the destination part of the instruction to the buffer, return the effective write mask */
static DWORD shader_glsl_append_dst(struct wined3d_string_buffer *buffer, const struct wined3d_shader_instruction *ins)
{
    return shader_glsl_append_dst_ext(buffer, ins, &ins->dst[0], ins->dst[0].reg.data_type);
}
 
static void shader_glsl_add_instruction_modifiers(const struct wined3d_shader_instruction *ins)
{
    struct glsl_dst_param dst_param;
    DWORD modifiers;
 
    if (!ins->dst_count) return;
 
    modifiers = ins->dst[0].modifiers;
    if (!modifiers) return;
 
    shader_glsl_add_dst_param(ins, &ins->dst[0], &dst_param);
 
    if (modifiers & WINED3DSPDM_SATURATE)
    {
        /* _SAT means to clamp the value of the register to between 0 and 1 */
        shader_addline(ins->ctx->buffer, "%s%s = clamp(%s%s, 0.0, 1.0);\n", dst_param.reg_name,
                dst_param.mask_str, dst_param.reg_name, dst_param.mask_str);
    }
 
    if (modifiers & WINED3DSPDM_MSAMPCENTROID)
    {
        FIXME("_centroid modifier not handled\n");
    }
 
    if (modifiers & WINED3DSPDM_PARTIALPRECISION)
    {
        /* MSDN says this modifier can be safely ignored, so that's what we'll do. */
    }
}
 
static const char *shader_glsl_get_rel_op(enum wined3d_shader_rel_op op)
{
    switch (op)
    {
        case WINED3D_SHADER_REL_OP_GT: return ">";
        case WINED3D_SHADER_REL_OP_EQ: return "==";
        case WINED3D_SHADER_REL_OP_GE: return ">=";
        case WINED3D_SHADER_REL_OP_LT: return "<";
        case WINED3D_SHADER_REL_OP_NE: return "!=";
        case WINED3D_SHADER_REL_OP_LE: return "<=";
        default:
            FIXME("Unrecognized operator %#x.\n", op);
            return "(\?\?)";
    }
}
 
static BOOL shader_glsl_has_core_grad(const struct wined3d_gl_info *gl_info)
{
    return shader_glsl_get_version(gl_info) >= 130 || gl_info->supported[EXT_GPU_SHADER4];
}
 
static void shader_glsl_get_coord_size(enum wined3d_shader_resource_type resource_type,
        unsigned int *coord_size, unsigned int *deriv_size)
{
    const BOOL is_array = resource_type == WINED3D_SHADER_RESOURCE_TEXTURE_1DARRAY
            || resource_type == WINED3D_SHADER_RESOURCE_TEXTURE_2DARRAY;
 
    *coord_size = resource_type_info[resource_type].coord_size;
    *deriv_size = *coord_size;
    if (is_array)
        --(*deriv_size);
}
 
static void shader_glsl_get_sample_function(const struct wined3d_shader_context *ctx,
        DWORD resource_idx, DWORD sampler_idx, DWORD flags, struct glsl_sample_function *sample_function)
{
    enum wined3d_shader_resource_type resource_type = ctx->reg_maps->resource_info[resource_idx].type;
    struct shader_glsl_ctx_priv *priv = ctx->backend_data;
    const struct wined3d_gl_info *gl_info = ctx->gl_info;
    BOOL legacy_syntax = needs_legacy_glsl_syntax(gl_info);
    BOOL shadow = glsl_is_shadow_sampler(ctx->shader, priv->cur_ps_args, resource_idx, sampler_idx);
    BOOL projected = flags & WINED3D_GLSL_SAMPLE_PROJECTED;
    BOOL texrect = ctx->reg_maps->shader_version.type == WINED3D_SHADER_TYPE_PIXEL
            && priv->cur_ps_args->np2_fixup & (1u << resource_idx)
            && gl_info->supported[ARB_TEXTURE_RECTANGLE];
    BOOL lod = flags & WINED3D_GLSL_SAMPLE_LOD;
    BOOL grad = flags & WINED3D_GLSL_SAMPLE_GRAD;
    BOOL offset = flags & WINED3D_GLSL_SAMPLE_OFFSET;
    const char *base = "texture", *type_part = "", *suffix = "";
    unsigned int coord_size, deriv_size;
 
    sample_function->data_type = ctx->reg_maps->resource_info[resource_idx].data_type;
    sample_function->emulate_lod = WINED3D_SHADER_RESOURCE_NONE;
 
    if (resource_type >= ARRAY_SIZE(resource_type_info))
    {
        ERR("Unexpected resource type %#x.\n", resource_type);
        resource_type = WINED3D_SHADER_RESOURCE_TEXTURE_2D;
    }
 
    /* Note that there's no such thing as a projected cube texture. */
    if (resource_type == WINED3D_SHADER_RESOURCE_TEXTURE_CUBE)
        projected = FALSE;
 
    if (shadow && lod)
    {
        switch (resource_type)
        {
            /* emulate textureLod(sampler2DArrayShadow, ...) using textureGradOffset */
            case WINED3D_SHADER_RESOURCE_TEXTURE_2DARRAY:
                sample_function->emulate_lod = resource_type;
                grad = offset = TRUE;
                lod = FALSE;
                break;
 
            /* emulate textureLod(samplerCubeShadow, ...) using shadowCubeGrad */
            case WINED3D_SHADER_RESOURCE_TEXTURE_CUBE:
                sample_function->emulate_lod = resource_type;
                grad = legacy_syntax = TRUE;
                lod = FALSE;
                break;
 
            default:
                break;
        }
    }
 
    if (legacy_syntax)
    {
        if (shadow)
            base = "shadow";
 
        type_part = resource_type_info[resource_type].type_part;
        if (resource_type == WINED3D_SHADER_RESOURCE_TEXTURE_2D && texrect)
            type_part = "2DRect";
        if (!type_part[0] && resource_type != WINED3D_SHADER_RESOURCE_TEXTURE_CUBEARRAY)
            FIXME("Unhandled resource type %#x.\n", resource_type);
 
        if (!lod && grad && !shader_glsl_has_core_grad(gl_info))
        {
            if (gl_info->supported[ARB_SHADER_TEXTURE_LOD])
                suffix = "ARB";
            else
                FIXME("Unsupported grad function.\n");
        }
    }
 
    if (flags & WINED3D_GLSL_SAMPLE_LOAD)
    {
        static const DWORD texel_fetch_flags = WINED3D_GLSL_SAMPLE_LOAD | WINED3D_GLSL_SAMPLE_OFFSET;
        if (flags & ~texel_fetch_flags)
            ERR("Unexpected flags %#x for texelFetch.\n", flags & ~texel_fetch_flags);
 
        base = "texelFetch";
        type_part = "";
    }
 
    sample_function->name = string_buffer_get(priv->string_buffers);
    string_buffer_sprintf(sample_function->name, "%s%s%s%s%s%s", base, type_part, projected ? "Proj" : "",
            lod ? "Lod" : grad ? "Grad" : "", offset ? "Offset" : "", suffix);
 
    shader_glsl_get_coord_size(resource_type, &coord_size, &deriv_size);
    if (shadow)
        ++coord_size;
    sample_function->offset_size = offset ? deriv_size : 0;
    sample_function->coord_mask = (1u << coord_size) - 1;
    sample_function->deriv_mask = (1u << deriv_size) - 1;
    sample_function->output_single_component = shadow && !legacy_syntax;
}
 
static void shader_glsl_release_sample_function(const struct wined3d_shader_context *ctx,
        struct glsl_sample_function *sample_function)
{
    const struct shader_glsl_ctx_priv *priv = ctx->backend_data;
 
    string_buffer_release(priv->string_buffers, sample_function->name);
}
 
static void shader_glsl_append_fixup_arg(char *arguments, const char *reg_name,
        BOOL sign_fixup, enum fixup_channel_source channel_source)
{
    switch(channel_source)
    {
        case CHANNEL_SOURCE_ZERO:
            strcat(arguments, "0.0");
            break;
 
        case CHANNEL_SOURCE_ONE:
            strcat(arguments, "1.0");
            break;
 
        case CHANNEL_SOURCE_X:
            strcat(arguments, reg_name);
            strcat(arguments, ".x");
            break;
 
        case CHANNEL_SOURCE_Y:
            strcat(arguments, reg_name);
            strcat(arguments, ".y");
            break;
 
        case CHANNEL_SOURCE_Z:
            strcat(arguments, reg_name);
            strcat(arguments, ".z");
            break;
 
        case CHANNEL_SOURCE_W:
            strcat(arguments, reg_name);
            strcat(arguments, ".w");
            break;
 
        default:
            FIXME("Unhandled channel source %#x\n", channel_source);
            strcat(arguments, "undefined");
            break;
    }
 
    if (sign_fixup) strcat(arguments, " * 2.0 - 1.0");
}
 
static void shader_glsl_color_correction_ext(struct wined3d_string_buffer *buffer,
        const char *reg_name, DWORD mask, struct color_fixup_desc fixup)
{
    unsigned int mask_size, remaining;
    DWORD fixup_mask = 0;
    char arguments[256];
    char mask_str[6];
 
    if (fixup.x_sign_fixup || fixup.x_source != CHANNEL_SOURCE_X) fixup_mask |= WINED3DSP_WRITEMASK_0;
    if (fixup.y_sign_fixup || fixup.y_source != CHANNEL_SOURCE_Y) fixup_mask |= WINED3DSP_WRITEMASK_1;
    if (fixup.z_sign_fixup || fixup.z_source != CHANNEL_SOURCE_Z) fixup_mask |= WINED3DSP_WRITEMASK_2;
    if (fixup.w_sign_fixup || fixup.w_source != CHANNEL_SOURCE_W) fixup_mask |= WINED3DSP_WRITEMASK_3;
    if (!(mask &= fixup_mask))
        return;
 
    if (is_complex_fixup(fixup))
    {
        enum complex_fixup complex_fixup = get_complex_fixup(fixup);
        FIXME("Complex fixup (%#x) not supported\n",complex_fixup);
        return;
    }
 
    shader_glsl_write_mask_to_str(mask, mask_str);
    mask_size = shader_glsl_get_write_mask_size(mask);
 
    arguments[0] = '\0';
    remaining = mask_size;
    if (mask & WINED3DSP_WRITEMASK_0)
    {
        shader_glsl_append_fixup_arg(arguments, reg_name, fixup.x_sign_fixup, fixup.x_source);
        if (--remaining) strcat(arguments, ", ");
    }
    if (mask & WINED3DSP_WRITEMASK_1)
    {
        shader_glsl_append_fixup_arg(arguments, reg_name, fixup.y_sign_fixup, fixup.y_source);
        if (--remaining) strcat(arguments, ", ");
    }
    if (mask & WINED3DSP_WRITEMASK_2)
    {
        shader_glsl_append_fixup_arg(arguments, reg_name, fixup.z_sign_fixup, fixup.z_source);
        if (--remaining) strcat(arguments, ", ");
    }
    if (mask & WINED3DSP_WRITEMASK_3)
    {
        shader_glsl_append_fixup_arg(arguments, reg_name, fixup.w_sign_fixup, fixup.w_source);
        if (--remaining) strcat(arguments, ", ");
    }
 
    if (mask_size > 1)
        shader_addline(buffer, "%s%s = vec%u(%s);\n", reg_name, mask_str, mask_size, arguments);
    else
        shader_addline(buffer, "%s%s = %s;\n", reg_name, mask_str, arguments);
}
 
static void shader_glsl_color_correction(const struct wined3d_shader_instruction *ins, struct color_fixup_desc fixup)
{
    char reg_name[256];
    BOOL is_color;
 
    shader_glsl_get_register_name(&ins->dst[0].reg, ins->dst[0].reg.data_type, reg_name, &is_color, ins);
    shader_glsl_color_correction_ext(ins->ctx->buffer, reg_name, ins->dst[0].write_mask, fixup);
}
 
static void PRINTF_ATTR(9, 10) shader_glsl_gen_sample_code(const struct wined3d_shader_instruction *ins,
        unsigned int sampler_bind_idx, const struct glsl_sample_function *sample_function, DWORD swizzle,
        const char *dx, const char *dy, const char *bias, const struct wined3d_shader_texel_offset *offset,
        const char *coord_reg_fmt, ...)
{
    static const struct wined3d_shader_texel_offset dummy_offset = {0, 0, 0};
    const struct wined3d_shader_version *version = &ins->ctx->reg_maps->shader_version;
    char dst_swizzle[6];
    struct color_fixup_desc fixup;
    BOOL np2_fixup = FALSE;
    va_list args;
    int ret;
 
    shader_glsl_swizzle_to_str(swizzle, FALSE, ins->dst[0].write_mask, dst_swizzle);
 
    /* If ARB_texture_swizzle is supported we don't need to do anything here.
     * We actually rely on it for vertex shaders and SM4+. */
    if (version->type == WINED3D_SHADER_TYPE_PIXEL && version->major < 4)
    {
        const struct shader_glsl_ctx_priv *priv = ins->ctx->backend_data;
        fixup = priv->cur_ps_args->color_fixup[sampler_bind_idx];
 
        if (priv->cur_ps_args->np2_fixup & (1u << sampler_bind_idx))
            np2_fixup = TRUE;
    }
    else
    {
        fixup = COLOR_FIXUP_IDENTITY;
    }
 
    shader_glsl_append_dst_ext(ins->ctx->buffer, ins, &ins->dst[0], sample_function->data_type);
 
    if (sample_function->output_single_component)
        shader_addline(ins->ctx->buffer, "vec4(");
 
    shader_addline(ins->ctx->buffer, "%s(%s_sampler%u, ",
            sample_function->name->buffer, shader_glsl_get_prefix(version->type), sampler_bind_idx);
 
    for (;;)
    {
        va_start(args, coord_reg_fmt);
        ret = shader_vaddline(ins->ctx->buffer, coord_reg_fmt, args);
        va_end(args);
        if (!ret)
            break;
        if (!string_buffer_resize(ins->ctx->buffer, ret))
            break;
    }
 
    if (np2_fixup)
    {
        const struct shader_glsl_ctx_priv *priv = ins->ctx->backend_data;
        const unsigned char idx = priv->cur_np2fixup_info->idx[sampler_bind_idx];
 
        switch (shader_glsl_get_write_mask_size(sample_function->coord_mask))
        {
            case 1:
                shader_addline(ins->ctx->buffer, " * ps_samplerNP2Fixup[%u].%s",
                        idx >> 1, (idx % 2) ? "z" : "x");
                break;
            case 2:
                shader_addline(ins->ctx->buffer, " * ps_samplerNP2Fixup[%u].%s",
                        idx >> 1, (idx % 2) ? "zw" : "xy");
                break;
            case 3:
                shader_addline(ins->ctx->buffer, " * vec3(ps_samplerNP2Fixup[%u].%s, 1.0)",
                        idx >> 1, (idx % 2) ? "zw" : "xy");
                break;
            case 4:
                shader_addline(ins->ctx->buffer, " * vec4(ps_samplerNP2Fixup[%u].%s, 1.0, 1.0)",
                        idx >> 1, (idx % 2) ? "zw" : "xy");
                break;
        }
    }
    if (sample_function->emulate_lod)
    {
        if (strcmp(bias, "0")) FIXME("Don't know how to emulate lod level %s\n", bias);
        switch (sample_function->emulate_lod)
        {
            case WINED3D_SHADER_RESOURCE_TEXTURE_2DARRAY:
                if (!dx) dx = "vec2(0.0, 0.0)";
                if (!dy) dy = "vec2(0.0, 0.0)";
                break;
 
            case WINED3D_SHADER_RESOURCE_TEXTURE_CUBE:
                if (!dx) dx = "vec3(0.0, 0.0, 0.0)";
                if (!dy) dy = "vec3(0.0, 0.0, 0.0)";
                break;
 
            default:
                break;
        }
        if (!offset) offset = &dummy_offset;
    }
    if (dx && dy)
        shader_addline(ins->ctx->buffer, ", %s, %s", dx, dy);
    else if (bias)
        shader_addline(ins->ctx->buffer, ", %s", bias);
    if (sample_function->offset_size)
    {
        int offset_immdata[4] = {offset->u, offset->v, offset->w};
        shader_addline(ins->ctx->buffer, ", ");
        shader_glsl_append_imm_ivec(ins->ctx->buffer, offset_immdata, sample_function->offset_size);
    }
    shader_addline(ins->ctx->buffer, ")");
 
    if (sample_function->output_single_component)
        shader_addline(ins->ctx->buffer, ")");
 
    shader_addline(ins->ctx->buffer, "%s);\n", dst_swizzle);
 
    if (!is_identity_fixup(fixup))
        shader_glsl_color_correction(ins, fixup);
}
 
static void shader_glsl_fixup_position(struct wined3d_string_buffer *buffer)
{
    /* 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
     * pos_fixup. pos_fixup.y contains 1.0 or -1.0 to turn the rendering
     * upside down for offscreen rendering. pos_fixup.x contains 1.0 to allow
     * a MAD. */
    shader_addline(buffer, "gl_Position.y = gl_Position.y * pos_fixup.y;\n");
    shader_addline(buffer, "gl_Position.xy += pos_fixup.zw * gl_Position.ww;\n");
 
    /* Z coord [0;1]->[-1;1] mapping, see comment in get_projection_matrix()
     * in utils.c
     *
     * Basically we want (in homogeneous coordinates) z = z * 2 - 1. However,
     * shaders are run before the homogeneous divide, so we have to take the w
     * into account: z = ((z / w) * 2 - 1) * w, which is the same as
     * z = z * 2 - w. */
    shader_addline(buffer, "gl_Position.z = gl_Position.z * 2.0 - gl_Position.w;\n");
}
 
/*****************************************************************************
 * Begin processing individual instruction opcodes
 ****************************************************************************/
 
static void shader_glsl_binop(const struct wined3d_shader_instruction *ins)
{
    struct wined3d_string_buffer *buffer = ins->ctx->buffer;
    struct glsl_src_param src0_param;
    struct glsl_src_param src1_param;
    DWORD write_mask;
    const char *op;
 
    /* Determine the GLSL operator to use based on the opcode */
    switch (ins->handler_idx)
    {
        case WINED3DSIH_ADD:  op = "+";  break;
        case WINED3DSIH_AND:  op = "&";  break;
        case WINED3DSIH_DIV:  op = "/";  break;
        case WINED3DSIH_IADD: op = "+";  break;
        case WINED3DSIH_ISHL: op = "<<"; break;
        case WINED3DSIH_ISHR: op = ">>"; break;
        case WINED3DSIH_MUL:  op = "*";  break;
        case WINED3DSIH_OR:   op = "|";  break;
        case WINED3DSIH_SUB:  op = "-";  break;
        case WINED3DSIH_USHR: op = ">>"; break;
        case WINED3DSIH_XOR:  op = "^";  break;
        default:
            op = "<unhandled operator>";
            FIXME("Opcode %s not yet handled in GLSL.\n", debug_d3dshaderinstructionhandler(ins->handler_idx));
            break;
    }
 
    write_mask = shader_glsl_append_dst(buffer, ins);
    shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
    shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
    shader_addline(buffer, "%s %s %s);\n", src0_param.param_str, op, src1_param.param_str);
}
 
static void shader_glsl_relop(const struct wined3d_shader_instruction *ins)
{
    struct wined3d_string_buffer *buffer = ins->ctx->buffer;
    struct glsl_src_param src0_param;
    struct glsl_src_param src1_param;
    unsigned int mask_size;
    DWORD write_mask;
    const char *op;
 
    write_mask = shader_glsl_append_dst(buffer, ins);
    mask_size = shader_glsl_get_write_mask_size(write_mask);
    shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
    shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
 
    if (mask_size > 1)
    {
        switch (ins->handler_idx)
        {
            case WINED3DSIH_EQ:  op = "equal"; break;
            case WINED3DSIH_IEQ: op = "equal"; break;
            case WINED3DSIH_GE:  op = "greaterThanEqual"; break;
            case WINED3DSIH_IGE: op = "greaterThanEqual"; break;
            case WINED3DSIH_UGE: op = "greaterThanEqual"; break;
            case WINED3DSIH_LT:  op = "lessThan"; break;
            case WINED3DSIH_ILT: op = "lessThan"; break;
            case WINED3DSIH_ULT: op = "lessThan"; break;
            case WINED3DSIH_NE:  op = "notEqual"; break;
            case WINED3DSIH_INE: op = "notEqual"; break;
            default:
                op = "<unhandled operator>";
                ERR("Unhandled opcode %#x.\n", ins->handler_idx);
                break;
        }
 
        shader_addline(buffer, "uvec%u(%s(%s, %s)) * 0xffffffffu);\n",
                mask_size, op, src0_param.param_str, src1_param.param_str);
    }
    else
    {
        switch (ins->handler_idx)
        {
            case WINED3DSIH_EQ:  op = "=="; break;
            case WINED3DSIH_IEQ: op = "=="; break;
            case WINED3DSIH_GE:  op = ">="; break;
            case WINED3DSIH_IGE: op = ">="; break;
            case WINED3DSIH_UGE: op = ">="; break;
            case WINED3DSIH_LT:  op = "<"; break;
            case WINED3DSIH_ILT: op = "<"; break;
            case WINED3DSIH_ULT: op = "<"; break;
            case WINED3DSIH_NE:  op = "!="; break;
            case WINED3DSIH_INE: op = "!="; break;
            default:
                op = "<unhandled operator>";
                ERR("Unhandled opcode %#x.\n", ins->handler_idx);
                break;
        }
 
        shader_addline(buffer, "%s %s %s ? 0xffffffffu : 0u);\n",
                src0_param.param_str, op, src1_param.param_str);
    }
}
 
static void shader_glsl_unary_op(const struct wined3d_shader_instruction *ins)
{
    struct glsl_src_param src_param;
    DWORD write_mask;
    const char *op;
 
    switch (ins->handler_idx)
    {
        case WINED3DSIH_INEG: op = "-"; break;
        case WINED3DSIH_NOT:  op = "~"; break;
        default:
            op = "<unhandled operator>";
            ERR("Unhandled opcode %s.\n",
                    debug_d3dshaderinstructionhandler(ins->handler_idx));
            break;
    }
 
    write_mask = shader_glsl_append_dst(ins->ctx->buffer, ins);
    shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src_param);
    shader_addline(ins->ctx->buffer, "%s%s);\n", op, src_param.param_str);
}
 
static void shader_glsl_mul_extended(const struct wined3d_shader_instruction *ins)
{
    struct wined3d_string_buffer *buffer = ins->ctx->buffer;
    struct glsl_src_param src0_param;
    struct glsl_src_param src1_param;
    DWORD write_mask;
 
    /* If we have ARB_gpu_shader5, we can use imulExtended() / umulExtended().
     * If not, we can emulate it. */
    if (ins->dst[0].reg.type != WINED3DSPR_NULL)
        FIXME("64-bit integer multiplies not implemented.\n");
 
    if (ins->dst[1].reg.type != WINED3DSPR_NULL)
    {
        write_mask = shader_glsl_append_dst_ext(buffer, ins, &ins->dst[1], ins->dst[1].reg.data_type);
        shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
        shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
 
        shader_addline(ins->ctx->buffer, "%s * %s);\n",
                src0_param.param_str, src1_param.param_str);
    }
}
 
static void shader_glsl_udiv(const struct wined3d_shader_instruction *ins)
{
    struct wined3d_string_buffer *buffer = ins->ctx->buffer;
    struct glsl_src_param src0_param, src1_param;
    DWORD write_mask;
 
    if (ins->dst[0].reg.type != WINED3DSPR_NULL)
    {
        if (ins->dst[1].reg.type != WINED3DSPR_NULL)
        {
            char dst_mask[6];
 
            write_mask = shader_glsl_get_write_mask(&ins->dst[0], dst_mask);
            shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
            shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
            shader_addline(buffer, "tmp0%s = uintBitsToFloat(%s / %s);\n",
                    dst_mask, src0_param.param_str, src1_param.param_str);
 
            write_mask = shader_glsl_append_dst_ext(buffer, ins, &ins->dst[1], ins->dst[1].reg.data_type);
            shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
            shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
            shader_addline(buffer, "%s %% %s);\n", src0_param.param_str, src1_param.param_str);
 
            shader_glsl_append_dst_ext(buffer, ins, &ins->dst[0], WINED3D_DATA_FLOAT);
            shader_addline(buffer, "tmp0%s);\n", dst_mask);
        }
        else
        {
            write_mask = shader_glsl_append_dst_ext(buffer, ins, &ins->dst[0], ins->dst[0].reg.data_type);
            shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
            shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
            shader_addline(buffer, "%s / %s);\n", src0_param.param_str, src1_param.param_str);
        }
    }
    else if (ins->dst[1].reg.type != WINED3DSPR_NULL)
    {
        write_mask = shader_glsl_append_dst_ext(buffer, ins, &ins->dst[1], ins->dst[1].reg.data_type);
        shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
        shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
        shader_addline(buffer, "%s %% %s);\n", src0_param.param_str, src1_param.param_str);
    }
}
 
/* Process the WINED3DSIO_MOV opcode using GLSL (dst = src) */
static void shader_glsl_mov(const struct wined3d_shader_instruction *ins)
{
    const struct wined3d_gl_info *gl_info = ins->ctx->gl_info;
    struct wined3d_string_buffer *buffer = ins->ctx->buffer;
    struct glsl_src_param src0_param;
    DWORD write_mask;
 
    write_mask = shader_glsl_append_dst(buffer, ins);
    shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
 
    /* In vs_1_1 WINED3DSIO_MOV can write to the address register. In later
     * shader versions WINED3DSIO_MOVA is used for this. */
    if (ins->ctx->reg_maps->shader_version.major == 1
            && ins->ctx->reg_maps->shader_version.type == WINED3D_SHADER_TYPE_VERTEX
            && ins->dst[0].reg.type == WINED3DSPR_ADDR)
    {
        /* This is a simple floor() */
        unsigned int mask_size = shader_glsl_get_write_mask_size(write_mask);
        if (mask_size > 1) {
            shader_addline(buffer, "ivec%d(floor(%s)));\n", mask_size, src0_param.param_str);
        } else {
            shader_addline(buffer, "int(floor(%s)));\n", src0_param.param_str);
        }
    }
    else if (ins->handler_idx == WINED3DSIH_MOVA)
    {
        unsigned int mask_size = shader_glsl_get_write_mask_size(write_mask);
 
        if (shader_glsl_get_version(gl_info) >= 130 || gl_info->supported[EXT_GPU_SHADER4])
        {
            if (mask_size > 1)
                shader_addline(buffer, "ivec%d(round(%s)));\n", mask_size, src0_param.param_str);
            else
                shader_addline(