diff --git a/video/out/gl_video.c b/video/out/gl_video.c
index 5cf832d201..e2b64968ce 100644
--- a/video/out/gl_video.c
+++ b/video/out/gl_video.c
@@ -826,9 +826,9 @@ static void shader_setup_scaler(char **shader, struct scaler *scaler, int pass)
 {
     const char *target = scaler->index == 0 ? "SAMPLE_L" : "SAMPLE_C";
     if (!scaler->kernel) {
-        *shader = talloc_asprintf_append(*shader, "#define %s sample_%s_%c\n",
-                                         target, scaler->name,
-                                         "lc"[scaler->index]);
+        *shader = talloc_asprintf_append(*shader, "#define %s(p0, p1, p2) "
+            "sample_%s(p0, p1, p2, filter_param1_%c)\n",
+            target, scaler->name, "lc"[scaler->index]);
     } else {
         int size = scaler->kernel->size;
         if (pass != -1) {
@@ -1033,13 +1033,13 @@ static void compile_shaders(struct gl_video *p)
         // Force using the luma scaler on chroma. If the "indirect" stage is
         // used, the actual scaling will happen in the next stage.
         shader_def(&header_conv, "SAMPLE_C",
-                   use_indirect ? "sample_bilinear_l" : "SAMPLE_L");
+                   use_indirect ? "SAMPLE_BILINEAR" : "SAMPLE_L");
     }
 
     if (use_indirect) {
         // We don't use filtering for the Y-plane (luma), because it's never
         // scaled in this scenario.
-        shader_def(&header_conv, "SAMPLE_L", "sample_bilinear_l");
+        shader_def(&header_conv, "SAMPLE_L", "SAMPLE_BILINEAR");
         shader_def_opt(&header_conv, "FIXED_SCALE", true);
         header_conv = t_concat(tmp, header, header_conv);
         p->indirect_program =
diff --git a/video/out/gl_video_shaders.glsl b/video/out/gl_video_shaders.glsl
index 6e135f0d11..7b6a1f07fa 100644
--- a/video/out/gl_video_shaders.glsl
+++ b/video/out/gl_video_shaders.glsl
@@ -172,6 +172,8 @@ vec4 sample_bilinear(VIDEO_SAMPLER tex, vec2 texsize, vec2 texcoord, float param
     return texture(tex, texcoord);
 }
 
+#define SAMPLE_BILINEAR(p0, p1, p2) sample_bilinear(p0, p1, p2, 0)
+
 // Explanation how bicubic scaling with only 4 texel fetches is done:
 //   http://www.mate.tue.nl/mate/pdfs/10318.pdf
 //   'Efficient GPU-Based Texture Interpolation using Uniform B-Splines'
@@ -352,22 +354,6 @@ vec4 sample_sharpen5(VIDEO_SAMPLER tex, vec2 texsize, vec2 texcoord, float param
     return p + t * param1;
 }
 
-#define CONCAT(a, b) a ## b
-
-#define SAMPLE_FILTER_LC(NAME)                                              \
-    vec4 CONCAT(NAME, _l)(VIDEO_SAMPLER tex, vec2 texsize, vec2 texcoord) { \
-        return NAME(tex, texsize, texcoord, filter_param1_l);               \
-    }                                                                       \
-                                                                            \
-    vec4 CONCAT(NAME, _c)(VIDEO_SAMPLER tex, vec2 texsize, vec2 texcoord) { \
-        return NAME(tex, texsize, texcoord, filter_param1_c);               \
-    }
-
-SAMPLE_FILTER_LC(sample_bilinear)
-SAMPLE_FILTER_LC(sample_bicubic_fast)
-SAMPLE_FILTER_LC(sample_sharpen3)
-SAMPLE_FILTER_LC(sample_sharpen5)
-
 void main() {
     vec2 chr_texcoord = texcoord;
 #ifdef USE_RECTANGLE