vo_gpu: simplify structs / names

Due to the plethora of historical baggage from different eras getting
confusing, I decided to simplify and unify the struct organization and
naming scheme.

Structs that got renamed:

1. fbodst     -> ra_fbo  (and moved to gpu/context.h)
2. fbotex     -> removed (redundant after 2af2fa7a)
3. fbosurface -> surface
4. img_tex    -> image

In addition to these structs being renamed, all of the names have been
made consistent. The new scheme is as follows:

struct image img;
struct ra_tex *tex;
struct ra_fbo fbo;

This also affects derived names, e.g. indirect_fbo -> indirect_tex.
Notably also, finish_pass_fbo -> finish_pass_tex and finish_pass_direct
-> finish_pass_fbo.

The new equivalent of fbotex_change() is called ra_tex_resize().

This commit (should) contain no logic changes, just renaming a bunch of
crap.

7 files changed, 317 insertions, 355 deletions

diff --git a/video/out/gpu/context.h b/video/out/gpu/context.h
index 42de59b75f..f74087592d 100644
--- a/video/out/gpu/context.h
+++ b/video/out/gpu/context.h
@@ -56,21 +56,26 @@ struct ra_swapchain {
     struct ra_ctx *ctx;
     struct priv *priv;
     const struct ra_swapchain_fns *fns;
+};
 
-    bool flip_v; // flip the rendered image vertically (set by the swapchain)
+// Represents a framebuffer / render target
+struct ra_fbo {
+    struct ra_tex *tex;
+    bool flip; // rendering needs to be inverted
 };
 
 struct ra_swapchain_fns {
     // Gets the current framebuffer depth in bits (0 if unknown). Optional.
     int (*color_depth)(struct ra_swapchain *sw);
 
-    // Retrieves a screenshot of the framebuffer. These are always the right
-    // side up, regardless of ra_swapchain->flip_v. Optional.
+    // Retrieves a screenshot of the framebuffer. Optional.
     struct mp_image *(*screenshot)(struct ra_swapchain *sw);
 
     // Called when rendering starts. Returns NULL on failure. This must be
-    // followed by submit_frame, to submit the rendered frame.
-    struct ra_tex *(*start_frame)(struct ra_swapchain *sw);
+    // followed by submit_frame, to submit the rendered frame. This function
+    // can also fail sporadically, and such errors should be ignored unless
+    // they persist.
+    bool (*start_frame)(struct ra_swapchain *sw, struct ra_fbo *out_fbo);
 
     // Present the frame. Issued in lockstep with start_frame, with rendering
     // commands in between. The `frame` is just there for timing data, for
diff --git a/video/out/gpu/osd.c b/video/out/gpu/osd.c
index f7c325d1db..350736461c 100644
--- a/video/out/gpu/osd.c
+++ b/video/out/gpu/osd.c
@@ -291,7 +291,7 @@ static void get_3d_side_by_side(int stereo_mode, int div[2])
 }
 
 void mpgl_osd_draw_finish(struct mpgl_osd *ctx, int index,
-                          struct gl_shader_cache *sc, struct fbodst target)
+                          struct gl_shader_cache *sc, struct ra_fbo fbo)
 {
     struct mpgl_osd_part *part = ctx->parts[index];
 
@@ -303,7 +303,7 @@ void mpgl_osd_draw_finish(struct mpgl_osd *ctx, int index,
     for (int x = 0; x < div[0]; x++) {
         for (int y = 0; y < div[1]; y++) {
             struct gl_transform t;
-            gl_transform_ortho_fbodst(&t, target);
+            gl_transform_ortho_fbo(&t, fbo);
 
             float a_x = ctx->osd_res.w * x;
             float a_y = ctx->osd_res.h * y;
@@ -317,7 +317,7 @@ void mpgl_osd_draw_finish(struct mpgl_osd *ctx, int index,
     const int *factors = &blend_factors[part->format][0];
     gl_sc_blend(sc, factors[0], factors[1], factors[2], factors[3]);
 
-    gl_sc_dispatch_draw(sc, target.tex, part->vertices, part->num_vertices);
+    gl_sc_dispatch_draw(sc, fbo.tex, part->vertices, part->num_vertices);
 }
 
 static void set_res(struct mpgl_osd *ctx, struct mp_osd_res res, int stereo_mode)
diff --git a/video/out/gpu/osd.h b/video/out/gpu/osd.h
index 6c2b886de3..00fbc4914f 100644
--- a/video/out/gpu/osd.h
+++ b/video/out/gpu/osd.h
@@ -18,7 +18,7 @@ void mpgl_osd_resize(struct mpgl_osd *ctx, struct mp_osd_res res, int stereo_mod
 bool mpgl_osd_draw_prepare(struct mpgl_osd *ctx, int index,
                            struct gl_shader_cache *sc);
 void mpgl_osd_draw_finish(struct mpgl_osd *ctx, int index,
-                          struct gl_shader_cache *sc, struct fbodst target);
+                          struct gl_shader_cache *sc, struct ra_fbo fbo);
 bool mpgl_osd_check_change(struct mpgl_osd *ctx, struct mp_osd_res *res,
                            double pts);
 
diff --git a/video/out/gpu/utils.c b/video/out/gpu/utils.c
index 38d1e9b7cd..daf570cba5 100644
--- a/video/out/gpu/utils.c
+++ b/video/out/gpu/utils.c
@@ -33,7 +33,7 @@ void gl_transform_trans(struct gl_transform t, struct gl_transform *x)
     gl_transform_vec(t, &x->t[0], &x->t[1]);
 }
 
-void gl_transform_ortho_fbodst(struct gl_transform *t, struct fbodst fbo)
+void gl_transform_ortho_fbo(struct gl_transform *t, struct ra_fbo fbo)
 {
     int y_dir = fbo.flip ? -1 : 1;
     gl_transform_ortho(t, 0, fbo.tex->params.w, 0, fbo.tex->params.h * y_dir);
@@ -167,35 +167,24 @@ struct ra_layout std430_layout(struct ra_renderpass_input *inp)
     };
 }
 
-// Create a texture and a FBO using the texture as color attachments.
-//  fmt: texture internal format
-// If the parameters are the same as the previous call, do not touch it.
-bool fbotex_change(struct fbotex *fbo, struct ra *ra, struct mp_log *log,
+// Resize a texture to a new desired size and format if necessary
+bool ra_tex_resize(struct ra *ra, struct mp_log *log, struct ra_tex **tex,
                    int w, int h, const struct ra_format *fmt)
 {
-    int lw = w, lh = h;
-
-    if (fbo->tex) {
-        int cw = w, ch = h;
-        int rw = fbo->tex->params.w, rh = fbo->tex->params.h;
-
-        if (rw == cw && rh == ch && fbo->tex->params.format == fmt)
-            goto done;
+    if (*tex) {
+        struct ra_tex_params cur_params = (*tex)->params;
+        if (cur_params.w == w && cur_params.h == h && cur_params.format == fmt)
+            return true;
     }
 
-    mp_verbose(log, "Create FBO: %dx%d (%dx%d)\n", lw, lh, w, h);
+    mp_verbose(log, "Resizing texture: %dx%d\n", w, h);
 
     if (!fmt || !fmt->renderable || !fmt->linear_filter) {
         mp_err(log, "Format %s not supported.\n", fmt ? fmt->name : "(unset)");
         return false;
     }
 
-    fbotex_uninit(fbo);
-
-    *fbo = (struct fbotex) {
-        .ra = ra,
-    };
-
+    ra_tex_free(ra, tex);
     struct ra_tex_params params = {
         .dimensions = 2,
         .w = w,
@@ -209,32 +198,11 @@ bool fbotex_change(struct fbotex *fbo, struct ra *ra, struct mp_log *log,
         .blit_src = true,
     };
 
-    fbo->tex = ra_tex_create(fbo->ra, &params);
-
-    if (!fbo->tex) {
-        mp_err(log, "Error: framebuffer could not be created.\n");
-        fbotex_uninit(fbo);
-        return false;
-    }
-
-done:
+    *tex = ra_tex_create(ra, &params);
+    if (!*tex)
+        mp_err(log, "Error: texture could not be created.\n");
 
-    fbo->lw = lw;
-    fbo->lh = lh;
-
-    fbo->fbo = (struct fbodst){
-        .tex = fbo->tex,
-    };
-
-    return true;
-}
-
-void fbotex_uninit(struct fbotex *fbo)
-{
-    if (fbo->ra) {
-        ra_tex_free(fbo->ra, &fbo->tex);
-        *fbo = (struct fbotex) {0};
-    }
+    return *tex;
 }
 
 struct timer_pool {
diff --git a/video/out/gpu/utils.h b/video/out/gpu/utils.h
index 9a52246ae7..ac0cbf28d7 100644
--- a/video/out/gpu/utils.h
+++ b/video/out/gpu/utils.h
@@ -4,6 +4,7 @@
 #include <math.h>
 
 #include "ra.h"
+#include "context.h"
 
 // A 3x2 matrix, with the translation part separate.
 struct gl_transform {
@@ -62,12 +63,7 @@ static inline bool gl_transform_eq(struct gl_transform a, struct gl_transform b)
 
 void gl_transform_trans(struct gl_transform t, struct gl_transform *x);
 
-struct fbodst {
-    struct ra_tex *tex;
-    bool flip; // mirror vertically
-};
-
-void gl_transform_ortho_fbodst(struct gl_transform *t, struct fbodst fbo);
+void gl_transform_ortho_fbo(struct gl_transform *t, struct ra_fbo fbo);
 
 // A pool of buffers, which can grow as needed
 struct ra_buf_pool {
@@ -92,15 +88,7 @@ bool ra_tex_upload_pbo(struct ra *ra, struct ra_buf_pool *pbo,
 struct ra_layout std140_layout(struct ra_renderpass_input *inp);
 struct ra_layout std430_layout(struct ra_renderpass_input *inp);
 
-struct fbotex {
-    struct ra *ra;
-    struct ra_tex *tex;
-    int lw, lh; // logical (configured) size, <= than texture size
-    struct fbodst fbo;
-};
-
-void fbotex_uninit(struct fbotex *fbo);
-bool fbotex_change(struct fbotex *fbo, struct ra *ra, struct mp_log *log,
+bool ra_tex_resize(struct ra *ra, struct mp_log *log, struct ra_tex **tex,
                    int w, int h, const struct ra_format *fmt);
 
 // A wrapper around ra_timer that does result pooling, averaging etc.
diff --git a/video/out/gpu/video.c b/video/out/gpu/video.c
index 9884b4b2e5..286a4aa9ee 100644
--- a/video/out/gpu/video.c
+++ b/video/out/gpu/video.c
@@ -115,7 +115,7 @@ static const char *plane_names[] = {
 
 // A self-contained description of a source image which can be bound to a
 // texture unit and sampled from. Contains metadata about how it's to be used
-struct img_tex {
+struct image {
     enum plane_type type; // must be set to something non-zero
     int components; // number of relevant coordinates
     float multiplier; // multiplier to be used when sampling
@@ -124,10 +124,10 @@ struct img_tex {
     struct gl_transform transform; // rendering transformation
 };
 
-// A named img_tex, for user scripting purposes
-struct saved_tex {
+// A named image, for user scripting purposes
+struct saved_img {
     const char *name;
-    struct img_tex tex;
+    struct image img;
 };
 
 // A texture hook. This is some operation that transforms a named texture as
@@ -138,18 +138,18 @@ struct tex_hook {
     const char *bind_tex[TEXUNIT_VIDEO_NUM];
     int components; // how many components are relevant (0 = same as input)
     void *priv; // this gets talloc_freed when the tex_hook is removed
-    void (*hook)(struct gl_video *p, struct img_tex tex, // generates GLSL
+    void (*hook)(struct gl_video *p, struct image img, // generates GLSL
                  struct gl_transform *trans, void *priv);
-    bool (*cond)(struct gl_video *p, struct img_tex tex, void *priv);
+    bool (*cond)(struct gl_video *p, struct image img, void *priv);
 };
 
-struct fbosurface {
-    struct fbotex fbotex;
+struct surface {
+    struct ra_tex *tex;
     uint64_t id;
     double pts;
 };
 
-#define FBOSURFACES_MAX 10
+#define SURFACES_MAX 10
 
 struct cached_file {
     char *path;
@@ -216,16 +216,16 @@ struct gl_video {
     bool forced_dumb_mode;
 
     const struct ra_format *fbo_format;
-    struct fbotex merge_fbo[4];
-    struct fbotex scale_fbo[4];
-    struct fbotex integer_fbo[4];
-    struct fbotex indirect_fbo;
-    struct fbotex blend_subs_fbo;
-    struct fbotex screen_fbo;
-    struct fbotex output_fbo;
-    struct fbosurface surfaces[FBOSURFACES_MAX];
-    struct fbotex vdpau_deinterleave_fbo[2];
+    struct ra_tex *merge_tex[4];
+    struct ra_tex *scale_tex[4];
+    struct ra_tex *integer_tex[4];
+    struct ra_tex *indirect_tex;
+    struct ra_tex *blend_subs_tex;
+    struct ra_tex *screen_tex;
+    struct ra_tex *output_tex;
+    struct ra_tex *vdpau_deinterleave_tex[2];
     struct ra_buf *hdr_peak_ssbo;
+    struct surface surfaces[SURFACES_MAX];
 
     // user pass descriptions and textures
     struct tex_hook tex_hooks[SHADER_MAX_PASSES];
@@ -237,7 +237,7 @@ struct gl_video {
     int surface_now;
     int frames_drawn;
     bool is_interpolated;
-    bool output_fbo_valid;
+    bool output_tex_valid;
 
     // state for configured scalers
     struct scaler scaler[SCALER_COUNT];
@@ -249,9 +249,9 @@ struct gl_video {
     struct mp_osd_res osd_rect; // OSD size/margins
 
     // temporary during rendering
-    struct img_tex pass_tex[TEXUNIT_VIDEO_NUM];
+    struct image pass_img[TEXUNIT_VIDEO_NUM];
     struct compute_info pass_compute; // compute shader metadata for this pass
-    int pass_tex_num;
+    int pass_img_num;
     int texture_w, texture_h;
     struct gl_transform texture_offset; // texture transform without rotation
     int components;
@@ -268,9 +268,9 @@ struct gl_video {
     struct timer_pool *osd_timer;
 
     // intermediate textures
-    struct saved_tex saved_tex[SHADER_MAX_SAVED];
-    int saved_tex_num;
-    struct fbotex hook_fbos[SHADER_MAX_SAVED];
+    struct saved_img saved_img[SHADER_MAX_SAVED];
+    int saved_img_num;
+    struct ra_tex *hook_fbos[SHADER_MAX_SAVED];
     int hook_fbo_num;
 
     int frames_uploaded;
@@ -466,14 +466,14 @@ static void debug_check_gl(struct gl_video *p, const char *msg)
 
 static void gl_video_reset_surfaces(struct gl_video *p)
 {
-    for (int i = 0; i < FBOSURFACES_MAX; i++) {
+    for (int i = 0; i < SURFACES_MAX; i++) {
         p->surfaces[i].id = 0;
         p->surfaces[i].pts = MP_NOPTS_VALUE;
     }
     p->surface_idx = 0;
     p->surface_now = 0;
     p->frames_drawn = 0;
-    p->output_fbo_valid = false;
+    p->output_tex_valid = false;
 }
 
 static void gl_video_reset_hooks(struct gl_video *p)
@@ -488,10 +488,10 @@ static void gl_video_reset_hooks(struct gl_video *p)
     p->user_tex_num = 0;
 }
 
-static inline int fbosurface_wrap(int id)
+static inline int surface_wrap(int id)
 {
-    id = id % FBOSURFACES_MAX;
-    return id < 0 ? id + FBOSURFACES_MAX : id;
+    id = id % SURFACES_MAX;
+    return id < 0 ? id + SURFACES_MAX : id;
 }
 
 static void reinit_osd(struct gl_video *p)
@@ -510,24 +510,24 @@ static void uninit_rendering(struct gl_video *p)
     ra_tex_free(p->ra, &p->dither_texture);
 
     for (int n = 0; n < 4; n++) {
-        fbotex_uninit(&p->merge_fbo[n]);
-        fbotex_uninit(&p->scale_fbo[n]);
-        fbotex_uninit(&p->integer_fbo[n]);
+        ra_tex_free(p->ra, &p->merge_tex[n]);
+        ra_tex_free(p->ra, &p->scale_tex[n]);
+        ra_tex_free(p->ra, &p->integer_tex[n]);
     }
 
-    fbotex_uninit(&p->indirect_fbo);
-    fbotex_uninit(&p->blend_subs_fbo);
-    fbotex_uninit(&p->screen_fbo);
-    fbotex_uninit(&p->output_fbo);
+    ra_tex_free(p->ra, &p->indirect_tex);
+    ra_tex_free(p->ra, &p->blend_subs_tex);
+    ra_tex_free(p->ra, &p->screen_tex);
+    ra_tex_free(p->ra, &p->output_tex);
 
-    for (int n = 0; n < FBOSURFACES_MAX; n++)
-        fbotex_uninit(&p->surfaces[n].fbotex);
+    for (int n = 0; n < SURFACES_MAX; n++)
+        ra_tex_free(p->ra, &p->surfaces[n].tex);
 
     for (int n = 0; n < SHADER_MAX_SAVED; n++)
-        fbotex_uninit(&p->hook_fbos[n]);
+        ra_tex_free(p->ra, &p->hook_fbos[n]);
 
     for (int n = 0; n < 2; n++)
-        fbotex_uninit(&p->vdpau_deinterleave_fbo[n]);
+        ra_tex_free(p->ra, &p->vdpau_deinterleave_tex[n]);
 
     gl_video_reset_surfaces(p);
     gl_video_reset_hooks(p);
@@ -613,29 +613,29 @@ static bool gl_video_get_lut3d(struct gl_video *p, enum mp_csp_prim prim,
     return true;
 }
 
-// Fill an img_tex struct from an FBO + some metadata
-static struct img_tex img_tex_fbo(struct fbotex *fbo, enum plane_type type,
-                                  int components)
+// Fill an image struct from a ra_tex + some metadata
+static struct image image_wrap(struct ra_tex *tex, enum plane_type type,
+                               int components)
 {
     assert(type != PLANE_NONE);
-    return (struct img_tex){
+    return (struct image){
         .type = type,
-        .tex = fbo->tex,
+        .tex = tex,
         .multiplier = 1.0,
-        .w = fbo->lw,
-        .h = fbo->lh,
+        .w = tex->params.w,
+        .h = tex->params.h,
         .transform = identity_trans,
         .components = components,
     };
 }
 
-// Bind an img_tex to a free texture unit and return its ID. At most
+// Bind an image to a free texture unit and return its ID. At most
 // TEXUNIT_VIDEO_NUM texture units can be bound at once
-static int pass_bind(struct gl_video *p, struct img_tex tex)
+static int pass_bind(struct gl_video *p, struct image img)
 {
-    assert(p->pass_tex_num < TEXUNIT_VIDEO_NUM);
-    p->pass_tex[p->pass_tex_num] = tex;
-    return p->pass_tex_num++;
+    assert(p->pass_img_num < TEXUNIT_VIDEO_NUM);
+    p->pass_img[p->pass_img_num] = img;
+    return p->pass_img_num++;
 }
 
 // Rotation by 90° and flipping.
@@ -684,11 +684,11 @@ static enum plane_type merge_plane_types(enum plane_type a, enum plane_type b)
     return a;
 }
 
-// Places a video_image's image textures + associated metadata into tex[]. The
+// Places a video_image's image textures + associated metadata into img[]. The
 // number of textures is equal to p->plane_count. Any necessary plane offsets
 // are stored in off. (e.g. chroma position)
-static void pass_get_img_tex(struct gl_video *p, struct video_image *vimg,
-                             struct img_tex tex[4], struct gl_transform off[4])
+static void pass_get_images(struct gl_video *p, struct video_image *vimg,
+                            struct image img[4], struct gl_transform off[4])
 {
     assert(vimg->mpi);
 
@@ -721,7 +721,7 @@ static void pass_get_img_tex(struct gl_video *p, struct video_image *vimg,
                                          msb_valid_bits,
                                          p->ra_format.component_bits);
 
-    memset(tex, 0, 4 * sizeof(tex[0]));
+    memset(img, 0, 4 * sizeof(img[0]));
     for (int n = 0; n < p->plane_count; n++) {
         struct texplane *t = &vimg->planes[n];
 
@@ -743,7 +743,7 @@ static void pass_get_img_tex(struct gl_video *p, struct video_image *vimg,
             type = merge_plane_types(type, ctype);
         }
 
-        tex[n] = (struct img_tex){
+        img[n] = (struct image){
             .type = type,
             .tex = t->tex,
             .multiplier = tex_mul,
@@ -752,12 +752,12 @@ static void pass_get_img_tex(struct gl_video *p, struct video_image *vimg,
         };
 
         for (int i = 0; i < 4; i++)
-            tex[n].components += !!p->ra_format.components[n][i];
+            img[n].components += !!p->ra_format.components[n][i];
 
         get_transform(t->w, t->h, p->image_params.rotate, t->flipped,
-                      &tex[n].transform);
+                      &img[n].transform);
         if (p->image_params.rotate % 180 == 90)
-            MPSWAP(int, tex[n].w, tex[n].h);
+            MPSWAP(int, img[n].w, img[n].h);
 
         off[n] = identity_trans;
 
@@ -1060,8 +1060,8 @@ static void pass_prepare_src_tex(struct gl_video *p)
 {
     struct gl_shader_cache *sc = p->sc;
 
-    for (int n = 0; n < p->pass_tex_num; n++) {
-        struct img_tex *s = &p->pass_tex[n];
+    for (int n = 0; n < p->pass_img_num; n++) {
+        struct image *s = &p->pass_img[n];
         if (!s->tex)
             continue;
 
@@ -1116,7 +1116,7 @@ static void dispatch_compute(struct gl_video *p, int w, int h,
     PRELUDE("#define outcoord(id) (out_scale * (vec2(id) + vec2(0.5)))\n");
 
     for (int n = 0; n < TEXUNIT_VIDEO_NUM; n++) {
-        struct img_tex *s = &p->pass_tex[n];
+        struct image *s = &p->pass_img[n];
         if (!s->tex)
             continue;
 
@@ -1144,18 +1144,18 @@ static void dispatch_compute(struct gl_video *p, int w, int h,
 
     pass_record(p, gl_sc_dispatch_compute(p->sc, num_x, num_y, 1));
 
-    memset(&p->pass_tex, 0, sizeof(p->pass_tex));
-    p->pass_tex_num = 0;
+    memset(&p->pass_img, 0, sizeof(p->pass_img));
+    p->pass_img_num = 0;
 }
 
 static struct mp_pass_perf render_pass_quad(struct gl_video *p,
-                                            struct fbodst target,
+                                            struct ra_fbo fbo,
                                             const struct mp_rect *dst)
 {
     struct vertex va[6] = {0};
 
     struct gl_transform t;
-    gl_transform_ortho_fbodst(&t, target);
+    gl_transform_ortho_fbo(&t, fbo);
 
     float x[2] = {dst->x0, dst->x1};
     float y[2] = {dst->y0, dst->y1};
@@ -1166,8 +1166,8 @@ static struct mp_pass_perf render_pass_quad(struct gl_video *p,
         struct vertex *v = &va[n];
         v->position.x = x[n / 2];
         v->position.y = y[n % 2];
-        for (int i = 0; i < p->pass_tex_num; i++) {
-            struct img_tex *s = &p->pass_tex[i];
+        for (int i = 0; i < p->pass_img_num; i++) {
+            struct image *s = &p->pass_img[i];
             if (!s->tex)
                 continue;
             struct gl_transform tr = s->transform;
@@ -1183,33 +1183,32 @@ static struct mp_pass_perf render_pass_quad(struct gl_video *p,
     va[4] = va[2];
     va[5] = va[1];
 
-    return gl_sc_dispatch_draw(p->sc, target.tex, va, 6);
+    return gl_sc_dispatch_draw(p->sc, fbo.tex, va, 6);
 }
 
-static void finish_pass_direct(struct gl_video *p, struct fbodst target,
+static void finish_pass_fbo(struct gl_video *p, struct ra_fbo fbo,
                                const struct mp_rect *dst)
 {
     pass_prepare_src_tex(p);
     gl_sc_set_vertex_format(p->sc, vertex_vao, sizeof(struct vertex));
-    pass_record(p, render_pass_quad(p, target, dst));
+    pass_record(p, render_pass_quad(p, fbo, dst));
     debug_check_gl(p, "after rendering");
-    memset(&p->pass_tex, 0, sizeof(p->pass_tex));
-    p->pass_tex_num = 0;
+    memset(&p->pass_img, 0, sizeof(p->pass_img));
+    p->pass_img_num = 0;
 }
 
 // dst_fbo: this will be used for rendering; possibly reallocating the whole
 //          FBO, if the required parameters have changed
 // w, h: required FBO target dimension, and also defines the target rectangle
 //       used for rasterization
-static void finish_pass_fbo(struct gl_video *p, struct fbotex *dst_fbo,
+static void finish_pass_tex(struct gl_video *p, struct ra_tex **dst_tex,
                             int w, int h)
 {
-    fbotex_change(dst_fbo, p->ra, p->log, w, h, p->fbo_format);
+    if (!ra_tex_resize(p->ra, p->log, dst_tex, w, h, p->fbo_format))
+        return;
 
     if (p->pass_compute.active) {
-        if (!dst_fbo->tex)
-            return;
-        gl_sc_uniform_image2D_wo(p->sc, "out_image", dst_fbo->tex);
+        gl_sc_uniform_image2D_wo(p->sc, "out_image", *dst_tex);
         if (!p->pass_compute.directly_writes)
             GLSL(imageStore(out_image, ivec2(gl_GlobalInvocationID), color);)
 
@@ -1218,11 +1217,12 @@ static void finish_pass_fbo(struct gl_video *p, struct fbotex *dst_fbo,
 
         debug_check_gl(p, "after dispatching compute shader");
     } else {
-        finish_pass_direct(p, dst_fbo->fbo, &(struct mp_rect){0, 0, w, h});
+        struct ra_fbo fbo = { .tex = *dst_tex, };
+        finish_pass_fbo(p, fbo, &(struct mp_rect){0, 0, w, h});
     }
 }
 
-static const char *get_tex_swizzle(struct img_tex *img)
+static const char *get_tex_swizzle(struct image *img)
 {
     if (!img->tex)
         return "rgba";
@@ -1231,7 +1231,7 @@ static const char *get_tex_swizzle(struct img_tex *img)
 
 // Copy a texture to the vec4 color, while increasing offset. Also applies
 // the texture multiplier to the sampled color
-static void copy_img_tex(struct gl_video *p, int *offset, struct img_tex img)
+static void copy_image(struct gl_video *p, int *offset, struct image img)
 {
     int count = img.components;
     assert(*offset + count <= 4);
@@ -1265,14 +1265,14 @@ static void skip_unused(struct gl_video *p, int num_components)
 
 static void uninit_scaler(struct gl_video *p, struct scaler *scaler)
 {
-    fbotex_uninit(&scaler->sep_fbo);
+    ra_tex_free(p->ra, &scaler->sep_fbo);
     ra_tex_free(p->ra, &scaler->lut);
     scaler->kernel = NULL;
     scaler->initialized = false;
 }
 
 static void hook_prelude(struct gl_video *p, const char *name, int id,
-                         struct img_tex tex)
+                         struct image img)
 {
     GLSLHF("#define %s_raw texture%d\n", name, id);
     GLSLHF("#define %s_pos texcoord%d\n", name, id);
@@ -1280,15 +1280,15 @@ static void hook_prelude(struct gl_video *p, const char *name, int id,
     GLSLHF("#define %s_rot texture_rot%d\n", name, id);
     GLSLHF("#define %s_pt pixel_size%d\n", name, id);
     GLSLHF("#define %s_map texmap%d\n", name, id);
-    GLSLHF("#define %s_mul %f\n", name, tex.multiplier);
+    GLSLHF("#define %s_mul %f\n", name, img.multiplier);
 
     // Set up the sampling functions
     GLSLHF("#define %s_tex(pos) (%s_mul * vec4(texture(%s_raw, pos)).%s)\n",
-           name, name, name, get_tex_swizzle(&tex));
+           name, name, name, get_tex_swizzle(&img));
 
     // Since the extra matrix multiplication impacts performance,
     // skip it unless the texture was actually rotated
-    if (gl_transform_eq(tex.transform, identity_trans)) {
+    if (gl_transform_eq(img.transform, identity_trans)) {
         GLSLHF("#define %s_texOff(off) %s_tex(%s_pos + %s_pt * vec2(off))\n",
                name, name, name, name);
     } else {
@@ -1298,15 +1298,15 @@ static void hook_prelude(struct gl_video *p, const char *name, int id,
     }
 }
 
-static bool saved_tex_find(struct gl_video *p, const char *name,
-                           struct img_tex *out)
+static bool saved_img_find(struct gl_video *p, const char *name,
+                           struct image *out)
 {
     if (!name || !out)
         return false;
 
-    for (int i = 0; i < p->saved_tex_num; i++) {
-        if (strcmp(p->saved_tex[i].name, name) == 0) {
-            *out = p->saved_tex[i].tex;
+    for (int i = 0; i < p->saved_img_num; i++) {
+        if (strcmp(p->saved_img[i].name, name) == 0) {
+            *out = p->saved_img[i].img;
             return true;
         }
     }
@@ -1314,27 +1314,27 @@ static bool saved_tex_find(struct gl_video *p, const char *name,
     return false;
 }
 
-static void saved_tex_store(struct gl_video *p, const char *name,
-                            struct img_tex tex)
+static void saved_img_store(struct gl_video *p, const char *name,
+                            struct image img)
 {
     assert(name);
 
-    for (int i = 0; i < p->saved_tex_num; i++) {
-        if (strcmp(p->saved_tex[i].name, name) == 0) {
-            p->saved_tex[i].tex = tex;
+    for (int i = 0; i < p->saved_img_num; i++) {
+        if (strcmp(p->saved_img[i].name, name) == 0) {
+            p->saved_img[i].img = img;
             return;
         }
     }
 
-    assert(p->saved_tex_num < SHADER_MAX_SAVED);
-    p->saved_tex[p->saved_tex_num++] = (struct saved_tex) {
+    assert(p->saved_img_num < SHADER_MAX_SAVED);
+    p->saved_img[p->saved_img_num++] = (struct saved_img) {
         .name = name,
-        .tex = tex
+        .img = img
     };
 }
 
 static bool pass_hook_setup_binds(struct gl_video *p, const char *name,
-                                  struct img_tex tex, struct tex_hook *hook)
+                                  struct image img, struct tex_hook *hook)
 {
     for (int t = 0; t < TEXUNIT_VIDEO_NUM; t++) {
         char *bind_name = (char *)hook->bind_tex[t];
@@ -1344,9 +1344,9 @@ static bool pass_hook_setup_binds(struct gl_video *p, const char *name,
 
         // This is a special name that means "currently hooked texture"
         if (strcmp(bind_name, "HOOKED") == 0) {
-            int id = pass_bind(p, tex);
-            hook_prelude(p, "HOOKED", id, tex);
-            hook_prelude(p, name, id, tex);
+            int id = pass_bind(p, img);
+            hook_prelude(p, "HOOKED", id, img);
+            hook_prelude(p, name, id, img);
             continue;
         }
 
@@ -1361,16 +1361,16 @@ static bool pass_hook_setup_binds(struct gl_video *p, const char *name,
             }
         }
 
-        struct img_tex bind_tex;
-        if (!saved_tex_find(p, bind_name, &bind_tex)) {
+        struct image bind_img;
+        if (!saved_img_find(p, bind_name, &bind_img)) {
             // Clean up texture bindings and move on to the next hook
             MP_DBG(p, "Skipping hook on %s due to no texture named %s.\n",
                    name, bind_name);
-            p->pass_tex_num -= t;
+            p->pass_img_num -= t;
             return false;
         }
 
-        hook_prelude(p, bind_name, pass_bind(p, bind_tex), bind_tex);
+        hook_prelude(p, bind_name, pass_bind(p, bind_img), bind_img);
 
 next_bind: ;
     }
@@ -1378,15 +1378,15 @@ next_bind: ;
     return true;
 }
 
-// Process hooks for a plane, saving the result and returning a new img_tex
-// If 'trans' is NULL, the shader is forbidden from transforming tex
-static struct img_tex pass_hook(struct gl_video *p, const char *name,
-                                struct img_tex tex, struct gl_transform *trans)
+// Process hooks for a plane, saving the result and returning a new image
+// If 'trans' is NULL, the shader is forbidden from transforming img
+static struct image pass_hook(struct gl_video *p, const char *name,
+                                struct image img, struct gl_transform *trans)
 {
     if (!name)
-        return tex;
+        return img;
 
-    saved_tex_store(p, name, tex);
+    saved_img_store(p, name, img);
 
     MP_DBG(p, "Running hooks for %s\n", name);
     for (int i = 0; i < p->tex_hook_num; i++) {
@@ -1402,34 +1402,34 @@ static struct img_tex pass_hook(struct gl_video *p, const char *name,
 
 found:
         // Check the hook's condition
-        if (hook->cond && !hook->cond(p, tex, hook->priv)) {
+        if (hook->cond && !hook->cond(p, img, hook->priv)) {
             MP_DBG(p, "Skipping hook on %s due to condition.\n", name);
             continue;
         }
 
-        if (!pass_hook_setup_binds(p, name, tex, hook))
+        if (!pass_hook_setup_binds(p, name, img, hook))
             continue;
 
         // Run the actual hook. This generates a series of GLSL shader
         // instructions sufficient for drawing the hook's output
         struct gl_transform hook_off = identity_trans;
-        hook->hook(p, tex, &hook_off, hook->priv);
+        hook->hook(p, img, &hook_off, hook->priv);
 
-        int comps = hook->components ? hook->components : tex.components;
+        int comps = hook->components ? hook->components : img.components;
         skip_unused(p, comps);
 
         // Compute the updated FBO dimensions and store the result
-        struct mp_rect_f sz = {0, 0, tex.w, tex.h};
+        struct mp_rect_f sz = {0, 0, img.w, img.h};
         gl_transform_rect(hook_off, &sz);
         int w = lroundf(fabs(sz.x1 - sz.x0));
         int h = lroundf(fabs(sz.y1 - sz.y0));
 
         assert(p->hook_fbo_num < SHADER_MAX_SAVED);
-        struct fbotex *fbo = &p->hook_fbos[p->hook_fbo_num++];
-        finish_pass_fbo(p, fbo, w, h);
+        struct ra_tex **fbo = &p->hook_fbos[p->hook_fbo_num++];
+        finish_pass_tex(p, fbo, w, h);
 
         const char *store_name = hook->save_tex ? hook->save_tex : name;
-        struct img_tex saved_tex = img_tex_fbo(fbo, tex.type, comps);
+        struct image saved_img = image_wrap(*fbo, img.type, comps);
 
         // If the texture we're saving overwrites the "current" texture, also
         // update the tex parameter so that the future loop cycles will use the
@@ -1438,18 +1438,18 @@ found:
             if (!trans && !gl_transform_eq(hook_off, identity_trans)) {
                 MP_ERR(p, "Hook tried changing size of unscalable texture %s!\n",
                        name);
-                return tex;
+                return img;
             }
 
-            tex = saved_tex;
+            img = saved_img;
             if (trans)
                 gl_transform_trans(hook_off, trans);
         }
 
-        saved_tex_store(p, store_name, saved_tex);
+        saved_img_store(p, store_name, saved_img);
     }
 
-    return tex;
+    return img;
 }
 
 // This can be used at any time in the middle of rendering to specify an
@@ -1482,12 +1482,12 @@ static void pass_opt_hook_point(struct gl_video *p, const char *name,
 
 found:
     assert(p->hook_fbo_num < SHADER_MAX_SAVED);
-    struct fbotex *fbo = &p->hook_fbos[p->hook_fbo_num++];
-    finish_pass_fbo(p, fbo, p->texture_w, p->texture_h);
+    struct ra_tex **tex = &p->hook_fbos[p->hook_fbo_num++];
+    finish_pass_tex(p, tex, p->texture_w, p->texture_h);
 
-    struct img_tex img = img_tex_fbo(fbo, PLANE_RGB, p->components);
+    struct image img = image_wrap(*tex, PLANE_RGB, p->components);
     img = pass_hook(p, name, img, tex_trans);
-    copy_img_tex(p, &(int){0}, img);
+    copy_image(p, &(int){0}, img);
     p->texture_w = img.w;
     p->texture_h = img.h;
     p->components = img.components;
@@ -1635,7 +1635,7 @@ static void reinit_scaler(struct gl_video *p, struct scaler *scaler,
 }
 
 // Special helper for sampling from two separated stages
-static void pass_sample_separated(struct gl_video *p, struct img_tex src,
+static void pass_sample_separated(struct gl_video *p, struct image src,
                                   struct scaler *scaler, int w, int h)
 {
     // Separate the transformation into x and y components, per pass
@@ -1654,10 +1654,10 @@ static void pass_sample_separated(struct gl_video *p, struct img_tex src,
     GLSLF("// first pass\n");
     pass_sample_separated_gen(p->sc, scaler, 0, 1);
     GLSLF("color *= %f;\n", src.multiplier);
-    finish_pass_fbo(p, &scaler->sep_fbo, src.w, h);
+    finish_pass_tex(p, &scaler->sep_fbo, src.w, h);
 
     // Second pass (scale only in the x dir)
-    src = img_tex_fbo(&scaler->sep_fbo, src.type, src.components);
+    src = image_wrap(scaler->sep_fbo, src.type, src.components);
     src.transform = t_x;
     pass_describe(p, "%s second pass", scaler->conf.kernel.name);
     sampler_prelude(p->sc, pass_bind(p, src));
@@ -1667,7 +1667,7 @@ static void pass_sample_separated(struct gl_video *p, struct img_tex src,
 // Picks either the compute shader version or the regular sampler version
 // depending on hardware support
 static void pass_dispatch_sample_polar(struct gl_video *p, struct scaler *scaler,
-