arm64/lib/strncmp.S

14  * ARMv8-a, AArch64.
71 	   (=> (X - 1) & ~(X | 0x7f)) is non-zero iff a byte is zero, and
75 	ldr	data1, [src1], #8
76 	ldr	data2, [src2], #8
78 	subs	limit, limit, #8
91 	add	limit, limit, 8	/* Rewind limit to before last subs. */
103 	   perform a signed 32-bit subtraction.  */
109 	/* Not reached the limit, must have found the end or a diff.  */
111 	add	tmp1, limit, 8
136 	/* Re-compute the NUL-byte detection, using a byte-reversed value.  */
152 	   perform a signed 32-bit subtraction.  */
166 	ldr	data1, [src1], #8
168 	ldr	data2, [src2], #8
195 	/* Align the SRC1 to a dword by doing a bytewise compare and then do
217 	   src1     | a a a a a a a a | b b b c c c c c | . . .
218 	   src2     | x x x x x a a a   a a a a a b b b | c c c c c . . .
222 	   data1    a a a a a a a a     b b b c c c c c     b b b c c c c c
223 	   data2    a a a a a a a a     b b b 0 0 0 0 0     0 0 0 c c c c c
227 	   Align SRC2 down to 16 bytes. This way we can read 16 bytes at a
231 	/* Calculate offset from 8 byte alignment to string start in bits. No
237 	ldr	data1, [src1], #8
245 	/* STEP_A: Compare full 8 bytes when there is enough data from SRC2.*/
248 	subs	limit, limit, #8
249 	orr	data2, data2, tmp1	/* 8 bytes from SRC2 combined from two regs.*/
258 	ldr	data1, [src1], #8
263 	/* For big-endian we do a byte reverse to avoid carry-propagation
286 	cmp	limit, #8
294 	ldr	data1, [src1], #8
295 	sub	limit, limit, #8