Cycles / GPU:

* Enable the Non-Progressive integrator on GPU (CUDA) for testing.

In order to compile the CUDA kernel with it, you need at least 6GB of system memory and CUDA Toolkit 5.0 or 5.5.
It should also work with CUDA Toolkit 4.2, but in this case you should have 12GB of RAM. 

In case any problems arise, just change line 65 of kernel_types.h to disable Non-Progressive again. 
-- #define __NON_PROGRESSIVE__
++ //#define __NON_PROGRESSIVE__

intern/cycles/blender/addon/ui.py

@@ -51,20 +51,17 @@ class CyclesRender_PT_sampling(CyclesButtonsPanel, Panel):
 
         scene = context.scene
         cscene = scene.cycles
-        device_type = context.user_preferences.system.compute_device_type
 
         split = layout.split()
 
         col = split.column()
-        sub = col.column()
+        col.prop(cscene, "progressive")
-        sub.active = (device_type == 'NONE' or cscene.device == 'CPU')
-        sub.prop(cscene, "progressive")
 
         sub = col.column(align=True)
         sub.prop(cscene, "seed")
         sub.prop(cscene, "sample_clamp")
 
-        if cscene.progressive or (device_type != 'NONE' and cscene.device == 'GPU'):
+        if cscene.progressive:
             col = split.column()
             col.label(text="Samples:")
             sub = col.column(align=True)
@@ -610,7 +607,6 @@ class CyclesLamp_PT_lamp(CyclesButtonsPanel, Panel):
         lamp = context.lamp
         clamp = lamp.cycles
         cscene = context.scene.cycles
-        device_type = context.user_preferences.system.compute_device_type
 
         layout.prop(lamp, "type", expand=True)
 
@@ -629,7 +625,7 @@ class CyclesLamp_PT_lamp(CyclesButtonsPanel, Panel):
                 sub.prop(lamp, "size", text="Size X")
                 sub.prop(lamp, "size_y", text="Size Y")
 
-        if not cscene.progressive and (device_type == 'NONE' or cscene.device == 'CPU'):
+        if not cscene.progressive:
             col.prop(clamp, "samples")
 
         col = split.column()
@@ -817,7 +813,6 @@ class CyclesWorld_PT_settings(CyclesButtonsPanel, Panel):
         world = context.world
         cworld = world.cycles
         cscene = context.scene.cycles
-        device_type = context.user_preferences.system.compute_device_type
 
         col = layout.column()
 
@@ -825,7 +820,7 @@ class CyclesWorld_PT_settings(CyclesButtonsPanel, Panel):
         sub = col.row(align=True)
         sub.active = cworld.sample_as_light
         sub.prop(cworld, "sample_map_resolution")
-        if not cscene.progressive and (device_type == 'NONE' or cscene.device == 'CPU'):
+        if not cscene.progressive:
             sub.prop(cworld, "samples")
 
 

intern/cycles/blender/blender_sync.cpp

@@ -385,7 +385,7 @@ SessionParams BlenderSync::get_session_params(BL::RenderEngine b_engine, BL::Use
 	params.background = background;
 
 	/* samples */
-	if(get_boolean(cscene, "progressive") == 0 && params.device.type == DEVICE_CPU) {
+	if(get_boolean(cscene, "progressive") == 0) {
 		if(background) {
 			params.samples = get_int(cscene, "aa_samples");
 		}

intern/cycles/kernel/kernel_shader.h

@@ -916,8 +916,15 @@ __device void shader_merge_closures(KernelGlobals *kg, ShaderData *sd)
 				sci->sample_weight += scj->sample_weight;
 
 				int size = sd->num_closure - (j+1);
-				if(size > 0)
+				if(size > 0) {
+#ifdef __KERNEL_GPU__
+					for(int k = 0; k < size; k++) {
+						scj[k] = scj[k+1];
+					}
+#else
 					memmove(scj, scj+1, size*sizeof(ShaderClosure));
+#endif
+				}
 
 				sd->num_closure--;
 				j--;

intern/cycles/kernel/kernel_types.h

@@ -62,6 +62,7 @@ CCL_NAMESPACE_BEGIN
 #define __KERNEL_SHADING__
 #if __CUDA_ARCH__ >= 200
 #define __KERNEL_ADV_SHADING__
+#define __NON_PROGRESSIVE__
 #endif
 #endif