path: root/src/shaders.c

/*
 * This file is part of libplacebo.
 *
 * libplacebo 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.
 *
 * libplacebo 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 libplacebo. If not, see <http://www.gnu.org/licenses/>.
 */

#include <stdio.h>
#include <math.h>

#include "common.h"
#include "log.h"
#include "shaders.h"

pl_shader_info pl_shader_info_ref(pl_shader_info pinfo)
{
    struct sh_info *info = (struct sh_info *) pinfo;
    if (!info)
        return NULL;

    pl_rc_ref(&info->rc);
    return &info->info;
}

void pl_shader_info_deref(pl_shader_info *pinfo)
{
    struct sh_info *info = (struct sh_info *) *pinfo;
    if (!info)
        return;

    if (pl_rc_deref(&info->rc))
        pl_free(info);
    *pinfo = NULL;
}

static struct sh_info *sh_info_alloc(void *alloc)
{
    struct sh_info *info = pl_zalloc_ptr(alloc, info);
    info->tmp = pl_tmp(info);
    pl_rc_init(&info->rc);
    return info;
}

// Re-use `sh_info` allocation if possible, allocate new otherwise
static struct sh_info *sh_info_recycle(struct sh_info *info)
{
    if (!pl_rc_deref(&info->rc))
        return sh_info_alloc(NULL);

    memset(&info->info, 0, sizeof(info->info)); // reset public fields
    pl_free_children(info->tmp);
    pl_rc_ref(&info->rc);
    info->desc.len = 0;
    info->steps.num = 0;
    return info;
}

static uint8_t reverse_bits(uint8_t x)
{
    static const uint8_t reverse_nibble[16] = {
        0x0, 0x8, 0x4, 0xc, 0x2, 0xa, 0x6, 0xe,
        0x1, 0x9, 0x5, 0xd, 0x3, 0xb, 0x7, 0xf,
    };

    return reverse_nibble[x & 0xF] << 4 | reverse_nibble[x >> 4];
}

static void init_shader(pl_shader sh, const struct pl_shader_params *params)
{
    if (params) {
        sh->info->info.params = *params;

        // To avoid collisions for shaders with very high number of
        // identifiers, pack the shader ID into the highest bits (MSB -> LSB)
        pl_static_assert(sizeof(sh->prefix) > sizeof(params->id));
        const int shift = 8 * (sizeof(sh->prefix) - sizeof(params->id));
        sh->prefix = reverse_bits(params->id) << shift;
    }

    sh->name = sh_fresh(sh, "main");
}

pl_shader pl_shader_alloc(pl_log log, const struct pl_shader_params *params)
{
    static const int glsl_ver_req = 130;
    if (params && params->glsl.version && params->glsl.version < 130) {
        pl_err(log, "Requested GLSL version %d too low (required: %d)",
               params->glsl.version, glsl_ver_req);
        return NULL;
    }

    pl_shader sh = pl_alloc_ptr(NULL, sh);
    *sh = (struct pl_shader_t) {
        .log        = log,
        .tmp        = pl_tmp(sh),
        .info       = sh_info_alloc(NULL),
        .mutable    = true,
    };

    for (int i = 0; i < PL_ARRAY_SIZE(sh->buffers); i++)
        sh->buffers[i] = pl_str_builder_alloc(sh);

    init_shader(sh, params);
    return sh;
}

static void sh_obj_deref(pl_shader_obj obj);

void sh_deref(pl_shader sh)
{
    pl_free_children(sh->tmp);

    for (int i = 0; i < sh->obj.num; i++)
        sh_obj_deref(sh->obj.elem[i]);
    sh->obj.num = 0;
}

void pl_shader_free(pl_shader *psh)
{
    pl_shader sh = *psh;
    if (!sh)
        return;

    sh_deref(sh);
    pl_shader_info_deref((pl_shader_info *) &sh->info);
    pl_free_ptr(psh);
}

void pl_shader_reset(pl_shader sh, const struct pl_shader_params *params)
{
    sh_deref(sh);

    struct pl_shader_t new = {
        .log            = sh->log,
        .tmp            = sh->tmp,
        .info           = sh_info_recycle(sh->info),
        .data.buf       = sh->data.buf,
        .mutable        = true,

        // Preserve array allocations
        .obj.elem       = sh->obj.elem,
        .vas.elem       = sh->vas.elem,
        .vars.elem      = sh->vars.elem,
        .descs.elem     = sh->descs.elem,
        .consts.elem    = sh->consts.elem,
    };

    // Preserve buffer allocations
    memcpy(new.buffers, sh->buffers, sizeof(new.buffers));
    for (int i = 0; i < PL_ARRAY_SIZE(new.buffers); i++)
        pl_str_builder_reset(new.buffers[i]);

    *sh = new;
    init_shader(sh, params);
}

static void *sh_alloc(pl_shader sh, size_t size, size_t align)
{
    const size_t offset = PL_ALIGN2(sh->data.len, align);
    const size_t req_size = offset + size;
    if (req_size <= pl_get_size(sh->data.buf)) {
        sh->data.len = offset + size;
        return sh->data.buf + offset;
    }

    // We can't realloc this buffer because various pointers will be left
    // dangling, so just reparent it onto `sh->tmp` (so it will be cleaned
    // up when the shader is next reset) and allocate a new, larger buffer
    // in its place
    const size_t new_size = PL_MAX(req_size << 1, 256);
    pl_steal(sh->tmp, sh->data.buf);
    sh->data.buf = pl_alloc(sh, new_size);
    sh->data.len = size;
    return sh->data.buf;
}

static void *sh_memdup(pl_shader sh, const void *data, size_t size, size_t align)
{
    if (!size)
        return NULL;

    void *dst = sh_alloc(sh, size, align);
    assert(data);
    memcpy(dst, data, size);
    return dst;
}

bool pl_shader_is_failed(const pl_shader sh)
{
    return sh->failed;
}

struct pl_glsl_version sh_glsl(const pl_shader sh)
{
    if (SH_PARAMS(sh).glsl.version)
        return SH_PARAMS(sh).glsl;

    if (SH_GPU(sh))
        return SH_GPU(sh)->glsl;

    return (struct pl_glsl_version) { .version = 130 };
}

bool sh_try_compute(pl_shader sh, int bw, int bh, bool flex, size_t mem)
{
    pl_assert(bw && bh);
    int *sh_bw = &sh->group_size[0];
    int *sh_bh = &sh->group_size[1];

    struct pl_glsl_version glsl = sh_glsl(sh);
    if (!glsl.compute) {
        PL_TRACE(sh, "Disabling compute shader due to missing `compute` support");
        return false;
    }

    if (sh->shmem + mem > glsl.max_shmem_size) {
        PL_TRACE(sh, "Disabling compute shader due to insufficient shmem");
        return false;
    }

    if (sh->type == SH_FRAGMENT) {
        PL_TRACE(sh, "Disabling compute shader because shader is already marked "
                 "as fragment shader");
        return false;
    }

    if (bw > glsl.max_group_size[0] ||
        bh > glsl.max_group_size[1] ||
        (bw * bh) > glsl.max_group_threads)
    {
        if (!flex) {
            PL_TRACE(sh, "Disabling compute shader due to exceeded group "
                     "thread count.");
            return false;
        } else {
            // Pick better group sizes
            bw = PL_MIN(bw, glsl.max_group_size[0]);
            bh = glsl.max_group_threads / bw;
        }
    }

    sh->shmem += mem;

    // If the current shader is either not a compute shader, or we have no
    // choice but to override the metadata, always do so
    if (sh->type != SH_COMPUTE || (sh->flexible_work_groups && !flex)) {
        *sh_bw = bw;
        *sh_bh = bh;
        sh->type = SH_COMPUTE;
        sh->flexible_work_groups = flex;
        return true;
    }

    // If both shaders are flexible, pick the larger of the two
    if (sh->flexible_work_groups && flex) {
        *sh_bw = PL_MAX(*sh_bw, bw);
        *sh_bh = PL_MAX(*sh_bh, bh);
        pl_assert(*sh_bw * *sh_bh <= glsl.max_group_threads);
        return true;
    }

    // At this point we're looking only at a non-flexible compute shader
    pl_assert(sh->type == SH_COMPUTE && !sh->flexible_work_groups);
    if (!flex) {
        // Ensure parameters match
        if (bw != *sh_bw || bh != *sh_bh) {
            PL_TRACE(sh, "Disabling compute shader due to incompatible group "
                     "sizes %dx%d and %dx%d", *sh_bw, *sh_bh, bw, bh);
            sh->shmem -= mem;
            return false;
        }
    }

    return true;
}

bool pl_shader_is_compute(const pl_shader sh)
{
    return sh->type == SH_COMPUTE;
}

bool pl_shader_output_size(const pl_shader sh, int *w, int *h)
{
    if (!sh->output_w || !sh->output_h)
        return false;

    *w = sh->transpose ? sh->output_h : sh->output_w;
    *h = sh->transpose ? sh->output_w : sh->output_h;
    return true;
}

ident_t sh_fresh(pl_shader sh, const char *name)
{
    unsigned short id = ++sh->fresh;
    assert(!(sh->prefix & id));
    id |= sh->prefix;

    assert(name);
    return sh_mkident(id, name);
}

static inline ident_t sh_fresh_name(pl_shader sh, const char **pname)
{
    ident_t id = sh_fresh(sh, *pname);
    *pname = sh_ident_pack(id);
    return id;
}

ident_t sh_var(pl_shader sh, struct pl_shader_var sv)
{
    ident_t id = sh_fresh_name(sh, &sv.var.name);
    struct pl_var_layout layout = pl_var_host_layout(0, &sv.var);
    sv.data = sh_memdup(sh, sv.data, layout.size, layout.stride);
    PL_ARRAY_APPEND(sh, sh->vars, sv);
    return id;
}

ident_t sh_var_int(pl_shader sh, const char *name, int val, bool dynamic)
{
    return sh_var(sh, (struct pl_shader_var) {
        .var     = pl_var_int(name),
        .data    = &val,
        .dynamic = dynamic,
    });
}

ident_t sh_var_uint(pl_shader sh, const char *name, unsigned int val, bool dynamic)
{
    return sh_var(sh, (struct pl_shader_var) {
        .var     = pl_var_uint(name),
        .data    = &val,
        .dynamic = dynamic,
    });
}

ident_t sh_var_float(pl_shader sh, const char *name, float val, bool dynamic)
{
    return sh_var(sh, (struct pl_shader_var) {
        .var     = pl_var_float(name),
        .data    = &val,
        .dynamic = dynamic,
    });
}

ident_t sh_var_mat3(pl_shader sh, const char *name, pl_matrix3x3 val)
{
    return sh_var(sh, (struct pl_shader_var) {
        .var     = pl_var_mat3(name),
        .data    = PL_TRANSPOSE_3X3(val.m),
    });
}

ident_t sh_desc(pl_shader sh, struct pl_shader_desc sd)
{
    switch (sd.desc.type) {
    case PL_DESC_BUF_UNIFORM:
    case PL_DESC_BUF_STORAGE:
        for (int i = 0; i < sh->descs.num; i++) // ensure uniqueness
            pl_assert(sh->descs.elem[i].binding.object != sd.binding.object);
        size_t bsize = sizeof(sd.buffer_vars[0]) * sd.num_buffer_vars;
        sd.buffer_vars = sh_memdup(sh, sd.buffer_vars, bsize,
                                   alignof(struct pl_buffer_var));
        for (int i = 0; i < sd.num_buffer_vars; i++) {
            struct pl_var *bv = &sd.buffer_vars[i].var;
            const char *name = bv->name;
            GLSLP("#define %s "$"\n", name, sh_fresh_name(sh, &bv->name));
        }
        break;

    case PL_DESC_BUF_TEXEL_UNIFORM:
    case PL_DESC_BUF_TEXEL_STORAGE:
    case PL_DESC_SAMPLED_TEX:
    case PL_DESC_STORAGE_IMG:
        pl_assert(!sd.num_buffer_vars);
        break;

    case PL_DESC_INVALID:
    case PL_DESC_TYPE_COUNT:
        pl_unreachable();
    }

    ident_t id = sh_fresh_name(sh, &sd.desc.name);
    PL_ARRAY_APPEND(sh, sh->descs, sd);
    return id;
}

ident_t sh_const(pl_shader sh, struct pl_shader_const sc)
{
    if (SH_PARAMS(sh).dynamic_constants && !sc.compile_time) {
        return sh_var(sh, (struct pl_shader_var) {
            .var = {
                .name = sc.name,
                .type = sc.type,
                .dim_v = 1,
                .dim_m = 1,
                .dim_a = 1,
            },
            .data = sc.data,
        });
    }

    ident_t id = sh_fresh_name(sh, &sc.name);

    pl_gpu gpu = SH_GPU(sh);
    if (gpu && gpu->limits.max_constants) {
        if (!sc.compile_time || gpu->limits.array_size_constants) {
            size_t size = pl_var_type_size(sc.type);
            sc.data = sh_memdup(sh, sc.data, size, size);
            PL_ARRAY_APPEND(sh, sh->consts, sc);
            return id;
        }
    }

    // Fallback for GPUs without specialization constants
    switch (sc.type) {
    case PL_VAR_SINT:
        GLSLH("const int "$" = %d; \n", id, *(int *) sc.data);
        return id;
    case PL_VAR_UINT:
        GLSLH("const uint "$" = uint(%u); \n", id, *(unsigned int *) sc.data);
        return id;
    case PL_VAR_FLOAT:
        GLSLH("const float "$" = float(%f); \n", id, *(float *) sc.data);
        return id;
    case PL_VAR_INVALID:
    case PL_VAR_TYPE_COUNT:
        break;
    }

    pl_unreachable();
}

ident_t sh_const_int(pl_shader sh, const char *name, int val)
{
    return sh_const(sh, (struct pl_shader_const) {
        .type = PL_VAR_SINT,
        .name = name,
        .data = &val,
    });
}

ident_t sh_const_uint(pl_shader sh, const char *name, unsigned int val)
{
    return sh_const(sh, (struct pl_shader_const) {
        .type = PL_VAR_UINT,
        .name = name,
        .data = &val,
    });
}

ident_t sh_const_float(pl_shader sh, const char *name, float val)
{
    return sh_const(sh, (struct pl_shader_const) {
        .type = PL_VAR_FLOAT,
        .name = name,
        .data = &val,
    });
}

ident_t sh_attr(pl_shader sh, struct pl_shader_va sva)
{
    const size_t vsize = sva.attr.fmt->texel_size;
    uint8_t *data = sh_alloc(sh, vsize * 4, vsize);
    for (int i = 0; i < 4; i++) {
        memcpy(data, sva.data[i], vsize);
        sva.data[i] = data;
        data += vsize;
    }

    ident_t id = sh_fresh_name(sh, &sva.attr.name);
    PL_ARRAY_APPEND(sh, sh->vas, sva);
    return id;
}

ident_t sh_attr_vec2(pl_shader sh, const char *name, const pl_rect2df *rc)
{
    pl_gpu gpu = SH_GPU(sh);
    if (!gpu) {
        SH_FAIL(sh, "Failed adding vertex attr '%s': No GPU available!", name);
        return NULL_IDENT;
    }

    pl_fmt fmt = pl_find_vertex_fmt(gpu, PL_FMT_FLOAT, 2);
    if (!fmt) {
        SH_FAIL(sh, "Failed adding vertex attr '%s': no vertex fmt!", name);
        return NULL_IDENT;
    }

    float verts[4][2] = {
        { rc->x0, rc->y0 },
        { rc->x1, rc->y0 },
        { rc->x0, rc->y1 },
        { rc->x1, rc->y1 },
    };

    return sh_attr(sh, (struct pl_shader_va) {
        .attr = {
            .name     = name,
            .fmt      = pl_find_vertex_fmt(gpu, PL_FMT_FLOAT, 2),
        },
        .data = { verts[0], verts[1], verts[2], verts[3] },
    });
}

ident_t sh_bind(pl_shader sh, pl_tex tex,
                enum pl_tex_address_mode address_mode,
                enum pl_tex_sample_mode sample_mode,
                const char *name, const pl_rect2df *rect,
                ident_t *out_pos, ident_t *out_pt)
{
    if (pl_tex_params_dimension(tex->params) != 2) {
        SH_FAIL(sh, "Failed binding texture '%s': not a 2D texture!", name);
        return NULL_IDENT;
    }

    if (!tex->params.sampleable) {
        SH_FAIL(sh, "Failed binding texture '%s': texture not sampleable!", name);
        return NULL_IDENT;
    }

    ident_t itex = sh_desc(sh, (struct pl_shader_desc) {
        .desc = {
            .name = name,
            .type = PL_DESC_SAMPLED_TEX,
        },
        .binding = {
            .object = tex,
            .address_mode = address_mode,
            .sample_mode = sample_mode,
        },
    });

    float sx, sy;
    if (tex->sampler_type == PL_SAMPLER_RECT) {
        sx = 1.0;
        sy = 1.0;
    } else {
        sx = 1.0 / tex->params.w;
        sy = 1.0 / tex->params.h;
    }

    if (out_pos) {
        pl_rect2df full = {
            .x1 = tex->params.w,
            .y1 = tex->params.h,
        };

        rect = PL_DEF(rect, &full);
        *out_pos = sh_attr_vec2(sh, "tex_coord", &(pl_rect2df) {
            .x0 = sx * rect->x0, .y0 = sy * rect->y0,
            .x1 = sx * rect->x1, .y1 = sy * rect->y1,
        });
    }

    if (out_pt) {
        *out_pt = sh_var(sh, (struct pl_shader_var) {
            .var  = pl_var_vec2("tex_pt"),
            .data = &(float[2]) {sx, sy},
        });
    }

    return itex;
}

bool sh_buf_desc_append(void *alloc, pl_gpu gpu,
                        struct pl_shader_desc *buf_desc,
                        struct pl_var_layout *out_layout,
                        const struct pl_var new_var)
{
    struct pl_buffer_var bv = { .var = new_var };
    size_t cur_size = sh_buf_desc_size(buf_desc);

    switch (buf_desc->desc.type) {
    case PL_DESC_BUF_UNIFORM:
        bv.layout = pl_std140_layout(cur_size, &new_var);
        if (bv.layout.offset + bv.layout.size > gpu->limits.max_ubo_size)
            return false;
        break;
    case PL_DESC_BUF_STORAGE:
        bv.layout = pl_std430_layout(cur_size, &new_var);
        if (bv.layout.offset + bv.layout.size > gpu->limits.max_ssbo_size)
            return false;
        break;
    case PL_DESC_INVALID:
    case PL_DESC_SAMPLED_TEX:
    case PL_DESC_STORAGE_IMG:
    case PL_DESC_BUF_TEXEL_UNIFORM:
    case PL_DESC_BUF_TEXEL_STORAGE:
    case PL_DESC_TYPE_COUNT:
        pl_unreachable();
    }

    if (out_layout)
        *out_layout = bv.layout;
    PL_ARRAY_APPEND_RAW(alloc, buf_desc->buffer_vars, buf_desc->num_buffer_vars, bv);
    return true;
}

size_t sh_buf_desc_size(const struct pl_shader_desc *buf_desc)
{
    if (!buf_desc->num_buffer_vars)
        return 0;

    const struct pl_buffer_var *last;
    last = &buf_desc->buffer_vars[buf_desc->num_buffer_vars - 1];
    return last->layout.offset + last->layout.size;
}

void sh_describef(pl_shader sh, const char *fmt, ...)
{
    va_list ap;
    va_start(ap, fmt);
    sh_describe(sh, pl_vasprintf(sh->info->tmp, fmt, ap));
    va_end(ap);
}

static const char *insigs[] = {
    [PL_SHADER_SIG_NONE]  = "",
    [PL_SHADER_SIG_COLOR] = "vec4 color",
};

static const char *outsigs[] = {
    [PL_SHADER_SIG_NONE]  = "void",
    [PL_SHADER_SIG_COLOR] = "vec4",
};

static const char *retvals[] = {
    [PL_SHADER_SIG_NONE]  = "",
    [PL_SHADER_SIG_COLOR] = "return color;",
};

// libplacebo currently only allows 2D samplers for shader signatures
static const char *samplers2D[] = {
    [PL_SAMPLER_NORMAL]     = "sampler2D",
    [PL_SAMPLER_RECT]       = "sampler2DRect",
    [PL_SAMPLER_EXTERNAL]   = "samplerExternalOES",
};

ident_t sh_subpass(pl_shader sh, pl_shader sub)
{
    pl_assert(sh->mutable);

    if (sh->prefix == sub->prefix) {
        PL_TRACE(sh, "Can't merge shaders: conflicting identifiers!");
        return NULL_IDENT;
    }

    // Check for shader compatibility
    int res_w = PL_DEF(sh->output_w, sub->output_w),
        res_h = PL_DEF(sh->output_h, sub->output_h);

    if ((sub->output_w && res_w != sub->output_w) ||
        (sub->output_h && res_h != sub->output_h))
    {
        PL_TRACE(sh, "Can't merge shaders: incompatible sizes: %dx%d and %dx%d",
                 sh->output_w, sh->output_h, sub->output_w, sub->output_h);
        return NULL_IDENT;
    }

    if (sub->type == SH_COMPUTE) {
        int subw = sub->group_size[0],
            subh = sub->group_size[1];
        bool flex = sub->flexible_work_groups;

        if (!sh_try_compute(sh, subw, subh, flex, sub->shmem)) {
            PL_TRACE(sh, "Can't merge shaders: incompatible block sizes or "
                     "exceeded shared memory resource capabilities");
            return NULL_IDENT;
        }
    }

    sh->output_w = res_w;
    sh->output_h = res_h;

    // Append the prelude and header
    pl_str_builder_concat(sh->buffers[SH_BUF_PRELUDE], sub->buffers[SH_BUF_PRELUDE]);
    pl_str_builder_concat(sh->buffers[SH_BUF_HEADER], sub->buffers[SH_BUF_HEADER]);

    // Append the body as a new header function
    if (sub->input == PL_SHADER_SIG_SAMPLER) {
        pl_assert(sub->sampler_prefix);
        GLSLH("%s "$"(%c%s src_tex, vec2 tex_coord) {\n",
              outsigs[sub->output], sub->name,
              sub->sampler_prefix, samplers2D[sub->sampler_type]);
    } else {
        GLSLH("%s "$"(%s) {\n",
              outsigs[sub->output], sub->name, insigs[sub->input]);
    }
    pl_str_builder_concat(sh->buffers[SH_BUF_HEADER], sub->buffers[SH_BUF_BODY]);
    GLSLH("%s\n}\n\n", retvals[sub->output]);

    // Steal all inputs and objects from the subpass
#define ARRAY_STEAL(arr) do                 \
{                                           \
    PL_ARRAY_CONCAT(sh, sh->arr, sub->arr); \
    sub->arr.num = 0;                       \
} while (0)

    ARRAY_STEAL(obj);
    ARRAY_STEAL(vas);
    ARRAY_STEAL(vars);
    ARRAY_STEAL(descs);
    ARRAY_STEAL(consts);
#undef ARRAY_STEAL

    // Steal the scratch buffer (if it holds data)
    if (sub->data.len) {
        pl_steal(sh->tmp, sub->data.buf);
        sub->data = (pl_str) {0};
    }

    // Steal all temporary allocations and mark the child as unusable
    pl_steal(sh->tmp, sub->tmp);
    sub->tmp = pl_tmp(sub);
    sub->failed = true;

    // Steal the shader steps array (and allocations)
    pl_assert(pl_rc_count(&sub->info->rc) == 1);
    PL_ARRAY_CONCAT(sh->info, sh->info->steps, sub->info->steps);
    pl_steal(sh->info->tmp, sub->info->tmp);
    sub->info->tmp = pl_tmp(sub->info);
    sub->info->steps.num = 0; // sanity

    return sub->name;
}

pl_str_builder sh_finalize_internal(pl_shader sh)
{
    pl_assert(sh->mutable); // this function should only ever be called once
    if (sh->failed)
        return NULL;

    // Padding for readability
    GLSLP("\n");

    // Concatenate everything onto the prelude to form the final output
    pl_str_builder_concat(sh->buffers[SH_BUF_PRELUDE], sh->buffers[SH_BUF_HEADER]);

    if (sh->input == PL_SHADER_SIG_SAMPLER) {
        pl_assert(sh->sampler_prefix);
        GLSLP("%s "$"(%c%s src_tex, vec2 tex_coord) {\n",
              outsigs[sh->output], sh->name,
              sh->sampler_prefix,
              samplers2D[sh->sampler_type]);
    } else {
        GLSLP("%s "$"(%s) {\n", outsigs[sh->output], sh->name, insigs[sh->input]);
    }

    pl_str_builder_concat(sh->buffers[SH_BUF_PRELUDE], sh->buffers[SH_BUF_BODY]);
    pl_str_builder_concat(sh->buffers[SH_BUF_PRELUDE], sh->buffers[SH_BUF_FOOTER]);
    GLSLP("%s\n}\n\n", retvals[sh->output]);

    // Generate the shader info
    struct sh_info *info = sh->info;
    info->info.steps = info->steps.elem;
    info->info.num_steps = info->steps.num;
    info->info.description = "(unknown shader)";

    // Generate pretty description
    for (int i = 0; i < info->steps.num; i++) {
        const char *step = info->steps.elem[i];

        // Prevent duplicates. We're okay using a weak equality check here
        // because most pass descriptions are static strings.
        for (int j = 0; j < i; j++) {
            if (info->steps.elem[j] == step)
                goto next_step;
        }

        int count = 1;
        for (int j = i+1; j < info->steps.num; j++) {
            if (info->steps.elem[j] == step)
                count++;
        }

        const char *prefix = i > 0 ? ", " : "";
        if (count > 1) {
            pl_str_append_asprintf(info, &info->desc, "%s%s x%d",
                                   prefix, step, count);
        } else {
            pl_str_append_asprintf(info, &info->desc, "%s%s", prefix, step);
        }

next_step: ;
    }

    if (info->desc.len)
        info->info.description = (char *) info->desc.buf;

    sh->mutable = false;
    return sh->buffers[SH_BUF_PRELUDE];
}

const struct pl_shader_res *pl_shader_finalize(pl_shader sh)
{
    if (sh->failed) {
        return NULL;
    } else if (!sh->mutable) {
        return &sh->result;
    }

    pl_shader_info info = &sh->info->info;
    pl_str_builder glsl = sh_finalize_internal(sh);

    // Turn ident_t into friendly strings before passing it to users
#define FIX_IDENT(name) \
    name = sh_ident_tostr(sh_ident_unpack(name))
    for (int i = 0; i < sh->vas.num; i++)
        FIX_IDENT(sh->vas.elem[i].attr.name);
    for (int i = 0; i < sh->vars.num; i++)
        FIX_IDENT(sh->vars.elem[i].var.name);
    for (int i = 0; i < sh->consts.num; i++)
        FIX_IDENT(sh->consts.elem[i].name);
    for (int i = 0; i < sh->descs.num; i++) {
        struct pl_shader_desc *sd = &sh->descs.elem[i];
        FIX_IDENT(sd->desc.name);
        for (int j = 0; j < sd->num_buffer_vars; sd++)
            FIX_IDENT(sd->buffer_vars[j].var.name);
    }
#undef FIX_IDENT

    sh->result = (struct pl_shader_res) {
        .info               = info,
        .glsl               = (char *) pl_str_builder_exec(glsl).buf,
        .name               = sh_ident_tostr(sh->name),
        .input              = sh->input,
        .output             = sh->output,
        .compute_group_size = { sh->group_size[0], sh->group_size[1] },
        .compute_shmem      = sh->shmem,
        .vertex_attribs     = sh->vas.elem,
        .num_vertex_attribs = sh->vas.num,
        .variables          = sh->vars.elem,
        .num_variables      = sh->vars.num,
        .descriptors        = sh->descs.elem,
        .num_descriptors    = sh->descs.num,
        .constants          = sh->consts.elem,
        .num_constants      = sh->consts.num,
        // deprecated fields
        .params             = info->params,
        .steps              = info->steps,
        .num_steps          = info->num_steps,
        .description        = info->description,
    };

    return &sh->result;
}

bool sh_require(pl_shader sh, enum pl_shader_sig insig, int w, int h)
{
    if (sh->failed) {
        SH_FAIL(sh, "Attempting to modify a failed shader!");
        return false;
    }

    if (!sh->mutable) {
        SH_FAIL(sh, "Attempted to modify an immutable shader!");
        return false;
    }

    if ((w && sh->output_w && sh->output_w != w) ||
        (h && sh->output_h && sh->output_h != h))
    {
        SH_FAIL(sh, "Illegal sequence of shader operations: Incompatible "
                "output size requirements %dx%d and %dx%d",
                sh->output_w, sh->output_h, w, h);
        return false;
    }

    static const char *names[] = {
        [PL_SHADER_SIG_NONE]  = "PL_SHADER_SIG_NONE",
        [PL_SHADER_SIG_COLOR] = "PL_SHADER_SIG_COLOR",
    };

    // If we require an input, but there is none available - just get it from
    // the user by turning it into an explicit input signature.
    if (!sh->output && insig) {
        pl_assert(!sh->input);
        sh->input = insig;
    } else if (sh->output != insig) {
        SH_FAIL(sh, "Illegal sequence of shader operations! Current output "
                "signature is '%s', but called operation expects '%s'!",
                names[sh->output], names[insig]);
        return false;
    }

    // All of our shaders end up returning a vec4 color
    sh->output = PL_SHADER_SIG_COLOR;
    sh->output_w = PL_DEF(sh->output_w, w);
    sh->output_h = PL_DEF(sh->output_h, h);
    return true;
}

static void sh_obj_deref(pl_shader_obj obj)
{
    if (!pl_rc_deref(&obj->rc))
        return;

    if (obj->uninit)
        obj->uninit(obj->gpu, obj->priv);

    pl_free(obj);
}

void pl_shader_obj_destroy(pl_shader_obj *ptr)
{
    pl_shader_obj obj = *ptr;
    if (!obj)
        return;

    sh_obj_deref(obj);
    *ptr = NULL;
}

void *sh_require_obj(pl_shader sh, pl_shader_obj *ptr,
                     enum pl_shader_obj_type type, size_t priv_size,
                     void (*uninit)(pl_gpu gpu, void *priv))
{
    if (!ptr)
        return NULL;

    pl_shader_obj obj = *ptr;
    if (obj && obj->gpu != SH_GPU(sh)) {
        SH_FAIL(sh, "Passed pl_shader_obj belongs to different GPU!");
        return NULL;
    }

    if (obj && obj->type != type) {
        SH_FAIL(sh, "Passed pl_shader_obj of wrong type! Shader objects must "
                "always be used with the same type of shader.");
        return NULL;
    }

    if (!obj) {
        obj = pl_zalloc_ptr(NULL, obj);
        pl_rc_init(&obj->rc);
        obj->gpu = SH_GPU(sh);
        obj->type = type;
        obj->priv = pl_zalloc(obj, priv_size);
        obj->uninit = uninit;
    }

    PL_ARRAY_APPEND(sh, sh->obj, obj);
    pl_rc_ref(&obj->rc);

    *ptr = obj;
    return obj->priv;
}

ident_t sh_prng(pl_shader sh, bool temporal, ident_t *p_state)
{
    ident_t randfun = sh_fresh(sh, "rand"),
            state = sh_fresh(sh, "state");

    // Based on pcg3d (http://jcgt.org/published/0009/03/02/)
    GLSLP("#define prng_t uvec3\n");
    GLSLH("vec3 "$"(inout uvec3 s) {                    \n"
          "    s = 1664525u * s + uvec3(1013904223u);   \n"
          "    s.x += s.y * s.z;                        \n"
          "    s.y += s.z * s.x;                        \n"
          "    s.z += s.x * s.y;                        \n"
          "    s ^= s >> 16u;                           \n"
          "    s.x += s.y * s.z;                        \n"
          "    s.y += s.z * s.x;                        \n"
          "    s.z += s.x * s.y;                        \n"
          "    return vec3(s) * 1.0/float(0xFFFFFFFFu); \n"
          "}                                            \n",
          randfun);

    if (temporal) {
        GLSL("uvec3 "$" = uvec3(gl_FragCoord.xy, "$"); \n",
             state, SH_UINT_DYN(SH_PARAMS(sh).index));
    } else {
        GLSL("uvec3 "$" = uvec3(gl_FragCoord.xy, 0.0); \n", state);
    }

    if (p_state)
        *p_state = state;

    ident_t res = sh_fresh(sh, "RAND");
    GLSLH("#define "$" ("$"("$"))\n", res, randfun, state);
    return res;
}