diff --git a/.gitignore b/.gitignore
index d957e69..c678d63 100644
--- a/.gitignore
+++ b/.gitignore
@@ -1,5 +1,8 @@
 .clang_complete
-src/*.o
-src/*.o.deps
+*/*.o
+*/*.o.deps
 neatvnc
 vgcore.*
+perf.data
+perf.data.old
+zrle-bench
diff --git a/Makefile b/Makefile
index 88b26a2..f4afc79 100644
--- a/Makefile
+++ b/Makefile
@@ -2,7 +2,7 @@ all: neatvnc
 
 DEPENDENCIES := pixman-1 libpng libuv
 
-CFLAGS := -g -O0 -std=gnu11 -D_GNU_SOURCE -Iinc \
+CFLAGS := -g -O3 -DNDEBUG -std=gnu11 -D_GNU_SOURCE -Iinc \
 	$(foreach dep,$(DEPENDENCIES),$(shell pkg-config --cflags $(dep)))
 
 LDFLAGS := $(foreach dep,$(DEPENDENCIES),$(shell pkg-config --libs $(dep)))
@@ -10,13 +10,20 @@ LDFLAGS := $(foreach dep,$(DEPENDENCIES),$(shell pkg-config --libs $(dep)))
 neatvnc: src/server.o src/util.o src/vec.o src/zrle.o src/pngfb.o
 	$(CC) $^ $(LDFLAGS) -o $@
 
+zrle-bench: bench/zrle-bench.o src/server.o src/util.o src/vec.o src/zrle.o \
+		src/pngfb.o
+	$(CC) $^ $(LDFLAGS) -o $@
+
 src/%.o: src/%.c
 	$(CC) -c $(CFLAGS) $< -o $@ -MMD -MP -MF $@.deps
 
+bench/%.o: bench/%.c
+	$(CC) -c $(CFLAGS) $< -o $@ -MMD -MP -MF $@.deps
+
 .PHONY: clean
 clean:
 	rm -f neatvnc
-	rm -f src/*.o src/*.deps
+	rm -f src/*.o src/*.deps bench/*.o bench/*.deps
 
 -include src/*.deps
 
diff --git a/bench/zrle-bench.c b/bench/zrle-bench.c
new file mode 100644
index 0000000..eead1f8
--- /dev/null
+++ b/bench/zrle-bench.c
@@ -0,0 +1,59 @@
+#include "zrle.h"
+#include "rfb-proto.h"
+#include "util.h"
+#include "vec.h"
+
+#include <stdlib.h>
+#include <libdrm/drm_fourcc.h>
+#include <pixman.h>
+
+int read_png_file(struct vnc_framebuffer* fb, const char *filename);
+
+int run_benchmark(const char *image)
+{
+	int rc = -1;
+
+	struct vnc_framebuffer fb;
+	rc = read_png_file(&fb, image);
+	if (rc < 0)
+		return -1;
+
+	struct rfb_pixel_format pixfmt;
+	rfb_pixfmt_from_fourcc(&pixfmt, DRM_FORMAT_ARGB8888);
+
+	struct pixman_region16 region;
+	pixman_region_init(&region);
+
+	pixman_region_union_rect(&region, &region, 0, 0, fb.width, fb.height);
+
+	struct vec frame;
+	vec_init(&frame, fb.width * fb.height * 3 / 2);
+
+	rc = zrle_encode_frame(&frame, &pixfmt, fb.addr, &pixfmt,
+			       fb.width, fb.height, &region);
+
+	if (rc < 0)
+		goto failure;
+
+	rc = 0;
+failure:
+	pixman_region_fini(&region);
+	vec_destroy(&frame);
+	free(fb.addr);
+	return 0;
+}
+
+int main(int argc, char *argv[])
+{
+	int rc = 0;
+
+	char *image = argv[1];
+
+	if (image)
+		return run_benchmark(image) < 0 ? 1 :0;
+
+	rc |= run_benchmark("test-images/tv-test-card.png") < 0 ? 1 : 0;
+	rc |= run_benchmark("test-images/lena-soderberg.png") < 0 ? 1 : 0;
+
+	return rc;
+}
diff --git a/inc/util.h b/inc/util.h
index 4567c52..2630c17 100644
--- a/inc/util.h
+++ b/inc/util.h
@@ -19,4 +19,6 @@ struct vnc_write_request {
 int vnc__write(uv_stream_t *stream, const void *payload, size_t size,
                uv_write_cb on_done);
 
+int rfb_pixfmt_from_fourcc(struct rfb_pixel_format *dst, uint32_t src);
+
 #endif /* _VNC_UTIL_H_ */
diff --git a/src/server.c b/src/server.c
index 17940d5..80fe04e 100644
--- a/src/server.c
+++ b/src/server.c
@@ -685,6 +685,7 @@ failure:
 
 int read_png_file(struct vnc_framebuffer* fb, const char *filename);
 
+/*
 int main(int argc, char *argv[])
 {
 	if (!argv[1]) {
@@ -709,3 +710,4 @@ int main(int argc, char *argv[])
 
 	uv_run(uv_default_loop(), UV_RUN_DEFAULT);
 }
+*/
diff --git a/src/zrle.c b/src/zrle.c
index 09149af..c4297b7 100644
--- a/src/zrle.c
+++ b/src/zrle.c
@@ -235,9 +235,6 @@ int zrle_encode_box(struct vec* out, const struct rfb_pixel_format *dst_fmt,
 		int tile_width = width - tile_x >= 64 ? 64 : width - tile_x;
 		int tile_height = height - tile_y >= 64 ? 64 : height - tile_y;
 
-		printf("Encoding tile @ %dx%d. width: %d, height: %d\n", tile_x,
-				tile_y, tile_width, tile_height);
-
 		zrle_encode_tile(&in, dst_fmt,
 				 ((uint32_t*)src) + tile_x + tile_y * width,
 				 src_fmt, stride, tile_width, tile_height);
@@ -251,7 +248,6 @@ int zrle_encode_box(struct vec* out, const struct rfb_pixel_format *dst_fmt,
 
 	/* There seems to be something extra at the end */
 	out->len -= 4;
-	printf("Encoded with size: %lu\n", out->len - size_index);
 
 	uint32_t out_size = htonl(out->len - size_index - 4);
 	memcpy(((uint8_t*)out->data) + size_index, &out_size, sizeof(out_size));