Merge pull request #37 from asmblur/fixed-crt0

Fixed crt0.s .bss clearing loop. It would write a qword even if .bss …

stub/crt0.s

@@ -54,14 +54,16 @@ zerobss:
 	la	$2, _fbss
 	la	$3, _end
 loop:
+	sltu	$1,$2,$3
+	beq     $1,$0,zbss_done
 	nop
 	nop
-	nop
-	sq	$0,($2)
-	sltu	$1,$2,$3
-	bne	$1,$0,loop
+	sq	    $0,($2)
+    bgez    $0, loop
 	addiu	$2,$2,16
 
+zbss_done:
+
 # Some program loaders (such as Pukklink) execute programs as a thread, but
 # support passing argc and argv values via a0.  This also indicates that when
 # the program terminates, control should be returned to the program loader