first it worked, then i broke it and finally remembered to commit

shaders/decay_cs.glsl

@@ -12,8 +12,17 @@ layout(r32f, set = 1, binding = 0) uniform restrict writeonly image2D output_ima
 layout(push_constant, std430) uniform Params {
     vec2 texture_size;
     float decay_factor;
+    float dissipation_factor;
 } params;
 
+float get_dissipation_from(ivec2 uv) {
+    float v = imageLoad(current_image, uv).r;
+    return v * params.dissipation_factor;
+}
+
+void store_value(ivec2 uv, vec4 v) {
+    imageStore(output_image, uv, v);
+}
 
 // The code we want to execute in each invocation
 void main() {
@@ -21,24 +30,31 @@ void main() {
     ivec2 size = ivec2(params.texture_size.x - 1, params.texture_size.y - 1);
     ivec2 uv = ivec2(gl_GlobalInvocationID.xy);
 
+    ivec2 tl = ivec2(0, 0);
+
+    ivec2 up_uv = clamp(uv - ivec2(0, 1), tl, size);
+    ivec2 down_uv = clamp(uv + ivec2(0, 1), tl, size);
+    ivec2 left_uv = clamp(uv - ivec2(1, 0), tl, size);
+    ivec2 right_uv = clamp(uv + ivec2(1, 0), tl, size);
+
     // Just in case the texture size is not divisable by 8.
     if ((uv.x > size.x) || (uv.y > size.y)) {
         return;
     }
 
-    ivec2 tl = ivec2(0, 0);
     float current_v = imageLoad(current_image, uv).r;
-    float up_v = imageLoad(current_image, clamp(uv - ivec2(0, 1), tl, size)).r;
-    float down_v = imageLoad(current_image, clamp(uv + ivec2(0, 1), tl, size)).r;
-    float left_v = imageLoad(current_image, clamp(uv - ivec2(1, 0), tl, size)).r;
-    float right_v = imageLoad(current_image, clamp(uv + ivec2(1, 0), tl, size)).r;
 
     float new_v = current_v * params.decay_factor;
+    new_v += get_dissipation_from(up_uv);
+    new_v += get_dissipation_from(down_uv);
+    new_v += get_dissipation_from(right_uv);
+    new_v += get_dissipation_from(left_uv);
+
+
+    new_v = clamp(new_v, 0, 1);
 
-    if (new_v < 0.0) {
-        new_v = 0.0;
-    }
     vec4 result = vec4(new_v, new_v, new_v, 1.0);
-
     imageStore(output_image, uv, result);
+
+
 }