[ARM PATCH] 1302/1: ARMv5 optimized findbit

Question: is there any reason to do all this with byte access rather than 
word access besides alignment issues?  Word access would be much faster.

arch/arm/lib/findbit.S

@@ -83,7 +83,15 @@ ENTRY(_find_next_zero_bit_be)
 /*
  * One or more bits in the LSB of r3 are assumed to be set.
  */
-.found:		tst	r3, #0x0f
+.found:
+#if __LINUX_ARM_ARCH__ >= 5
+		rsb	r1, r3, #0
+		and	r3, r3, r1
+		clz	r3, r3
+		rsb	r3, r3, #31
+		add	r0, r2, r3
+#else
+		tst	r3, #0x0f
 		addeq	r2, r2, #4
 		movne	r3, r3, lsl #4
 		tst	r3, #0x30
@@ -92,5 +100,6 @@ ENTRY(_find_next_zero_bit_be)
 		tst	r3, #0x40
 		addeq	r2, r2, #1
 		mov	r0, r2
+#endif
 		RETINSTR(mov,pc,lr)