[client] egl: switch to using GL_TEXTURE_EXTERNAL_OES for DMABUF

According to Erik @ NVidia the open source NVidia driver will not
create a EGLImage from a DMABUF if the target is not
GL_TEXTURE_EXTERNAL_OES. This change set converts the dmabuf texture
from GL_TEXTURE_2D to GL_TEXTURE_EXTERNAL_OES and at runtime performs a
global search & replace on fragment shaders as needed to remain
compatible, replacing `sampler2D` with `samplerExternalOES`.

Ref: https://github.com/NVIDIA/open-gpu-kernel-modules/discussions/243#discussioncomment-3283415

client/renderers/EGL/cursor.c

@@ -100,7 +100,7 @@ static bool cursorTexInit(struct CursorTex * t,
   }
 
   if (!egl_shaderCompile(t->shader,
-        vertex_code, vertex_size, fragment_code, fragment_size))
+        vertex_code, vertex_size, fragment_code, fragment_size, false))
   {
     DEBUG_ERROR("Failed to compile the cursor shader");
     return false;

client/renderers/EGL/damage.c

@@ -76,7 +76,8 @@ bool egl_damageInit(EGL_Damage ** damage)
 
   if (!egl_shaderCompile((*damage)->shader,
         b_shader_damage_vert, b_shader_damage_vert_size,
-        b_shader_damage_frag, b_shader_damage_frag_size))
+        b_shader_damage_frag, b_shader_damage_frag_size,
+        false))
   {
     DEBUG_ERROR("Failed to compile the damage shader");
     return false;

client/renderers/EGL/desktop.c

@@ -46,6 +46,7 @@ struct DesktopShader
   EGL_Shader * shader;
   GLint uTransform;
   GLint uDesktopSize;
+  GLint uSamplerType;
   GLint uScaleAlgo;
   GLint uNVGain;
   GLint uCBMode;
@@ -57,7 +58,7 @@ struct EGL_Desktop
   EGLDisplay * display;
 
   EGL_Texture          * texture;
-  struct DesktopShader shader;
+  struct DesktopShader dmaShader, shader;
   EGL_DesktopRects     * mesh;
   CountedBuffer        * matrix;
 
@@ -92,7 +93,8 @@ void toggleNV(int key, void * opaque);
 static bool egl_initDesktopShader(
   struct DesktopShader * shader,
   const char * vertex_code  , size_t vertex_size,
-  const char * fragment_code, size_t fragment_size
+  const char * fragment_code, size_t fragment_size,
+  bool useDMA
 )
 {
   if (!egl_shaderInit(&shader->shader))
@@ -100,7 +102,8 @@ static bool egl_initDesktopShader(
 
   if (!egl_shaderCompile(shader->shader,
         vertex_code  , vertex_size,
-        fragment_code, fragment_size))
+        fragment_code, fragment_size,
+        useDMA))
   {
     return false;
   }
@@ -135,15 +138,6 @@ bool egl_desktopInit(EGL * egl, EGL_Desktop ** desktop_, EGLDisplay * display,
     return false;
   }
 
-  if (!egl_initDesktopShader(
-    &desktop->shader,
-    b_shader_desktop_vert    , b_shader_desktop_vert_size,
-    b_shader_desktop_rgb_frag, b_shader_desktop_rgb_frag_size))
-  {
-    DEBUG_ERROR("Failed to initialize the generic desktop shader");
-    return false;
-  }
-
   if (!egl_desktopRectsInit(&desktop->mesh, maxRects))
   {
     DEBUG_ERROR("Failed to initialize the desktop mesh");
@@ -157,6 +151,27 @@ bool egl_desktopInit(EGL * egl, EGL_Desktop ** desktop_, EGLDisplay * display,
     return false;
   }
 
+  if (!egl_initDesktopShader(
+    &desktop->shader,
+    b_shader_desktop_vert    , b_shader_desktop_vert_size,
+    b_shader_desktop_rgb_frag, b_shader_desktop_rgb_frag_size,
+    false))
+  {
+    DEBUG_ERROR("Failed to initialize the desktop shader");
+    return false;
+  }
+
+  if (useDMA)
+    if (!egl_initDesktopShader(
+      &desktop->dmaShader,
+      b_shader_desktop_vert    , b_shader_desktop_vert_size,
+      b_shader_desktop_rgb_frag, b_shader_desktop_rgb_frag_size,
+      true))
+    {
+      DEBUG_ERROR("Failed to initialize the desktop DMA shader");
+      return false;
+    }
+
   app_registerKeybind(0, 'N', toggleNV, desktop,
       "Toggle night vision mode");
 
@@ -205,11 +220,12 @@ void egl_desktopFree(EGL_Desktop ** desktop)
   if (!*desktop)
     return;
 
-  egl_textureFree    (&(*desktop)->texture      );
+  egl_textureFree    (&(*desktop)->texture         );
-  egl_textureFree    (&(*desktop)->spiceTexture );
+  egl_textureFree    (&(*desktop)->spiceTexture    );
-  egl_shaderFree     (&(*desktop)->shader.shader);
+  egl_shaderFree     (&(*desktop)->shader   .shader);
-  egl_desktopRectsFree(&(*desktop)->mesh        );
+  egl_shaderFree     (&(*desktop)->dmaShader.shader);
-  countedBufferRelease(&(*desktop)->matrix      );
+  egl_desktopRectsFree(&(*desktop)->mesh           );
+  countedBufferRelease(&(*desktop)->matrix         );
 
   egl_postProcessFree(&(*desktop)->pp);
 
@@ -399,18 +415,17 @@ bool egl_desktopRender(EGL_Desktop * desktop, unsigned int outputWidth,
   if (atomic_exchange(&desktop->processFrame, false) ||
       egl_postProcessConfigModified(desktop->pp))
     egl_postProcessRun(desktop->pp, tex, desktop->mesh,
-        width, height, outputWidth, outputHeight);
+        width, height, outputWidth, outputHeight, desktop->useDMA);
 
   unsigned int finalSizeX, finalSizeY;
-  GLuint texture = egl_postProcessGetOutput(desktop->pp,
+  EGL_Texture * texture = egl_postProcessGetOutput(desktop->pp,
       &finalSizeX, &finalSizeY);
 
   glBindFramebuffer(GL_FRAMEBUFFER, 0);
   egl_resetViewport(desktop->egl);
 
   glActiveTexture(GL_TEXTURE0);
-  glBindTexture(GL_TEXTURE_2D, texture);
+  egl_textureBind(texture);
-  glBindSampler(0, tex->sampler);
 
   if (finalSizeX > width || finalSizeY > height)
     scaleType = EGL_DESKTOP_DOWNSCALE;
@@ -435,7 +450,10 @@ bool egl_desktopRender(EGL_Desktop * desktop, unsigned int outputWidth,
       scaleAlgo = desktop->scaleAlgo;
   }
 
-  const struct DesktopShader * shader = &desktop->shader;
+  const struct DesktopShader * shader =
+    desktop->useDMA && texture == desktop->texture ?
+      &desktop->dmaShader : &desktop->shader;
+
   EGL_Uniform uniforms[] =
   {
     {

client/renderers/EGL/filter.h

@@ -68,9 +68,11 @@ typedef struct EGL_FilterOps
   /* reads filter state from options */
   void (*loadState)(EGL_Filter * filter);
 
-  /* set the input format of the filter */
+  /* set the input format of the filter
+   * useDMA will be true if the texture provided needs to use samplerExternalOES
+   */
   bool (*setup)(EGL_Filter * filter, enum EGL_PixelFormat pixFmt,
-      unsigned int width, unsigned int height);
+      unsigned int width, unsigned int height, bool useDMA);
 
   /* set the output resolution hint for the filter
    * this is optional and only a hint */
@@ -87,8 +89,8 @@ typedef struct EGL_FilterOps
 
   /* runs the filter on the provided texture
    * returns the processed texture as the output */
-  GLuint (*run)(EGL_Filter * filter, EGL_FilterRects * rects,
+  EGL_Texture * (*run)(EGL_Filter * filter, EGL_FilterRects * rects,
-      GLuint texture);
+      EGL_Texture * texture);
 
   /* called when the filter output is no loger needed so it can release memory
    * this is optional */
@@ -133,9 +135,10 @@ static inline void egl_filterLoadState(EGL_Filter * filter)
 }
 
 static inline bool egl_filterSetup(EGL_Filter * filter,
-    enum EGL_PixelFormat pixFmt, unsigned int width, unsigned int height)
+    enum EGL_PixelFormat pixFmt, unsigned int width, unsigned int height,
+    bool useDMA)
 {
-  return filter->ops.setup(filter, pixFmt, width, height);
+  return filter->ops.setup(filter, pixFmt, width, height, useDMA);
 }
 
 static inline void egl_filterSetOutputResHint(EGL_Filter * filter,
@@ -156,8 +159,8 @@ static inline bool egl_filterPrepare(EGL_Filter * filter)
   return filter->ops.prepare(filter);
 }
 
-static inline GLuint egl_filterRun(EGL_Filter * filter,
+static inline EGL_Texture * egl_filterRun(EGL_Filter * filter,
-    EGL_FilterRects * rects, GLuint texture)
+    EGL_FilterRects * rects, EGL_Texture * texture)
 {
   return filter->ops.run(filter, rects, texture);
 }

client/renderers/EGL/filter_downscale.c

@@ -60,6 +60,7 @@ typedef struct EGL_FilterDownscale
   EGL_Shader * lanczos2;
 
   DownscaleFilter filter;
+  int useDMA;
   enum EGL_PixelFormat pixFmt;
   unsigned int width, height;
   float pixelSize;
@@ -157,55 +158,26 @@ static bool egl_filterDownscaleInit(EGL_Filter ** filter)
     return false;
   }
 
+  this->useDMA = -1;
+
   if (!egl_shaderInit(&this->nearest))
   {
     DEBUG_ERROR("Failed to initialize the shader");
     goto error_this;
   }
 
-  if (!egl_shaderCompile(this->nearest,
-        b_shader_basic_vert    , b_shader_basic_vert_size,
-        b_shader_downscale_frag, b_shader_downscale_frag_size)
-     )
-  {
-    DEBUG_ERROR("Failed to compile the shader");
-    goto error_shader;
-  }
-
   if (!egl_shaderInit(&this->linear))
   {
     DEBUG_ERROR("Failed to initialize the shader");
     goto error_this;
   }
 
-  if (!egl_shaderCompile(this->linear,
-        b_shader_basic_vert, b_shader_basic_vert_size,
-        b_shader_downscale_linear_frag, b_shader_downscale_linear_frag_size)
-     )
-  {
-    DEBUG_ERROR("Failed to compile the shader");
-    goto error_shader;
-  }
-
   if (!egl_shaderInit(&this->lanczos2))
   {
     DEBUG_ERROR("Failed to initialize the shader");
     goto error_this;
   }
 
-  if (!egl_shaderCompile(this->lanczos2,
-        b_shader_basic_vert, b_shader_basic_vert_size,
-        b_shader_downscale_lanczos2_frag, b_shader_downscale_lanczos2_frag_size)
-     )
-  {
-    DEBUG_ERROR("Failed to compile the shader");
-    goto error_shader;
-  }
-
-  this->uNearest.type = EGL_UNIFORM_TYPE_3F;
-  this->uNearest.location =
-    egl_shaderGetUniform(this->nearest, "uConfig");
-
   if (!egl_framebufferInit(&this->fb))
   {
     DEBUG_ERROR("Failed to initialize the framebuffer");
@@ -326,7 +298,8 @@ static bool egl_filterDownscaleImguiConfig(EGL_Filter * filter)
 }
 
 static bool egl_filterDownscaleSetup(EGL_Filter * filter,
-    enum EGL_PixelFormat pixFmt, unsigned int width, unsigned int height)
+    enum EGL_PixelFormat pixFmt, unsigned int width, unsigned int height,
+    bool useDMA)
 {
   EGL_FilterDownscale * this = UPCAST(EGL_FilterDownscale, filter);
 
@@ -336,6 +309,48 @@ static bool egl_filterDownscaleSetup(EGL_Filter * filter,
   if (!this->enable)
     return false;
 
+  if (this->useDMA != useDMA)
+  {
+    if (!egl_shaderCompile(this->nearest,
+          b_shader_basic_vert    , b_shader_basic_vert_size,
+          b_shader_downscale_frag,
+          b_shader_downscale_frag_size,
+          useDMA)
+       )
+    {
+      DEBUG_ERROR("Failed to compile the shader");
+      return false;
+    }
+
+    if (!egl_shaderCompile(this->linear,
+          b_shader_basic_vert, b_shader_basic_vert_size,
+          b_shader_downscale_linear_frag,
+          b_shader_downscale_linear_frag_size,
+          useDMA)
+       )
+    {
+      DEBUG_ERROR("Failed to compile the shader");
+      return false;
+    }
+
+    if (!egl_shaderCompile(this->lanczos2,
+          b_shader_basic_vert, b_shader_basic_vert_size,
+          b_shader_downscale_lanczos2_frag,
+          b_shader_downscale_lanczos2_frag_size,
+          useDMA)
+       )
+    {
+      DEBUG_ERROR("Failed to compile the shader");
+      return false;
+    }
+
+    this->uNearest.type = EGL_UNIFORM_TYPE_3F;
+    this->uNearest.location =
+      egl_shaderGetUniform(this->nearest, "uConfig");
+
+    this->useDMA = useDMA;
+  }
+
   if (this->prepared               &&
       pixFmt       == this->pixFmt &&
       this->width  == width        &&
@@ -385,15 +400,15 @@ static bool egl_filterDownscalePrepare(EGL_Filter * filter)
   return true;
 }
 
-static GLuint egl_filterDownscaleRun(EGL_Filter * filter,
+static EGL_Texture * egl_filterDownscaleRun(EGL_Filter * filter,
-    EGL_FilterRects * rects, GLuint texture)
+    EGL_FilterRects * rects, EGL_Texture * texture)
 {
   EGL_FilterDownscale * this = UPCAST(EGL_FilterDownscale, filter);
 
   egl_framebufferBind(this->fb);
 
   glActiveTexture(GL_TEXTURE0);
-  glBindTexture(GL_TEXTURE_2D, texture);
+  egl_textureBind(texture);
 
   EGL_Shader * shader;
 

client/renderers/EGL/filter_ffx_cas.c

@@ -37,6 +37,7 @@ typedef struct EGL_FilterFFXCAS
   EGL_Shader * shader;
   bool         enable;
 
+  int useDMA;
   enum EGL_PixelFormat pixFmt;
   unsigned int width, height;
   float sharpness;
@@ -106,21 +107,14 @@ static bool egl_filterFFXCASInit(EGL_Filter ** filter)
     return false;
   }
 
+  this->useDMA = -1;
+
   if (!egl_shaderInit(&this->shader))
   {
     DEBUG_ERROR("Failed to initialize the shader");
     goto error_this;
   }
 
-  if (!egl_shaderCompile(this->shader,
-        b_shader_basic_vert  , b_shader_basic_vert_size,
-        b_shader_ffx_cas_frag, b_shader_ffx_cas_frag_size)
-     )
-  {
-    DEBUG_ERROR("Failed to compile the shader");
-    goto error_shader;
-  }
-
   this->consts = countedBufferNew(8 * sizeof(GLuint));
   if (!this->consts)
   {
@@ -128,12 +122,6 @@ static bool egl_filterFFXCASInit(EGL_Filter ** filter)
     goto error_shader;
   }
 
-  egl_shaderSetUniforms(this->shader, &(EGL_Uniform) {
-    .type     = EGL_UNIFORM_TYPE_4UIV,
-    .location = egl_shaderGetUniform(this->shader, "uConsts"),
-    .v        = this->consts,
-  }, 1);
-
   egl_filterFFXCASLoadState(&this->base);
 
   if (!egl_framebufferInit(&this->fb))
@@ -220,13 +208,35 @@ static bool egl_filterFFXCASImguiConfig(EGL_Filter * filter)
 }
 
 static bool egl_filterFFXCASSetup(EGL_Filter * filter,
-    enum EGL_PixelFormat pixFmt, unsigned int width, unsigned int height)
+    enum EGL_PixelFormat pixFmt, unsigned int width, unsigned int height,
+    bool useDMA)
 {
   EGL_FilterFFXCAS * this = UPCAST(EGL_FilterFFXCAS, filter);
 
   if (!this->enable)
     return false;
 
+  if (this->useDMA != useDMA)
+  {
+    if (!egl_shaderCompile(this->shader,
+          b_shader_basic_vert  , b_shader_basic_vert_size,
+          b_shader_ffx_cas_frag, b_shader_ffx_cas_frag_size,
+          useDMA)
+       )
+    {
+      DEBUG_ERROR("Failed to compile the shader");
+      return false;
+    }
+
+    egl_shaderSetUniforms(this->shader, &(EGL_Uniform) {
+      .type     = EGL_UNIFORM_TYPE_4UIV,
+      .location = egl_shaderGetUniform(this->shader, "uConsts"),
+      .v        = this->consts,
+    }, 1);
+
+    this->useDMA = useDMA;
+  }
+
   if (pixFmt == this->pixFmt && this->width == width && this->height == height)
     return true;
 
@@ -262,15 +272,15 @@ static bool egl_filterFFXCASPrepare(EGL_Filter * filter)
   return true;
 }
 
-static GLuint egl_filterFFXCASRun(EGL_Filter * filter,
+static EGL_Texture * egl_filterFFXCASRun(EGL_Filter * filter,
-    EGL_FilterRects * rects, GLuint texture)
+    EGL_FilterRects * rects, EGL_Texture * texture)
 {
   EGL_FilterFFXCAS * this = UPCAST(EGL_FilterFFXCAS, filter);
 
   egl_framebufferBind(this->fb);
 
   glActiveTexture(GL_TEXTURE0);
-  glBindTexture(GL_TEXTURE_2D, texture);
+  egl_textureBind(texture);
   glBindSampler(0, this->sampler);
 
   egl_shaderUse(this->shader);

client/renderers/EGL/filter_ffx_fsr1.c

@@ -42,6 +42,7 @@ typedef struct EGL_FilterFFXFSR1
   CountedBuffer * consts;
   EGL_Uniform     easuUniform[2], rcasUniform;
 
+  int useDMA;
   enum EGL_PixelFormat pixFmt;
   unsigned int width, height;
   unsigned int inWidth, inHeight;
@@ -109,6 +110,8 @@ static bool egl_filterFFXFSR1Init(EGL_Filter ** filter)
     return false;
   }
 
+  this->useDMA = -1;
+
   if (!egl_shaderInit(&this->easu))
   {
     DEBUG_ERROR("Failed to initialize the Easu shader");
@@ -121,18 +124,10 @@ static bool egl_filterFFXFSR1Init(EGL_Filter ** filter)
     goto error_esau;
   }
 
-  if (!egl_shaderCompile(this->easu,
-        b_shader_basic_vert        , b_shader_basic_vert_size,
-        b_shader_ffx_fsr1_easu_frag, b_shader_ffx_fsr1_easu_frag_size)
-     )
-  {
-    DEBUG_ERROR("Failed to compile the Easu shader");
-    goto error_rcas;
-  }
-
   if (!egl_shaderCompile(this->rcas,
         b_shader_basic_vert        , b_shader_basic_vert_size,
-        b_shader_ffx_fsr1_rcas_frag, b_shader_ffx_fsr1_rcas_frag_size)
+        b_shader_ffx_fsr1_rcas_frag, b_shader_ffx_fsr1_rcas_frag_size,
+        false)
      )
   {
     DEBUG_ERROR("Failed to compile the Rcas shader");
@@ -148,14 +143,6 @@ static bool egl_filterFFXFSR1Init(EGL_Filter ** filter)
 
   egl_filterFFXFSR1LoadState(&this->base);
 
-  this->easuUniform[0].type = EGL_UNIFORM_TYPE_4UIV;
-  this->easuUniform[0].location =
-    egl_shaderGetUniform(this->easu, "uConsts");
-  this->easuUniform[0].v = this->consts;
-  this->easuUniform[1].type = EGL_UNIFORM_TYPE_2F;
-  this->easuUniform[1].location =
-    egl_shaderGetUniform(this->easu, "uOutRes");
-
   this->rcasUniform.type = EGL_UNIFORM_TYPE_4UI;
   this->rcasUniform.location = egl_shaderGetUniform(this->rcas, "uConsts");
   rcasUpdateUniform(this);
@@ -335,13 +322,37 @@ static void egl_filterFFXFSR1SetOutputResHint(EGL_Filter * filter,
 }
 
 static bool egl_filterFFXFSR1Setup(EGL_Filter * filter,
-    enum EGL_PixelFormat pixFmt, unsigned int width, unsigned int height)
+    enum EGL_PixelFormat pixFmt, unsigned int width, unsigned int height,
+    bool useDMA)
 {
   EGL_FilterFFXFSR1 * this = UPCAST(EGL_FilterFFXFSR1, filter);
 
   if (!this->enable)
     return false;
 
+  if (this->useDMA != useDMA)
+  {
+    if (!egl_shaderCompile(this->easu,
+          b_shader_basic_vert        , b_shader_basic_vert_size,
+          b_shader_ffx_fsr1_easu_frag, b_shader_ffx_fsr1_easu_frag_size,
+          useDMA)
+       )
+    {
+      DEBUG_ERROR("Failed to compile the Easu shader");
+      return false;
+    }
+
+    this->easuUniform[0].type = EGL_UNIFORM_TYPE_4UIV;
+    this->easuUniform[0].location =
+      egl_shaderGetUniform(this->easu, "uConsts");
+    this->easuUniform[0].v = this->consts;
+    this->easuUniform[1].type = EGL_UNIFORM_TYPE_2F;
+    this->easuUniform[1].location =
+      egl_shaderGetUniform(this->easu, "uOutRes");
+
+    this->useDMA = useDMA;
+  }
+
   this->active = this->width > width && this->height > height;
   if (!this->active)
     return false;
@@ -395,15 +406,15 @@ static bool egl_filterFFXFSR1Prepare(EGL_Filter * filter)
   return true;
 }
 
-static GLuint egl_filterFFXFSR1Run(EGL_Filter * filter,
+static EGL_Texture * egl_filterFFXFSR1Run(EGL_Filter * filter,
-    EGL_FilterRects * rects, GLuint texture)
+    EGL_FilterRects * rects, EGL_Texture * texture)
 {
   EGL_FilterFFXFSR1 * this = UPCAST(EGL_FilterFFXFSR1, filter);
 
   // pass 1, Easu
   egl_framebufferBind(this->easuFb);
   glActiveTexture(GL_TEXTURE0);
-  glBindTexture(GL_TEXTURE_2D, texture);
+  egl_textureBind(texture);
   glBindSampler(0, this->sampler);
   egl_shaderUse(this->easu);
   egl_filterRectsRender(this->easu, rects);
@@ -412,7 +423,7 @@ static GLuint egl_filterFFXFSR1Run(EGL_Filter * filter,
   // pass 2, Rcas
   egl_framebufferBind(this->rcasFb);
   glActiveTexture(GL_TEXTURE0);
-  glBindTexture(GL_TEXTURE_2D, texture);
+  egl_textureBind(texture);
   glBindSampler(0, this->sampler);
   egl_shaderUse(this->rcas);
   egl_filterRectsRender(this->rcas, rects);

client/renderers/EGL/framebuffer.c

@@ -100,9 +100,7 @@ void egl_framebufferBind(EGL_Framebuffer * this)
   glViewport(0, 0, this->tex->format.width, this->tex->format.height);
 }
 
-GLuint egl_framebufferGetTexture(EGL_Framebuffer * this)
+EGL_Texture * egl_framebufferGetTexture(EGL_Framebuffer * this)
 {
-  GLuint output;
+  return this->tex;
-  egl_textureGet(this->tex, &output, NULL, NULL);
-  return output;
 }

client/renderers/EGL/framebuffer.h

@@ -32,4 +32,4 @@ bool egl_framebufferSetup(EGL_Framebuffer * this, enum EGL_PixelFormat pixFmt,
 
 void egl_framebufferBind(EGL_Framebuffer * this);
 
-GLuint egl_framebufferGetTexture(EGL_Framebuffer * this);
+EGL_Texture * egl_framebufferGetTexture(EGL_Framebuffer * this);

client/renderers/EGL/postprocess.c

@@ -49,7 +49,7 @@ static const EGL_FilterOps * EGL_Filters[] =
 struct EGL_PostProcess
 {
   Vector filters;
-  GLuint output;
+  EGL_Texture * output;
   unsigned int outputX, outputY;
   _Atomic(bool) modified;
 
@@ -606,7 +606,7 @@ bool egl_postProcessConfigModified(EGL_PostProcess * this)
 
 bool egl_postProcessRun(EGL_PostProcess * this, EGL_Texture * tex,
     EGL_DesktopRects * rects, int desktopWidth, int desktopHeight,
-    unsigned int targetX, unsigned int targetY)
+    unsigned int targetX, unsigned int targetY, bool useDMA)
 {
   if (targetX == 0 && targetY == 0)
     DEBUG_FATAL("targetX || targetY == 0");
@@ -614,8 +614,9 @@ bool egl_postProcessRun(EGL_PostProcess * this, EGL_Texture * tex,
   EGL_Filter * lastFilter = NULL;
   unsigned int sizeX, sizeY;
 
-  GLuint texture;
+  //TODO: clean this up
-  if (egl_textureGet(tex, &texture, &sizeX, &sizeY) != EGL_TEX_STATUS_OK)
+  GLuint _unused;
+  if (egl_textureGet(tex, &_unused, &sizeX, &sizeY) != EGL_TEX_STATUS_OK)
     return false;
 
   if (atomic_exchange(&this->modified, false))
@@ -636,11 +637,12 @@ bool egl_postProcessRun(EGL_PostProcess * this, EGL_Texture * tex,
   };
 
   EGL_Filter * filter;
+  EGL_Texture * texture = tex;
   vector_forEach(filter, &this->filters)
   {
     egl_filterSetOutputResHint(filter, targetX, targetY);
 
-    if (!egl_filterSetup(filter, tex->format.pixFmt, sizeX, sizeY) ||
+    if (!egl_filterSetup(filter, tex->format.pixFmt, sizeX, sizeY, useDMA) ||
         !egl_filterPrepare(filter))
       continue;
 
@@ -651,6 +653,9 @@ bool egl_postProcessRun(EGL_PostProcess * this, EGL_Texture * tex,
       egl_filterRelease(lastFilter);
 
     lastFilter = filter;
+
+    // the first filter to run will convert to a normal texture
+    useDMA = false;
   }
 
   this->output  = texture;
@@ -659,7 +664,7 @@ bool egl_postProcessRun(EGL_PostProcess * this, EGL_Texture * tex,
   return true;
 }
 
-GLuint egl_postProcessGetOutput(EGL_PostProcess * this,
+EGL_Texture * egl_postProcessGetOutput(EGL_PostProcess * this,
     unsigned int * outputX, unsigned int * outputY)
 {
   *outputX = this->outputX;

client/renderers/EGL/postprocess.h

@@ -41,7 +41,7 @@ bool egl_postProcessConfigModified(EGL_PostProcess * this);
  * targetX/Y is the final target output dimension hint if scalers are present */
 bool egl_postProcessRun(EGL_PostProcess * this, EGL_Texture * tex,
     EGL_DesktopRects * rects, int desktopWidth, int desktopHeight,
-    unsigned int targetX, unsigned int targetY);
+    unsigned int targetX, unsigned int targetY, bool useDMA);
 
-GLuint egl_postProcessGetOutput(EGL_PostProcess * this,
+EGL_Texture * egl_postProcessGetOutput(EGL_PostProcess * this,
     unsigned int * outputX, unsigned int * outputY);

client/renderers/EGL/shader.c

@@ -64,7 +64,8 @@ void egl_shaderFree(EGL_Shader ** shader)
   *shader = NULL;
 }
 
-bool egl_shaderLoad(EGL_Shader * this, const char * vertex_file, const char * fragment_file)
+bool egl_shaderLoad(EGL_Shader * this,
+    const char * vertex_file, const char * fragment_file, bool useDMA)
 {
   char   * vertex_code, * fragment_code;
   size_t   vertex_size,   fragment_size;
@@ -86,13 +87,16 @@ bool egl_shaderLoad(EGL_Shader * this, const char * vertex_file, const char * fr
 
   DEBUG_INFO("Loaded fragment shader: %s", fragment_file);
 
-  bool ret = egl_shaderCompile(this, vertex_code, vertex_size, fragment_code, fragment_size);
+  bool ret = egl_shaderCompile(this,
+      vertex_code, vertex_size, fragment_code, fragment_size,
+      useDMA);
+
   free(vertex_code);
   free(fragment_code);
   return ret;
 }
 
-bool egl_shaderCompile(EGL_Shader * this, const char * vertex_code,
+static bool shaderCompile(EGL_Shader * this, const char * vertex_code,
     size_t vertex_size, const char * fragment_code, size_t fragment_size)
 {
   if (this->hasShader)
@@ -204,6 +208,64 @@ bool egl_shaderCompile(EGL_Shader * this, const char * vertex_code,
   return true;
 }
 
+bool egl_shaderCompile(EGL_Shader * this, const char * vertex_code,
+    size_t vertex_size, const char * fragment_code, size_t fragment_size,
+    bool useDMA)
+{
+  if (useDMA)
+  {
+    const char * search  = "sampler2D";
+    const char * replace = "samplerExternalOES";
+
+    const char * src = fragment_code;
+    int instances = 0;
+    while((src = strstr(src, search)))
+    {
+      ++instances;
+      src += strlen(search);
+    }
+
+    const int diff = (strlen(replace) - strlen(search)) * instances;
+    char * newCode = malloc(fragment_size + diff);
+    if (!newCode)
+    {
+      DEBUG_ERROR("Out of memory");
+      return false;
+    }
+
+    src        = fragment_code;
+    char * dst = newCode;
+    for(int i = 0; i < instances; ++i)
+    {
+      const char * pos = strstr(src, search);
+      const int offset = pos - src;
+
+      memcpy(dst, src, offset);
+      dst += offset;
+      src  = pos + strlen(search);
+
+      memcpy(dst, replace, strlen(replace));
+      dst += strlen(replace);
+    }
+
+    const int final = fragment_size - (src - fragment_code);
+    memcpy(dst, src, final);
+    dst[final] = 0;
+
+    bool result = shaderCompile(
+        this,
+        vertex_code, vertex_size,
+        newCode    , fragment_size + diff);
+
+    free(newCode);
+    return result;
+  }
+
+  return shaderCompile(this,
+      vertex_code  , vertex_size,
+      fragment_code, fragment_size);
+}
+
 void egl_shaderSetUniforms(EGL_Shader * this, EGL_Uniform * uniforms, int count)
 {
   egl_shaderFreeUniforms(this);

client/renderers/EGL/shader.h

@@ -97,10 +97,11 @@ bool egl_shaderInit(EGL_Shader ** shader);
 void egl_shaderFree(EGL_Shader ** shader);
 
 bool egl_shaderLoad(EGL_Shader * model, const char * vertex_file,
-    const char * fragment_file);
+    const char * fragment_file, bool useDMA);
 
 bool egl_shaderCompile(EGL_Shader * model, const char * vertex_code,
-    size_t vertex_size, const char * fragment_code, size_t fragment_size);
+    size_t vertex_size, const char * fragment_code, size_t fragment_size,
+    bool useDMA);
 
 void egl_shaderSetUniforms(EGL_Shader * shader, EGL_Uniform * uniforms,
     int count);

client/renderers/EGL/shader/compat.h

@@ -1,16 +1,21 @@
 #if __VERSION__ == 300
   vec4 textureGather(sampler2D tex, vec2 uv, int comp)
   {
-    vec4 c0 = textureOffset(tex, uv, ivec2(0,1));
+    vec2 res = vec2(textureSize(tex, 0));
-    vec4 c1 = textureOffset(tex, uv, ivec2(1,1));
+    ivec2 p = ivec2((uv * res) - 0.5f);
-    vec4 c2 = textureOffset(tex, uv, ivec2(1,0));
+
-    vec4 c3 = textureOffset(tex, uv, ivec2(0,0));
+    // NOTE: we can't use texelFecthOffset because sampler2D may actually be samplerExternalOES
+    vec4 c0 = texelFetch(tex, p+ivec2(0,1), 0);
+    vec4 c1 = texelFetch(tex, p+ivec2(1,1), 0);
+    vec4 c2 = texelFetch(tex, p+ivec2(1,0), 0);
+    vec4 c3 = texelFetch(tex, p+ivec2(0,0), 0);
+
     return vec4(c0[comp], c1[comp], c2[comp],c3[comp]);
   }
 #elif __VERSION__ < 300
   vec4 textureGather(sampler2D tex, vec2 uv, int comp)
   {
-    vec4 c3 = texture2D(tex, uv);
+    vec4 c3 = texture(tex, uv);
     return vec4(c3[comp], c3[comp], c3[comp],c3[comp]);
   }
 #endif

client/renderers/EGL/shader/desktop_rgb.frag

@@ -1,4 +1,6 @@
 #version 300 es
+#extension GL_OES_EGL_image_external_essl3 : enable
+
 precision highp float;
 
 #define EGL_SCALE_AUTO    0
@@ -23,13 +25,17 @@ void main()
   switch (scaleAlgo)
   {
     case EGL_SCALE_NEAREST:
+    {
       vec2 ts = vec2(textureSize(sampler1, 0));
       color   = texelFetch(sampler1, ivec2(uv * ts), 0);
       break;
+    }
 
     case EGL_SCALE_LINEAR:
+    {
       color = texture(sampler1, uv);
       break;
+    }
   }
 
   if (cbMode > 0)

client/renderers/EGL/shader/downscale.frag

@@ -1,10 +1,12 @@
 #version 300 es
+#extension GL_OES_EGL_image_external_essl3 : enable
+
 precision highp float;
 
 in  vec2  fragCoord;
 out vec4  fragColor;
 
-uniform sampler2D texture;
+uniform sampler2D sampler1;
 uniform vec3      uConfig;
 
 void main()
@@ -13,7 +15,7 @@ void main()
   float vOffset   = uConfig.y;
   float hOffset   = uConfig.z;
 
-  vec2 inRes  = vec2(textureSize(texture, 0));
+  vec2 inRes  = vec2(textureSize(sampler1, 0));
   ivec2 point = ivec2(
     (floor((fragCoord * inRes) / pixelSize) * pixelSize) +
     pixelSize / 2.0f
@@ -22,5 +24,5 @@ void main()
   point.x += int(pixelSize * hOffset);
   point.y += int(pixelSize * vOffset);
 
-  fragColor = texelFetch(texture, point, 0);
+  fragColor = texelFetch(sampler1, point, 0);
 }

client/renderers/EGL/shader/downscale_lanczos2.frag

@@ -1,4 +1,6 @@
 #version 300 es
+#extension GL_OES_EGL_image_external_essl3 : enable
+
 precision highp float;
 
 #define PI 3.141592653589793
@@ -6,7 +8,7 @@ precision highp float;
 in  vec2  fragCoord;
 out vec4  fragColor;
 
-uniform sampler2D texture;
+uniform sampler2D sampler1;
 float sinc(float x)
 {
   return x == 0.0 ? 1.0 : sin(x * PI) / (x * PI);
@@ -24,7 +26,7 @@ float lanczos(vec2 v)
 
 void main()
 {
-  vec2 size = vec2(textureSize(texture, 0));
+  vec2 size = vec2(textureSize(sampler1, 0));
   vec2 pos = fragCoord * size;
   vec2 invSize = 1.0 / size;
   vec2 uvc = floor(pos) + vec2(0.5, 0.5);
@@ -54,7 +56,7 @@ void main()
 
   vec3 color = vec3(0.0);
   for (int i = 0; i < 9; ++i)
-    color += texture2D(texture, uvs[i] * invSize).rgb * factors[i];
+    color += texture(sampler1, uvs[i] * invSize).rgb * factors[i];
 
   fragColor = vec4(color, 1.0);
 }

client/renderers/EGL/shader/downscale_linear.frag

@@ -1,12 +1,14 @@
 #version 300 es
+#extension GL_OES_EGL_image_external_essl3 : enable
+
 precision highp float;
 
 in  vec2  fragCoord;
 out vec4  fragColor;
 
-uniform sampler2D texture;
+uniform sampler2D sampler1;
 
 void main()
 {
-  fragColor = texture2D(texture, fragCoord);
+  fragColor = texture(sampler1, fragCoord);
 }

client/renderers/EGL/shader/ffx_cas.frag

@@ -1,4 +1,6 @@
 #version 300 es
+#extension GL_OES_EGL_image_external_essl3 : enable
+
 precision highp float;
 
 #include "compat.h"
@@ -6,7 +8,7 @@ precision highp float;
 in  vec2  fragCoord;
 out vec4  fragColor;
 
-uniform sampler2D texture;
+uniform sampler2D sampler1;
 uniform uvec4     uConsts[2];
 
 #define A_GPU 1
@@ -16,7 +18,7 @@ uniform uvec4     uConsts[2];
 
 vec3 imageLoad(ivec2 point)
 {
-  return texelFetch(texture, point, 0).rgb;
+  return texelFetch(sampler1, point, 0).rgb;
 }
 
 AF3 CasLoad(ASU2 p)
@@ -30,7 +32,7 @@ void CasInput(inout AF1 r,inout AF1 g,inout AF1 b) {}
 
 void main()
 {
-  vec2  res   = vec2(textureSize(texture, 0));
+  vec2  res   = vec2(textureSize(sampler1, 0));
   uvec2 point = uvec2(fragCoord * res);
 
   CasFilter(

client/renderers/EGL/shader/ffx_fsr1_easu.frag

@@ -1,4 +1,6 @@
 #version 300 es
+#extension GL_OES_EGL_image_external_essl3 : enable
+
 precision highp float;
 
 #include "compat.h"
@@ -6,7 +8,7 @@ precision highp float;
 in  vec2  fragCoord;
 out vec4  fragColor;
 
-uniform sampler2D texture;
+uniform sampler2D sampler1;
 uniform vec2      uOutRes;
 uniform uvec4     uConsts[4];
 
@@ -18,20 +20,9 @@ uniform uvec4     uConsts[4];
 
 #define FSR_EASU_F 1
 
-vec4 _textureGather(sampler2D tex, vec2 uv, int comp)
+AF4 FsrEasuRF(AF2 p){return AF4(textureGather(sampler1, p, 0));}
-{
+AF4 FsrEasuGF(AF2 p){return AF4(textureGather(sampler1, p, 1));}
-  vec2 res = vec2(textureSize(tex, 0));
+AF4 FsrEasuBF(AF2 p){return AF4(textureGather(sampler1, p, 2));}
-  ivec2 p = ivec2((uv * res) - 0.5f);
-  vec4 c0 = texelFetchOffset(tex, p, 0, ivec2(0,1));
-  vec4 c1 = texelFetchOffset(tex, p, 0, ivec2(1,1));
-  vec4 c2 = texelFetchOffset(tex, p, 0, ivec2(1,0));
-  vec4 c3 = texelFetchOffset(tex, p, 0, ivec2(0,0));
-  return vec4(c0[comp], c1[comp], c2[comp],c3[comp]);
-}
-
-AF4 FsrEasuRF(AF2 p){return AF4(_textureGather(texture, p, 0));}
-AF4 FsrEasuGF(AF2 p){return AF4(_textureGather(texture, p, 1));}
-AF4 FsrEasuBF(AF2 p){return AF4(_textureGather(texture, p, 2));}
 
 #include "ffx_fsr1.h"
 

client/renderers/EGL/shader/ffx_fsr1_rcas.frag

@@ -6,7 +6,7 @@ precision highp float;
 in  vec2  fragCoord;
 out vec4  fragColor;
 
-uniform sampler2D texture;
+uniform sampler2D sampler1;
 uniform uvec4     uConsts;
 
 #define A_GPU  1
@@ -15,7 +15,7 @@ uniform uvec4     uConsts;
 
 #include "ffx_a.h"
 
-AF4 FsrRcasLoadF(ASU2 p) { return texelFetch(texture, ASU2(p), 0); }
+AF4 FsrRcasLoadF(ASU2 p) { return texelFetch(sampler1, ASU2(p), 0); }
 void FsrRcasInputF(inout AF1 r, inout AF1 g, inout AF1 b) {}
 
 #define FSR_RCAS_F       1
@@ -24,7 +24,7 @@ void FsrRcasInputF(inout AF1 r, inout AF1 g, inout AF1 b) {}
 
 void main()
 {
-  vec2  inRes = vec2(textureSize(texture, 0));
+  vec2  inRes = vec2(textureSize(sampler1, 0));
   uvec2 point = uvec2(fragCoord * (inRes + 0.5f));
 
   FsrRcasF(fragColor.r, fragColor.g, fragColor.b, point, uConsts);

client/renderers/EGL/texture.c

@@ -25,6 +25,7 @@
 #include "shader.h"
 #include "common/framebuffer.h"
 #include "common/debug.h"
+#include "common/array.h"
 
 #include <EGL/egl.h>
 #include <EGL/eglext.h>
@@ -76,6 +77,7 @@ bool egl_textureInit(EGL_Texture ** texture_, EGLDisplay * display,
   glSamplerParameteri(this->sampler, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
   glSamplerParameteri(this->sampler, GL_TEXTURE_WRAP_S    , GL_CLAMP_TO_EDGE);
   glSamplerParameteri(this->sampler, GL_TEXTURE_WRAP_T    , GL_CLAMP_TO_EDGE);
+
   return true;
 }
 
@@ -203,14 +205,7 @@ enum EGL_TexStatus egl_textureProcess(EGL_Texture * this)
 
 enum EGL_TexStatus egl_textureBind(EGL_Texture * this)
 {
-  GLuint tex;
-  EGL_TexStatus status;
-
-  if ((status = this->ops.get(this, &tex)) != EGL_TEX_STATUS_OK)
-    return status;
-
   glActiveTexture(GL_TEXTURE0);
-  glBindTexture(GL_TEXTURE_2D, tex);
   glBindSampler(0, this->sampler);
-  return EGL_TEX_STATUS_OK;
+  return this->ops.bind(this);
 }

client/renderers/EGL/texture.h

@@ -93,6 +93,9 @@ typedef struct EGL_TextureOps
 
   /* get the texture for use */
   enum EGL_TexStatus (*get)(EGL_Texture * texture, GLuint * tex);
+
+  /* bind the texture for use */
+  enum EGL_TexStatus (*bind)(EGL_Texture * texture);
 }
 EGL_TextureOps;
 

client/renderers/EGL/texture_buffer.c

@@ -292,6 +292,19 @@ EGL_TexStatus egl_texBufferStreamGet(EGL_Texture * texture, GLuint * tex)
   return EGL_TEX_STATUS_OK;
 }
 
+EGL_TexStatus egl_texBufferBind(EGL_Texture * texture)
+{
+  GLuint tex;
+  EGL_TexStatus status;
+
+  if ((status = texture->ops.get(texture, &tex)) != EGL_TEX_STATUS_OK)
+    return status;
+
+  glBindTexture(GL_TEXTURE_2D, tex);
+  return EGL_TEX_STATUS_OK;
+}
+
+
 const EGL_TextureOps EGL_TextureBuffer =
 {
   .init        = egl_texBufferInit,
@@ -299,7 +312,8 @@ const EGL_TextureOps EGL_TextureBuffer =
   .setup       = egl_texBufferSetup,
   .update      = egl_texBufferUpdate,
   .process     = egl_texBufferProcess,
-  .get         = egl_texBufferGet
+  .get         = egl_texBufferGet,
+  .bind        = egl_texBufferBind
 };
 
 const EGL_TextureOps EGL_TextureBufferStream =
@@ -309,5 +323,6 @@ const EGL_TextureOps EGL_TextureBufferStream =
   .setup       = egl_texBufferStreamSetup,
   .update      = egl_texBufferStreamUpdate,
   .process     = egl_texBufferStreamProcess,
-  .get         = egl_texBufferStreamGet
+  .get         = egl_texBufferStreamGet,
+  .bind        = egl_texBufferBind
 };

client/renderers/EGL/texture_buffer.h

@@ -48,6 +48,7 @@ void egl_texBufferFree(EGL_Texture * texture_);
 bool egl_texBufferSetup(EGL_Texture * texture_, const EGL_TexSetup * setup);
 EGL_TexStatus egl_texBufferProcess(EGL_Texture * texture_);
 EGL_TexStatus egl_texBufferGet(EGL_Texture * texture_, GLuint * tex);
+EGL_TexStatus egl_texBufferBind(EGL_Texture * texture_);
 
 bool egl_texBufferStreamInit(EGL_Texture ** texture_, EGL_TexType type,
     EGLDisplay * display);

client/renderers/EGL/texture_dmabuf.c

@@ -46,12 +46,26 @@ EGL_TextureOps EGL_TextureDMABUF;
 
 // internal functions
 
-static void egl_texDMABUFCleanup(TexDMABUF * this)
+static void egl_texDMABUFCleanup(EGL_Texture * texture)
 {
+  TextureBuffer * parent = UPCAST(TextureBuffer, texture);
+  TexDMABUF     * this   = UPCAST(TexDMABUF    , parent);
+
   struct FdImage * image;
   vector_forEachRef(image, &this->images)
     g_egl_dynProcs.eglDestroyImage(this->display, image->image);
   vector_clear(&this->images);
+
+  egl_texUtilFreeBuffers(parent->buf, parent->texCount);
+
+  if (parent->tex[0])
+    glDeleteTextures(parent->texCount, parent->tex);
+
+  if (parent->sync)
+  {
+    glDeleteSync(parent->sync);
+    parent->sync = 0;
+  }
 }
 
 // dmabuf functions
@@ -92,7 +106,7 @@ static void egl_texDMABUFFree(EGL_Texture * texture)
   TextureBuffer * parent = UPCAST(TextureBuffer, texture);
   TexDMABUF     * this   = UPCAST(TexDMABUF    , parent);
 
-  egl_texDMABUFCleanup(this);
+  egl_texDMABUFCleanup(texture);
   vector_destroy(&this->images);
 
   egl_texBufferFree(&parent->base);
@@ -102,11 +116,27 @@ static void egl_texDMABUFFree(EGL_Texture * texture)
 static bool egl_texDMABUFSetup(EGL_Texture * texture, const EGL_TexSetup * setup)
 {
   TextureBuffer * parent = UPCAST(TextureBuffer, texture);
-  TexDMABUF     * this   = UPCAST(TexDMABUF    , parent);
 
-  egl_texDMABUFCleanup(this);
+  egl_texDMABUFCleanup(texture);
 
-  return egl_texBufferSetup(&parent->base, setup);
+  glGenTextures(parent->texCount, parent->tex);
+  for(int i = 0; i < parent->texCount; ++i)
+  {
+    glBindTexture(GL_TEXTURE_EXTERNAL_OES, parent->tex[i]);
+    glTexImage2D(GL_TEXTURE_EXTERNAL_OES,
+        0,
+        texture->format.intFormat,
+        texture->format.width,
+        texture->format.height,
+        0,
+        texture->format.format,
+        texture->format.dataType,
+        NULL);
+  }
+
+  glBindTexture(GL_TEXTURE_EXTERNAL_OES, 0);
+  parent->rIndex = -1;
+  return true;
 }
 
 static bool egl_texDMABUFUpdate(EGL_Texture * texture,
@@ -173,8 +203,8 @@ static bool egl_texDMABUFUpdate(EGL_Texture * texture,
 
   INTERLOCKED_SECTION(parent->copyLock,
   {
-    glBindTexture(GL_TEXTURE_2D, parent->tex[parent->bufIndex]);
+    glBindTexture(GL_TEXTURE_EXTERNAL_OES, parent->tex[parent->bufIndex]);
-    g_egl_dynProcs.glEGLImageTargetTexture2DOES(GL_TEXTURE_2D, image);
+    g_egl_dynProcs.glEGLImageTargetTexture2DOES(GL_TEXTURE_EXTERNAL_OES, image);
 
     if (parent->sync)
       glDeleteSync(parent->sync);
@@ -238,6 +268,18 @@ static EGL_TexStatus egl_texDMABUFGet(EGL_Texture * texture, GLuint * tex)
   return EGL_TEX_STATUS_OK;
 }
 
+static EGL_TexStatus egl_texDMABUFBind(EGL_Texture * texture)
+{
+  GLuint tex;
+  EGL_TexStatus status;
+
+  if ((status = egl_texDMABUFGet(texture, &tex)) != EGL_TEX_STATUS_OK)
+    return status;
+
+  glBindTexture(GL_TEXTURE_EXTERNAL_OES, tex);
+  return EGL_TEX_STATUS_OK;
+}
+
 EGL_TextureOps EGL_TextureDMABUF =
 {
   .init        = egl_texDMABUFInit,
@@ -245,5 +287,6 @@ EGL_TextureOps EGL_TextureDMABUF =
   .setup       = egl_texDMABUFSetup,
   .update      = egl_texDMABUFUpdate,
   .process     = egl_texDMABUFProcess,
-  .get         = egl_texDMABUFGet
+  .get         = egl_texDMABUFGet,
+  .bind        = egl_texDMABUFBind
 };

client/renderers/EGL/texture_framebuffer.c

@@ -147,5 +147,6 @@ EGL_TextureOps EGL_TextureFrameBuffer =
   .setup   = egl_texFBSetup,
   .update  = egl_texFBUpdate,
   .process = egl_texBufferStreamProcess,
-  .get     = egl_texBufferStreamGet
+  .get     = egl_texBufferStreamGet,
+  .bind    = egl_texBufferBind
 };