Add OriginalAspect and OriginalAspectRot uniforms (#17123)

* Initial implementation of CoreAspect uniform

* float -> float_t

* Possibly fix wii_u building

* vulkan: use float instead of float_t;

* slangp: Advertise support of CoreAspect uniform
by defining _RARCH_HAS_COREASPECT_UNIFORM early in the shader source, just after "#extension GL_GOOGLE_cpp_style_line_directive : require"

* CoreAspect + glsl fix: use glUniform1f()

* Add CoreAspectRot uniform.
It reports CoreAspect value or 1/CoreAspect when the content is rotated by 90 or 270 deg.

* Fixed stupid typo

* Just use _HAS_COREASPECT_UNIFORMS to check for CoreAspect uniforms support (was _RARCH_HAS_COREASPECT_UNIFORMS)

* Rename CoreAspect, CoreAspectRot, _HAS_COREASPECT_UNIFORMS to OriginalAspect, OriginlAspectRot, _HAS_ORIGINALASPECT_UNIFORMS

* GLCore: void Pass::build_semantic_float needs glUniform1f.
...how on earth did it worked for UBO !?

* d3d10,11,12, wrong function called by overlook.

* Add test shader, will remove that before PR

* Fix metal rotated aspect reporting

* remove test shader

* Fix C89 Build

* Use OriginalAspectRotated instead of OriginalAspectRot

gfx/common/d3d10_defines.h

@@ -218,6 +218,9 @@ typedef struct
       uint32_t                   rotation;
       uint32_t                   total_subframes;
       uint32_t                   current_subframe;
+      float                      core_aspect;
+      float                      core_aspect_rot;
+
    } pass[GFX_MAX_SHADERS];
 
    struct video_shader* shader_preset;

gfx/common/d3d11_defines.h

@@ -294,6 +294,8 @@ typedef struct
       uint32_t                   rotation;
       uint32_t                   total_subframes;
       uint32_t                   current_subframe;
+      float                      core_aspect;
+      float                      core_aspect_rot;
    } pass[GFX_MAX_SHADERS];
 
    struct video_shader* shader_preset;

gfx/common/d3d12_defines.h

@@ -331,6 +331,8 @@ typedef struct
       uint32_t                        rotation;
       uint32_t                        total_subframes;
       uint32_t                        current_subframe;
+      float                           core_aspect;
+      float                           core_aspect_rot;
       D3D12_GPU_DESCRIPTOR_HANDLE     textures;
       D3D12_GPU_DESCRIPTOR_HANDLE     samplers;
    } pass[GFX_MAX_SHADERS];

gfx/drivers/d3d10.c

@@ -1336,6 +1336,8 @@ static bool d3d10_gfx_set_shader(void* data,
             &d3d10->pass[i].frame_count,     /* FrameCount */
             &d3d10->pass[i].frame_direction, /* FrameDirection */
             &d3d10->pass[i].rotation,        /* Rotation */
+            &d3d10->pass[i].core_aspect,     /* OriginalAspect */
+            &d3d10->pass[i].core_aspect_rot, /* OriginalAspectRotated */
             &d3d10->pass[i].total_subframes, /* TotalSubFrames */
             &d3d10->pass[i].current_subframe,/* CurrentSubFrame */
          }
@@ -2331,6 +2333,14 @@ static bool d3d10_gfx_frame(
 
          d3d10->pass[i].rotation = retroarch_get_rotation();
 
+         d3d10->pass[i].core_aspect = video_driver_get_core_aspect();
+
+         /* OriginalAspectRotated: return 1/aspect for 90 and 270 rotated content */
+         d3d10->pass[i].core_aspect_rot = video_driver_get_core_aspect();
+         uint32_t rot = retroarch_get_rotation();
+         if (rot == 1 || rot == 3)
+            d3d10->pass[i].core_aspect_rot = 1/d3d10->pass[i].core_aspect_rot;
+
          /* Sub-frame info for multiframe shaders (per real content frame).
             Should always be 1 for non-use of subframes */
          if (!(d3d10->flags & D3D10_ST_FLAG_FRAME_DUPE_LOCK))

gfx/drivers/d3d11.c

@@ -1538,6 +1538,8 @@ static bool d3d11_gfx_set_shader(void* data, enum rarch_shader_type type, const
             &d3d11->pass[i].frame_count,     /* FrameCount */
             &d3d11->pass[i].frame_direction, /* FrameDirection */
             &d3d11->pass[i].rotation,        /* Rotation */
+            &d3d11->pass[i].core_aspect,     /* OriginalAspect */
+            &d3d11->pass[i].core_aspect_rot, /* OriginalAspectRotated */
             &d3d11->pass[i].total_subframes, /* TotalSubFrames */
             &d3d11->pass[i].current_subframe,/* CurrentSubFrame */
          }
@@ -3094,6 +3096,14 @@ static bool d3d11_gfx_frame(
 
          d3d11->pass[i].rotation = retroarch_get_rotation();
 
+         d3d11->pass[i].core_aspect = video_driver_get_core_aspect();
+
+         /* OriginalAspectRotated: return 1/aspect for 90 and 270 rotated content */
+         d3d11->pass[i].core_aspect_rot = video_driver_get_core_aspect();
+         uint32_t rot = retroarch_get_rotation();
+         if (rot == 1 || rot == 3)
+            d3d11->pass[i].core_aspect_rot = 1/d3d11->pass[i].core_aspect_rot;
+
          /* Sub-frame info for multiframe shaders (per real content frame).
             Should always be 1 for non-use of subframes */
          if (!(d3d11->flags & D3D11_ST_FLAG_FRAME_DUPE_LOCK))

gfx/drivers/d3d12.c

@@ -1725,6 +1725,8 @@ static bool d3d12_gfx_set_shader(void* data, enum rarch_shader_type type, const
             &d3d12->pass[i].frame_count,     /* FrameCount */
             &d3d12->pass[i].frame_direction, /* FrameDirection */
             &d3d12->pass[i].rotation,        /* Rotation */
+            &d3d12->pass[i].core_aspect,     /* OriginalAspect */
+            &d3d12->pass[i].core_aspect_rot, /* OriginalAspectRotated */
             &d3d12->pass[i].total_subframes, /* TotalSubFrames */
             &d3d12->pass[i].current_subframe,/* CurrentSubFrame */
          }
@@ -3595,6 +3597,14 @@ static bool d3d12_gfx_frame(
 
          d3d12->pass[i].rotation = retroarch_get_rotation();
 
+         d3d12->pass[i].core_aspect = video_driver_get_core_aspect();
+
+         /* OriginalAspectRotated: return 1/aspect for 90 and 270 rotated content */
+         d3d12->pass[i].core_aspect_rot = video_driver_get_core_aspect();
+         uint32_t rot = retroarch_get_rotation();
+         if (rot == 1 || rot == 3)
+            d3d12->pass[i].core_aspect_rot = 1/d3d12->pass[i].core_aspect_rot;
+
          /* Sub-frame info for multiframe shaders (per real content frame).
             Should always be 1 for non-use of subframes */
          if (!(d3d12->flags & D3D12_ST_FLAG_FRAME_DUPE_LOCK))

gfx/drivers/gl3.c

@@ -2744,6 +2744,15 @@ static bool gl3_frame(void *data, const void *frame,
 #endif
    gl3_filter_chain_set_rotation(gl->filter_chain, retroarch_get_rotation());
 
+   gl3_filter_chain_set_core_aspect(gl->filter_chain, video_driver_get_core_aspect());
+
+   /* OriginalAspectRotated: return 1/aspect for 90 and 270 rotated content */
+   uint32_t rot = retroarch_get_rotation();
+   float core_aspect_rot = video_driver_get_core_aspect();
+   if (rot == 1 || rot == 3)
+      core_aspect_rot = 1/core_aspect_rot;
+   gl3_filter_chain_set_core_aspect_rot(gl->filter_chain, core_aspect_rot);
+
    /* Sub-frame info for multiframe shaders (per real content frame).
       Should always be 1 for non-use of subframes*/
    if (!(gl->flags & GL3_FLAG_FRAME_DUPE_LOCK))

gfx/drivers/gx2_gfx.c

@@ -1665,7 +1665,7 @@ static bool wiiu_init_frame_textures(wiiu_video_t *wiiu, unsigned width, unsigne
 static void gx2_update_uniform_block(wiiu_video_t *wiiu,
       int pass, float *ubo, int id,
       int size, int uniformVarCount, GX2UniformVar *uniformVars,
-      uint64_t frame_count, int32_t frame_direction, uint32_t rotation)
+      uint64_t frame_count, int32_t frame_direction, uint32_t rotation, float core_aspect, float core_aspect_rot)
 {
    unsigned i;
    for (i = 0; i < uniformVarCount; i++)
@@ -1723,6 +1723,18 @@ static void gx2_update_uniform_block(wiiu_video_t *wiiu,
          continue;
       }
 
+      if (string_is_equal(id, "OriginalAspect"))
+      {
+         *dst        = core_aspect;
+         continue;
+      }
+
+      if (string_is_equal(id, "OriginalAspectRotated"))
+      {
+         *dst        = core_aspect_rot;
+         continue;
+      }
+
       if (string_is_equal(id, "OriginalSize"))
       {
          ((GX2_vec4 *)dst)->x = wiiu->texture.surface.width;
@@ -1940,6 +1952,13 @@ static bool gx2_frame(void *data, const void *frame,
 #endif
       uint32_t rotation       = retroarch_get_rotation();
 
+      float core_aspect = video_driver_get_core_aspect();
+
+      /* OriginalAspectRotated: return 1/aspect for 90 and 270 rotated content */
+      float core_aspect_rot = core_aspect;
+      if (rotation == 1 || rotation == 3)
+         core_aspect_rot = 1/core_aspect;
+
       for (i = 0; i < wiiu->shader_preset->passes; i++)
       {
          unsigned j;
@@ -1950,7 +1969,7 @@ static bool gx2_frame(void *data, const void *frame,
             gx2_update_uniform_block(wiiu, i, wiiu->pass[i].vs_ubos[j], j,
                   wiiu->pass[i].gfd->vs->uniformBlocks[j].size,
                   wiiu->pass[i].gfd->vs->uniformVarCount, wiiu->pass[i].gfd->vs->uniformVars,
-                  frame_count, frame_direction, rotation);
+                  frame_count, frame_direction, rotation, core_aspect, core_aspect_rot);
             GX2SetVertexUniformBlock(wiiu->pass[i].gfd->vs->uniformBlocks[j].offset,
                   wiiu->pass[i].gfd->vs->uniformBlocks[j].size, wiiu->pass[i].vs_ubos[j]);
          }
@@ -1962,7 +1981,7 @@ static bool gx2_frame(void *data, const void *frame,
             gx2_update_uniform_block(wiiu, i, wiiu->pass[i].ps_ubos[j], j,
                   wiiu->pass[i].gfd->ps->uniformBlocks[j].size,
                   wiiu->pass[i].gfd->ps->uniformVarCount, wiiu->pass[i].gfd->ps->uniformVars,
-                  frame_count, frame_direction, rotation);
+                  frame_count, frame_direction, rotation, core_aspect, core_aspect_rot);
             GX2SetPixelUniformBlock(wiiu->pass[i].gfd->ps->uniformBlocks[j].offset,
                   wiiu->pass[i].gfd->ps->uniformBlocks[j].size, wiiu->pass[i].ps_ubos[j]);
          }

gfx/drivers/metal.m

@@ -1206,6 +1206,8 @@ typedef struct MTLALIGN(16)
       uint32_t frame_count;
       int32_t frame_direction;
       uint32_t rotation;
+      float_t core_aspect;
+      float_t core_aspect_rot;
       pass_semantics_t semantics;
       MTLViewport viewport;
       __unsafe_unretained id<MTLRenderPipelineState> _state;
@@ -1579,6 +1581,15 @@ typedef struct MTLALIGN(16)
 
       _engine.pass[i].rotation = retroarch_get_rotation();
 
+      _engine.pass[i].core_aspect = video_driver_get_core_aspect();
+
+      /* OriginalAspectRotated: return 1/aspect for 90 and 270 rotated content */
+      int rot = retroarch_get_rotation();
+      float core_aspect_rot = video_driver_get_core_aspect();
+      if (rot == 1 || rot == 3)
+         core_aspect_rot = 1/core_aspect_rot;
+      _engine.pass[i].core_aspect_rot  = core_aspect_rot;
+
       for (j = 0; j < SLANG_CBUFFER_MAX; j++)
       {
          id<MTLBuffer> buffer      = _engine.pass[i].buffers[j];
@@ -1836,6 +1847,8 @@ typedef struct MTLALIGN(16)
                &_engine.pass[i].frame_count,     /* FrameCount */
                &_engine.pass[i].frame_direction, /* FrameDirection */
                &_engine.pass[i].rotation,        /* Rotation */
+               &_engine.pass[i].core_aspect,     /* OriginalAspect */
+               &_engine.pass[i].core_aspect_rot, /* OriginalAspectRotated */
             }
          };
          /* clang-format on */

gfx/drivers/vulkan.c

@@ -4574,6 +4574,17 @@ static bool vulkan_frame(void *data, const void *frame,
    vulkan_filter_chain_set_rotation(
          (vulkan_filter_chain_t*)vk->filter_chain, retroarch_get_rotation());
 
+   vulkan_filter_chain_set_core_aspect(
+         (vulkan_filter_chain_t*)vk->filter_chain, video_driver_get_core_aspect());
+
+   /* OriginalAspectRotated: return 1/aspect for 90 and 270 rotated content */
+   uint32_t rot = retroarch_get_rotation();
+   float core_aspect_rot = video_driver_get_core_aspect();
+   if (rot == 1 || rot == 3)
+      core_aspect_rot = 1/core_aspect_rot;
+   vulkan_filter_chain_set_core_aspect_rot(
+         (vulkan_filter_chain_t*)vk->filter_chain, core_aspect_rot);
+
    /* Render offscreen filter chain passes. */
    {
       /* Set the source texture in the filter chain */

gfx/drivers_shader/glslang_util.c

@@ -177,17 +177,21 @@ bool glslang_read_shader_file(const char *path,
          goto error;
    }
 
+   /* Add defines about supported retroarch features */
+   if (!string_list_append(output, "#define _HAS_ORIGINALASPECT_UNIFORMS", attr))
+      goto error;
+
    /* At least VIM treats the first line as line #1,
     * so offset everything by one. */
    snprintf(tmp, sizeof(tmp), "#line %u \"%s\"", root_file ? 2 : 1, basename);
    if (!string_list_append(output, tmp, attr))
       goto error;
-   
+
    /* Loop through lines of file */
    for (i = root_file ? 1 : 0; i < lines.size; i++)
    {
       const char *line   = lines.elems[i].data;
-   
+
       /* Check for 'include' statements */
       bool include_optional = !strncmp("#pragma include_optional ", line, STRLEN_CONST("#pragma include_optional "));
       if ( !strncmp("#include ", line, STRLEN_CONST("#include ")) || include_optional )

gfx/drivers_shader/shader_gl3.cpp

@@ -757,6 +757,16 @@ public:
       rotation = rot;
    }
 
+   void set_core_aspect(float coreaspect)
+   {
+      core_aspect = coreaspect;
+   }
+
+   void set_core_aspect_rot(float coreaspectrot)
+   {
+      core_aspect_rot = coreaspectrot;
+   }
+
    void set_shader_subframes(uint32_t tot_subframes)
    {
       total_subframes = tot_subframes;
@@ -855,6 +865,8 @@ private:
          slang_semantic semantic, uint32_t value);
    void build_semantic_int(uint8_t *data,
          slang_semantic semantic, int32_t value);
+   void build_semantic_float(uint8_t *data,
+         slang_semantic semantic, float value);
    void build_semantic_parameter(uint8_t *data, unsigned index, float value);
    void build_semantic_texture_vec4(uint8_t *data,
          slang_texture_semantic semantic,
@@ -872,6 +884,8 @@ private:
    unsigned frame_count_period = 0;
    int32_t frame_direction = 1;
    uint32_t rotation = 0;
+   float core_aspect = 0;
+   float core_aspect_rot = 0;
    unsigned pass_number = 0;
    uint32_t total_subframes = 1;
    uint32_t current_subframe = 1;
@@ -1080,6 +1094,8 @@ bool Pass::init_pipeline()
    reflect_parameter("FrameCount", reflection.semantics[SLANG_SEMANTIC_FRAME_COUNT]);
    reflect_parameter("FrameDirection", reflection.semantics[SLANG_SEMANTIC_FRAME_DIRECTION]);
    reflect_parameter("Rotation", reflection.semantics[SLANG_SEMANTIC_ROTATION]);
+   reflect_parameter("OriginalAspect", reflection.semantics[SLANG_SEMANTIC_CORE_ASPECT]);
+   reflect_parameter("OriginalAspectRotated", reflection.semantics[SLANG_SEMANTIC_CORE_ASPECT_ROT]);
    reflect_parameter("TotalSubFrames", reflection.semantics[SLANG_SEMANTIC_TOTAL_SUBFRAMES]);
    reflect_parameter("CurrentSubFrame", reflection.semantics[SLANG_SEMANTIC_CURRENT_SUBFRAME]);
 
@@ -1319,6 +1335,38 @@ void Pass::build_semantic_int(uint8_t *data, slang_semantic semantic,
    }
 }
 
+void Pass::build_semantic_float(uint8_t *data, slang_semantic semantic,
+                              float value)
+{
+   auto &refl = reflection.semantics[semantic];
+
+   if (data && refl.uniform)
+   {
+      if (refl.location.ubo_vertex >= 0 || refl.location.ubo_fragment >= 0)
+      {
+         if (refl.location.ubo_vertex >= 0)
+            glUniform1f(refl.location.ubo_vertex, value);
+         if (refl.location.ubo_fragment >= 0)
+            glUniform1f(refl.location.ubo_fragment, value);
+      }
+      else
+         *reinterpret_cast<float *>(data + reflection.semantics[semantic].ubo_offset) = value;
+   }
+
+   if (refl.push_constant)
+   {
+      if (refl.location.push_vertex >= 0 || refl.location.push_fragment >= 0)
+      {
+         if (refl.location.push_vertex >= 0)
+            glUniform1f(refl.location.push_vertex, value);
+         if (refl.location.push_fragment >= 0)
+            glUniform1f(refl.location.push_fragment, value);
+      }
+      else
+         *reinterpret_cast<float *>(push_constant_buffer.data() + refl.push_constant_offset) = value;
+   }
+}
+
 void Pass::build_semantic_texture(uint8_t *buffer,
       slang_texture_semantic semantic, const Texture &texture)
 {
@@ -1515,6 +1563,12 @@ void Pass::build_semantics(uint8_t *buffer,
    build_semantic_uint(buffer, SLANG_SEMANTIC_ROTATION,
                       rotation);
 
+   build_semantic_float(buffer, SLANG_SEMANTIC_CORE_ASPECT,
+                      core_aspect);
+
+   build_semantic_float(buffer, SLANG_SEMANTIC_CORE_ASPECT_ROT,
+                      core_aspect_rot);
+
    build_semantic_uint(buffer, SLANG_SEMANTIC_TOTAL_SUBFRAMES,
                       total_subframes);
    build_semantic_uint(buffer, SLANG_SEMANTIC_CURRENT_SUBFRAME,
@@ -1764,6 +1818,8 @@ public:
    void set_frame_count_period(unsigned pass, unsigned period);
    void set_frame_direction(int32_t direction);
    void set_rotation(uint32_t rot);
+   void set_core_aspect(float coreaspect);
+   void set_core_aspect_rot(float coreaspectrot);
    void set_shader_subframes(uint32_t tot_subframes);
    void set_current_shader_subframe(uint32_t cur_subframe);
 #ifdef GL3_ROLLING_SCANLINE_SIMULATION
@@ -2259,6 +2315,21 @@ void gl3_filter_chain::set_rotation(uint32_t rot)
       passes[i]->set_rotation(rot);
 }
 
+void gl3_filter_chain::set_core_aspect(float coreaspect)
+{
+   unsigned i;
+   for (i = 0; i < passes.size(); i++)
+      passes[i]->set_core_aspect(coreaspect);
+}
+
+void gl3_filter_chain::set_core_aspect_rot(float coreaspectrot)
+{
+   unsigned i;
+   for (i = 0; i < passes.size(); i++)
+      passes[i]->set_core_aspect_rot(coreaspectrot);
+}
+
+
 void gl3_filter_chain::set_shader_subframes(uint32_t tot_subframes)
 {
    unsigned i;
@@ -2702,6 +2773,20 @@ void gl3_filter_chain_set_rotation(
    chain->set_rotation(rot);
 }
 
+void gl3_filter_chain_set_core_aspect(
+      gl3_filter_chain_t *chain,
+      float coreaspect)
+{
+   chain->set_core_aspect(coreaspect);
+}
+
+void gl3_filter_chain_set_core_aspect_rot(
+      gl3_filter_chain_t *chain,
+      float coreaspectrot)
+{
+   chain->set_core_aspect_rot(coreaspectrot);
+}
+
 void gl3_filter_chain_set_shader_subframes(
       gl3_filter_chain_t *chain,
       uint32_t tot_subframes)

gfx/drivers_shader/shader_gl3.h

@@ -56,7 +56,7 @@ struct gl3_filter_chain_pass_info
 
    float scale_x;
    float scale_y;
-   /* Ignored for the last pass, swapchain info 
+   /* Ignored for the last pass, swapchain info
     * will be used instead. */
    GLenum rt_format;
    /* For the last pass, make sure VIEWPORT scale
@@ -119,6 +119,14 @@ void gl3_filter_chain_set_rotation(
       gl3_filter_chain_t *chain,
       uint32_t rotation);
 
+void gl3_filter_chain_set_core_aspect(
+      gl3_filter_chain_t *chain,
+      float coreaspect);
+
+void gl3_filter_chain_set_core_aspect_rot(
+      gl3_filter_chain_t *chain,
+      float coreaspectrot);
+
 void gl3_filter_chain_set_shader_subframes(
       gl3_filter_chain_t *chain,
       uint32_t tot_subframes);
@@ -131,7 +139,7 @@ void gl3_filter_chain_set_current_shader_subframe(
 void gl3_filter_chain_set_simulate_scanline(
       gl3_filter_chain_t *chain,
       bool simulate_scanline);
-#endif // GL3_ROLLING_SCANLINE_SIMULATION     
+#endif // GL3_ROLLING_SCANLINE_SIMULATION
 
 void gl3_filter_chain_set_pass_name(
       gl3_filter_chain_t *chain,

gfx/drivers_shader/shader_gl_cg.c

@@ -87,12 +87,16 @@ struct shader_program_cg
    CGparameter frame_cnt_f;
    CGparameter frame_dir_f;
    CGparameter rotation_f;
+   CGparameter originalaspect_f;
+   CGparameter originalaspectrotated_f;
    CGparameter vid_size_v;
    CGparameter tex_size_v;
    CGparameter out_size_v;
    CGparameter frame_cnt_v;
    CGparameter frame_dir_v;
    CGparameter rotation_v;
+   CGparameter originalaspect_v;
+   CGparameter originalaspectrotated_v;
    CGparameter mvp;
 
    struct cg_fbo_params fbo[GFX_MAX_SHADERS];
@@ -310,7 +314,9 @@ static void gl_cg_set_texture_info(
 
 static void gl_cg_set_params(void *dat, void *shader_data)
 {
-   unsigned i;
+   unsigned int i;
+   int rot;
+   float originalaspectrotated;
    video_shader_ctx_params_t          *params =
       (video_shader_ctx_params_t*)dat;
    unsigned width                             = params->width;
@@ -361,11 +367,22 @@ static void gl_cg_set_params(void *dat, void *shader_data)
    cg_gl_set_param_1f(cg->prg[cg->active_idx].rotation_f, (float)retroarch_get_rotation());
    cg_gl_set_param_1f(cg->prg[cg->active_idx].rotation_v, (float)retroarch_get_rotation());
 
+   cg_gl_set_param_1f(cg->prg[cg->active_idx].originalaspect_f, (float)video_driver_get_core_aspect());
+   cg_gl_set_param_1f(cg->prg[cg->active_idx].originalaspect_v, (float)video_driver_get_core_aspect());
+
+   /* OriginalAspectRotated: return 1/aspect for 90 and 270 rotated content */
+   rot = retroarch_get_rotation();
+   originalaspectrotated = video_driver_get_core_aspect();
+   if (rot == 1 || rot == 3)
+      originalaspectrotated = 1/originalaspectrotated;
+   cg_gl_set_param_1f(cg->prg[cg->active_idx].originalaspectrotated_f, (float)originalaspectrotated);
+   cg_gl_set_param_1f(cg->prg[cg->active_idx].originalaspectrotated_v, (float)originalaspectrotated);
+
    set_param_2f(cg->prg[cg->active_idx].vid_size_v, width, height);
    set_param_2f(cg->prg[cg->active_idx].tex_size_v, tex_width, tex_height);
    set_param_2f(cg->prg[cg->active_idx].out_size_v, out_width, out_height);
 
-   if (  cg->prg[cg->active_idx].frame_cnt_f || 
+   if (  cg->prg[cg->active_idx].frame_cnt_f ||
          cg->prg[cg->active_idx].frame_cnt_v)
    {
       unsigned modulo = cg->shader->pass[cg->active_idx - 1].frame_count_mod;
@@ -797,12 +814,16 @@ static void gl_cg_set_program_attributes(void *data, unsigned i)
    cg->prg[i].frame_cnt_f = cgGetNamedParameter(cg->prg[i].fprg, "IN.frame_count");
    cg->prg[i].frame_dir_f = cgGetNamedParameter(cg->prg[i].fprg, "IN.frame_direction");
    cg->prg[i].rotation_f  = cgGetNamedParameter(cg->prg[i].fprg, "IN.rotation");
+   cg->prg[i].originalaspect_f = cgGetNamedParameter(cg->prg[i].fprg, "IN.originalaspect");
+   cg->prg[i].originalaspectrotated_f = cgGetNamedParameter(cg->prg[i].fprg, "IN.originalaspectrotated");
    cg->prg[i].vid_size_v = cgGetNamedParameter (cg->prg[i].vprg, "IN.video_size");
    cg->prg[i].tex_size_v = cgGetNamedParameter (cg->prg[i].vprg, "IN.texture_size");
    cg->prg[i].out_size_v = cgGetNamedParameter (cg->prg[i].vprg, "IN.output_size");
    cg->prg[i].frame_cnt_v = cgGetNamedParameter(cg->prg[i].vprg, "IN.frame_count");
    cg->prg[i].frame_dir_v = cgGetNamedParameter(cg->prg[i].vprg, "IN.frame_direction");
    cg->prg[i].rotation_v  = cgGetNamedParameter(cg->prg[i].vprg, "IN.rotation");
+   cg->prg[i].originalaspect_v = cgGetNamedParameter(cg->prg[i].fprg, "IN.originalaspect");
+   cg->prg[i].originalaspectrotated_v = cgGetNamedParameter(cg->prg[i].fprg, "IN.originalaspectrotated");
 
    cg->prg[i].mvp                 = cgGetNamedParameter(cg->prg[i].vprg, "modelViewProj");
    if (!cg->prg[i].mvp)

gfx/drivers_shader/shader_glsl.c

@@ -101,6 +101,9 @@ struct shader_uniforms
    /* Use int for maximal compatibility despite other drivers using uint. */
    int rotation;
 
+   float core_aspect;
+   float core_aspect_rot;
+
    int lut_texture[GFX_MAX_TEXTURES];
    unsigned frame_count_mod;
 
@@ -287,7 +290,7 @@ static const XXH64_hash_t gl_glsl_hash_shader(
    {
       XXH64_update(state, source[n], strlen(source[n]));
    }
-   
+
    XXH64_hash_t const hash = XXH64_digest(state);
 
    XXH64_freeState(state);
@@ -417,10 +420,10 @@ static bool gl_glsl_compile_shader(glsl_shader_data_t *glsl,
    source[2] = glsl->alias_define;
    source[3] = program;
 
-#if defined(ORBIS) 
+#if defined(ORBIS)
    {
       char save_path[250];
-      XXH64_hash_t const hash = 
+      XXH64_hash_t const hash =
          gl_glsl_hash_shader(source, ARRAY_SIZE(source));
       snprintf(save_path, sizeof(save_path),
             "/data/retroarch/temp/%lx.sb", hash);
@@ -468,7 +471,7 @@ static bool gl_glsl_compile_program(
       struct shader_program_info *program_info)
 {
    glsl_shader_data_t                 *glsl = (glsl_shader_data_t*)data;
-   struct shader_program_glsl_data *program = 
+   struct shader_program_glsl_data *program =
       (struct shader_program_glsl_data*)program_data;
    GLuint prog                              = glCreateProgram();
 
@@ -486,7 +489,8 @@ static bool gl_glsl_compile_program(
       if (!gl_glsl_compile_shader(
                glsl,
                program->vprg,
-               "#define VERTEX\n#define PARAMETER_UNIFORM\n", program_info->vertex))
+               "#define VERTEX\n#define PARAMETER_UNIFORM\n#define _HAS_ORIGINALASPECT_UNIFORMS\n",
+               program_info->vertex))
       {
          RARCH_ERR("Failed to compile vertex shader #%u\n", idx);
          goto error;
@@ -500,7 +504,7 @@ static bool gl_glsl_compile_program(
       RARCH_LOG("[GLSL]: Found GLSL fragment shader.\n");
       program->fprg = glCreateShader(GL_FRAGMENT_SHADER);
       if (!gl_glsl_compile_shader(glsl, program->fprg,
-               "#define FRAGMENT\n#define PARAMETER_UNIFORM\n",
+               "#define FRAGMENT\n#define PARAMETER_UNIFORM\n#define _HAS_ORIGINALASPECT_UNIFORMS\n",
                program_info->fragment))
       {
          RARCH_ERR("Failed to compile fragment shader #%u\n", idx);
@@ -750,6 +754,8 @@ static void gl_glsl_find_uniforms(glsl_shader_data_t *glsl,
    uni->frame_count     = gl_glsl_get_uniform(glsl, prog, "FrameCount");
    uni->frame_direction = gl_glsl_get_uniform(glsl, prog, "FrameDirection");
    uni->rotation        = gl_glsl_get_uniform(glsl, prog, "Rotation");
+   uni->core_aspect     = gl_glsl_get_uniform(glsl, prog, "OriginalAspect");
+   uni->core_aspect_rot = gl_glsl_get_uniform(glsl, prog, "OriginalAspectRotated");
 
    for (i = 0; i < glsl->shader->luts; i++)
       uni->lut_texture[i] = glGetUniformLocation(prog, glsl->shader->lut[i].id);
@@ -1112,7 +1118,7 @@ static void *gl_glsl_init(void *data, const char *path)
       goto error;
    }
 #else
-   if (      glsl_core 
+   if (      glsl_core
          && (!(glsl->shader->flags & SHDR_FLAG_MODERN)))
    {
       RARCH_ERR("[GL]: GL core context is used, but shader is not core compatible. Cannot use it.\n");
@@ -1358,6 +1364,18 @@ static void gl_glsl_set_params(void *dat, void *shader_data)
   if (uni->rotation >= 0)
       glUniform1i(uni->rotation, retroarch_get_rotation());
 
+  if (uni->core_aspect >= 0)
+      glUniform1f(uni->core_aspect, video_driver_get_core_aspect());
+
+  if (uni->core_aspect_rot >= 0) {
+      /* OriginalAspectRotated: return 1/aspect for 90 and 270 rotated content */
+      float core_aspect_rot = video_driver_get_core_aspect();
+      uint32_t rot = retroarch_get_rotation();
+      if (rot == 1 || rot == 3)
+         core_aspect_rot = 1/core_aspect_rot;
+      glUniform1f(uni->core_aspect_rot, core_aspect_rot);
+  }
+
    /* Set lookup textures. */
    for (i = 0; i < glsl->shader->luts; i++)
    {
@@ -1545,7 +1563,7 @@ static bool gl_glsl_set_mvp(void *shader_data, const void *mat_data)
    int loc;
    glsl_shader_data_t *glsl   = (glsl_shader_data_t*)shader_data;
 
-   if (      !glsl 
+   if (      !glsl
          || (!(glsl->shader->flags & SHDR_FLAG_MODERN)))
       return false;
 
@@ -1589,7 +1607,7 @@ static bool gl_glsl_set_coords(void *shader_data,
    const struct shader_uniforms *uni = glsl
       ? &glsl->uniforms[glsl->active_idx] : NULL;
 
-   if (     !glsl 
+   if (     !glsl
          || (!(glsl->shader->flags & SHDR_FLAG_MODERN))
          || !coords)
    {

gfx/drivers_shader/shader_vulkan.cpp

@@ -254,6 +254,8 @@ class Pass
 #endif /* VULKAN_ROLLING_SCANLINE_SIMULATION */
       void set_frame_direction(int32_t dir) { frame_direction = dir; }
       void set_rotation(uint32_t rot) { rotation = rot; }
+      void set_core_aspect(float coreaspect) { core_aspect = coreaspect; }
+      void set_core_aspect_rot(float coreaspectrot) { core_aspect_rot = coreaspectrot; }
       void set_name(const char *name) { pass_name = name; }
       const std::string &get_name() const { return pass_name; }
       glslang_filter_chain_filter get_source_filter() const {
@@ -328,6 +330,7 @@ class Pass
             unsigned width, unsigned height);
       void build_semantic_uint(uint8_t *data, slang_semantic semantic, uint32_t value);
       void build_semantic_int(uint8_t *data, slang_semantic semantic, int32_t value);
+      void build_semantic_float(uint8_t *data,slang_semantic semantic, float value);
       void build_semantic_parameter(uint8_t *data, unsigned index, float value);
       void build_semantic_texture_vec4(uint8_t *data,
             slang_texture_semantic semantic,
@@ -343,6 +346,8 @@ class Pass
       uint64_t frame_count        = 0;
       int32_t frame_direction     = 1;
       uint32_t rotation           = 0;
+      float core_aspect           = 0;
+      float core_aspect_rot       = 0;
       unsigned frame_count_period = 0;
       unsigned pass_number        = 0;
       uint32_t total_subframes    = 1;
@@ -411,6 +416,8 @@ struct vulkan_filter_chain
 #endif /* VULKAN_ROLLING_SCANLINE_SIMULATION */
       void set_frame_direction(int32_t direction);
       void set_rotation(uint32_t rot);
+      void set_core_aspect(float coreaspect);
+      void set_core_aspect_rot(float coreaspect);
       void set_pass_name(unsigned pass, const char *name);
 
       void add_static_texture(std::unique_ptr<StaticTexture> texture);
@@ -1441,6 +1448,21 @@ void vulkan_filter_chain::set_rotation(uint32_t rot)
       passes[i]->set_rotation(rot);
 }
 
+
+void vulkan_filter_chain::set_core_aspect(float coreaspect)
+{
+   unsigned i;
+   for (i = 0; i < passes.size(); i++)
+      passes[i]->set_core_aspect(coreaspect);
+}
+
+void vulkan_filter_chain::set_core_aspect_rot(float coreaspectrot)
+{
+   unsigned i;
+   for (i = 0; i < passes.size(); i++)
+      passes[i]->set_core_aspect_rot(coreaspectrot);
+}
+
 void vulkan_filter_chain::set_pass_name(unsigned pass, const char *name)
 {
    passes[pass]->set_name(name);
@@ -2279,6 +2301,19 @@ void Pass::build_semantic_int(uint8_t *data, slang_semantic semantic,
       *reinterpret_cast<int32_t*>(push.buffer.data() + (refl.push_constant_offset >> 2)) = value;
 }
 
+void Pass::build_semantic_float(uint8_t *data, slang_semantic semantic,
+                              float value)
+{
+   auto &refl = reflection.semantics[semantic];
+
+   if (data && refl.uniform)
+      *reinterpret_cast<float*>(data + reflection.semantics[semantic].ubo_offset) = value;
+
+   if (refl.push_constant)
+      *reinterpret_cast<float*>(push.buffer.data() + (refl.push_constant_offset >> 2)) = value;
+}
+
+
 void Pass::build_semantic_texture(VkDescriptorSet set, uint8_t *buffer,
       slang_texture_semantic semantic, const Texture &texture)
 {
@@ -2344,6 +2379,12 @@ void Pass::build_semantics(VkDescriptorSet set, uint8_t *buffer,
    build_semantic_uint(buffer, SLANG_SEMANTIC_ROTATION,
                       rotation);
 
+   build_semantic_float(buffer, SLANG_SEMANTIC_CORE_ASPECT,
+                      core_aspect);
+
+   build_semantic_float(buffer, SLANG_SEMANTIC_CORE_ASPECT_ROT,
+                      core_aspect_rot);
+
    /* Standard inputs */
    build_semantic_texture(set, buffer, SLANG_TEXTURE_SEMANTIC_ORIGINAL, original);
    build_semantic_texture(set, buffer, SLANG_TEXTURE_SEMANTIC_SOURCE, source);
@@ -3188,6 +3229,20 @@ void vulkan_filter_chain_set_rotation(
    chain->set_rotation(rot);
 }
 
+void vulkan_filter_chain_set_core_aspect(
+      vulkan_filter_chain_t *chain,
+      float coreaspect)
+{
+   chain->set_core_aspect(coreaspect);
+}
+
+void vulkan_filter_chain_set_core_aspect_rot(
+      vulkan_filter_chain_t *chain,
+      float coreaspectrot)
+{
+   chain->set_core_aspect_rot(coreaspectrot);
+}
+
 void vulkan_filter_chain_set_pass_name(
       vulkan_filter_chain_t *chain,
       unsigned pass,

gfx/drivers_shader/shader_vulkan.h

@@ -51,7 +51,7 @@ struct vulkan_filter_chain_pass_info
    float scale_x;
    float scale_y;
 
-   /* Ignored for the last pass, swapchain info 
+   /* Ignored for the last pass, swapchain info
     * will be used instead. */
    VkFormat rt_format;
    /* For the last pass, make sure VIEWPORT scale
@@ -102,7 +102,7 @@ void vulkan_filter_chain_set_shader(vulkan_filter_chain_t *chain,
 
 VkFormat vulkan_filter_chain_get_pass_rt_format(
       vulkan_filter_chain_t *chain,
-      unsigned pass);      
+      unsigned pass);
 
 bool vulkan_filter_chain_update_swapchain_info(vulkan_filter_chain_t *chain,
       const struct vulkan_filter_chain_swapchain_info *info);
@@ -139,6 +139,12 @@ void vulkan_filter_chain_set_frame_direction(vulkan_filter_chain_t *chain,
 void vulkan_filter_chain_set_rotation(vulkan_filter_chain_t *chain,
       uint32_t rot);
 
+void vulkan_filter_chain_set_core_aspect(vulkan_filter_chain_t *chain,
+      float coreaspect);
+
+void vulkan_filter_chain_set_core_aspect_rot(vulkan_filter_chain_t *chain,
+      float coreaspectrot);
+
 void vulkan_filter_chain_build_offscreen_passes(vulkan_filter_chain_t *chain,
       VkCommandBuffer cmd, const VkViewport *vp);
 void vulkan_filter_chain_build_viewport_pass(vulkan_filter_chain_t *chain,

gfx/drivers_shader/slang_process.cpp

@@ -157,6 +157,8 @@ static bool slang_process_reflection(
             "FrameCount",
             "FrameDirection",
             "Rotation",
+            "OriginalAspect",
+            "OriginalAspectRotated",
             "TotalSubFrames",
             "CurrentSubFrame",
          };

gfx/drivers_shader/slang_reflection.cpp

@@ -51,6 +51,8 @@ static const char *semantic_uniform_names[] = {
    "FrameCount",
    "FrameDirection",
    "Rotation",
+   "OriginalAspect",
+   "OriginalAspectRotated",
    "TotalSubFrames",
    "CurrentSubFrame",
 };
@@ -270,6 +272,14 @@ static bool validate_type_for_semantic(const spirv_cross::SPIRType &type, slang_
          return type.basetype == spirv_cross::SPIRType::UInt
             &&  type.vecsize  == 1
             &&  type.columns  == 1;
+      case SLANG_SEMANTIC_CORE_ASPECT:
+         return type.basetype == spirv_cross::SPIRType::Float
+            &&  type.vecsize  == 1
+            &&  type.columns  == 1;
+      case SLANG_SEMANTIC_CORE_ASPECT_ROT:
+         return type.basetype == spirv_cross::SPIRType::Float
+            &&  type.vecsize  == 1
+            &&  type.columns  == 1;
          /* float */
       case SLANG_SEMANTIC_FLOAT_PARAMETER:
          return type.basetype == spirv_cross::SPIRType::Float

gfx/drivers_shader/slang_reflection.h

@@ -72,14 +72,18 @@ enum slang_semantic
    SLANG_SEMANTIC_FRAME_DIRECTION  = 4,
    /* uint, rotation */
    SLANG_SEMANTIC_ROTATION         = 5,
+   /* float, rotation */
+   SLANG_SEMANTIC_CORE_ASPECT      = 6,
+   SLANG_SEMANTIC_CORE_ASPECT_ROT  = 7,
+
    /* uint, sub frames per content frame */
-   SLANG_SEMANTIC_TOTAL_SUBFRAMES  = 6,
+   SLANG_SEMANTIC_TOTAL_SUBFRAMES  = 8,
    /* uint, current sub frame */
-   SLANG_SEMANTIC_CURRENT_SUBFRAME = 7,
+   SLANG_SEMANTIC_CURRENT_SUBFRAME = 9,
    SLANG_NUM_BASE_SEMANTICS,
 
    /* float, user defined parameter, arrayed */
-   SLANG_SEMANTIC_FLOAT_PARAMETER  = 8,
+   SLANG_SEMANTIC_FLOAT_PARAMETER  = 10,
 
    SLANG_NUM_SEMANTICS,
    SLANG_INVALID_SEMANTIC          = -1

tools/cg2glsl.py

@@ -50,6 +50,8 @@ def replace_global_in(source):
         ('IN.frame_count', 'FrameCount'),
         ('IN.frame_direction', 'FrameDirection'),
         ('IN.rotation', 'Rotation'),
+        ('IN.core_aspect', 'OriginalAspect'),
+        ('IN.core_aspect_rot', 'OriginalAspectRotated'),
     ]
 
     for line in source.splitlines():
@@ -109,6 +111,8 @@ def no_uniform(elem):
         '_frame_count',
         '_frame_direction',
         '_rotation',
+        '_core_aspect',
+        '_core_aspect_rot',
         '_mvp_matrix',
         '_vertex_coord',
         'sampler2D'