diff --git a/src/bits.c b/src/bits.c
index bb6db02..e733961 100644
--- a/src/bits.c
+++ b/src/bits.c
@@ -9,9 +9,11 @@
 
 #define BUFFER_SIZE 4096
 #define IGNORE(param) do { (void)(param); } while(0)
+#define LIKELY(ex) __builtin_expect((ex), 1)
+#define UNLIKELY(ex) __builtin_expect((ex), 0)
 
-//TODO: For explicitly threaded version: Make a lazy thread spawner, spawning up to 8 threads which will each work on a byte and atomically write back to a central array of `uint64_t [num_of_bytes / num_of_threads]`. 
-//TODO: But how to sync the fwrites of this buffer? An atomic counter of which threads have completed their operations, and a condvar that allows the controlling thread to check if the counter is full before `fwrite_all()`ing the array, resetting the counter, and carrying on
+#define LIF(ex) if(LIKELY(!!(ex)))
+#define UIF(ex) if(UNLIKELY(!!(ex)))
 
 static inline int sfread(void* out, size_t *restrict size, FILE* f)
 {
@@ -28,15 +30,15 @@ static int fwrite_all(const void* _buf, size_t sz, size_t num, FILE* out)
 	else return 1;
 }
 
-static void pbits(char out[restrict 8], unsigned char byte, const char bit[static 2])
+static inline void pbits(char out[restrict 8], unsigned char byte, const char bit[static 2])
 {
 	for(register int i=8; i --> 0; byte >>= 1) *out++=bit[byte & 1];
-	//TODO: In explicitly threaded version, we can use an atomic (atomic.h) 64-bit integer `store()` to output those 8 bytes in `out`, which will be re-cast as `uint64_t out[restrict 1]`
 }
 
-static inline int check(int val, const char* msg)
+// Light assertion (likely to succeed)
+static inline int lcheck(int val, const char* msg)
 {
-	if(val) return val;
+	LIF(val) return val;
 	fprintf(stderr, "error: %s\n", msg);
 	exit(1);
 }
@@ -45,7 +47,7 @@ int main(int argc, char **argv)
 {
 	IGNORE(argc);
 
-	const char* mask = argv[1] && check(strlen(argv[1]) > 1, "expected 2 character mask") ? argv[1] : "01";
+	const char* mask = argv[1] && lcheck(strlen(argv[1]) > 1, "expected 2 character mask") ? argv[1] : "01";
 	unsigned char buf[BUFFER_SIZE];
 	char tbuf[BUFFER_SIZE * 8];
 	size_t bsz = BUFFER_SIZE;