diff --git a/piet-gpu/bin/winit.rs b/piet-gpu/bin/winit.rs
index 3568732..0b471b3 100644
--- a/piet-gpu/bin/winit.rs
+++ b/piet-gpu/bin/winit.rs
@@ -37,7 +37,7 @@ fn main() -> Result<(), Error> {
             .map(|_| device.create_cmd_buf())
             .collect::<Result<Vec<_>, Error>>()?;
         let query_pools = (0..NUM_FRAMES)
-            .map(|_| device.create_query_pool(5))
+            .map(|_| device.create_query_pool(8))
             .collect::<Result<Vec<_>, Error>>()?;
 
         let mut ctx = PietGpuRenderContext::new();
@@ -70,13 +70,17 @@ fn main() -> Result<(), Error> {
                     if current_frame >= NUM_FRAMES {
                         device.wait_and_reset(&[frame_fences[frame_idx]]).unwrap();
 
-                        let timestamps = device.reap_query_pool(query_pool).unwrap();
+                        let ts = device.reap_query_pool(query_pool).unwrap();
                         window.set_title(&format!(
-                            "e: {:.3}ms, b: {:.3}ms, c: {:.3}ms, f: {:.3}ms",
-                            timestamps[0] * 1e3,
-                            (timestamps[1] - timestamps[0]) * 1e3,
-                            (timestamps[2] - timestamps[1]) * 1e3,
-                            (timestamps[3] - timestamps[2]) * 1e3,
+                            "{:.3}ms :: e:{:.3}ms|alloc:{:.3}ms|cp:{:.3}ms|bd:{:.3}ms|bin:{:.3}ms|cr:{:.3}ms|r:{:.3}ms",
+                            ts[6] * 1e3,
+                            ts[0] * 1e3,
+                            (ts[1] - ts[0]) * 1e3,
+                            (ts[2] - ts[1]) * 1e3,
+                            (ts[3] - ts[2]) * 1e3,
+                            (ts[4] - ts[3]) * 1e3,
+                            (ts[5] - ts[4]) * 1e3,
+                            (ts[6] - ts[5]) * 1e3,
                         ));
                     }
 
@@ -84,8 +88,6 @@ fn main() -> Result<(), Error> {
                     let swap_image = swapchain.image(image_idx);
                     let cmd_buf = &mut cmd_buffers[frame_idx];
                     cmd_buf.begin();
-                    cmd_buf.reset_query_pool(&query_pool);
-
                     renderer.record(cmd_buf, &query_pool);
 
                     // Image -> Swapchain
diff --git a/piet-gpu/shader/kernel4.comp b/piet-gpu/shader/kernel4.comp
index e00320b..a88bc3d 100644
--- a/piet-gpu/shader/kernel4.comp
+++ b/piet-gpu/shader/kernel4.comp
@@ -8,9 +8,11 @@
 #extension GL_GOOGLE_include_directive : enable
 #extension GL_KHR_shader_subgroup_basic : enable
 
+#include "setup.h"
+
 #define CHUNK 8
-#define CHUNK_DY (16 / CHUNK)
-layout(local_size_x = 16, local_size_y = 2) in;
+#define CHUNK_DY (TILE_HEIGHT_PX / CHUNK)
+layout(local_size_x = TILE_WIDTH_PX, local_size_y = CHUNK_DY) in;
 
 // Same concern that this should be readonly as in kernel 3.
 layout(set = 0, binding = 0) buffer PtclBuf {
@@ -26,8 +28,6 @@ layout(rgba8, set = 0, binding = 2) uniform writeonly image2D image;
 #include "ptcl.h"
 #include "tile.h"
 
-#include "setup.h"
-
 void main() {
     uint tile_ix = gl_WorkGroupID.y * WIDTH_IN_TILES + gl_WorkGroupID.x;
     CmdRef cmd_ref = CmdRef(tile_ix * PTCL_INITIAL_ALLOC);
diff --git a/piet-gpu/src/lib.rs b/piet-gpu/src/lib.rs
index 5813ed3..8c720a2 100644
--- a/piet-gpu/src/lib.rs
+++ b/piet-gpu/src/lib.rs
@@ -263,8 +263,8 @@ impl<D: Device> Renderer<D> {
         let k4_code = include_bytes!("../shader/kernel4.spv");
         let k4_pipeline = device.create_simple_compute_pipeline(k4_code, 2, 1)?;
         let k4_ds = device.create_descriptor_set(
-            &k4_pipeline, 
-            &[&ptcl_buf, &tile_buf], 
+            &k4_pipeline,
+            &[&ptcl_buf, &tile_buf],
             &[&image_dev]
         )?;