#define data2w w3
#define has_nul x4
#define diff x5
+ #define off1 x5
#define syndrome x6
- #define tmp1 x7
- #define tmp2 x8
- #define tmp3 x9
- #define zeroones x10
- #define pos x11
-
- /* Start of performance-critical section -- one 64B cache line. */
- .align 6
+ #define tmp x6
+ #define data3 x7
+ #define zeroones x8
+ #define shift x9
+ #define off2 x10
+
+ /* On big-endian early bytes are at MSB and on little-endian LSB.
+ LS_FW means shifting towards early bytes. */
+ #ifdef __AARCH64EB__
+ # define LS_FW lsl
+ #else
+ # define LS_FW lsr
+ #endif
+
+ /* NUL detection works on the principle that (X - 1) & (~X) & 0x80
+ (=> (X - 1) & ~(X | 0x7f)) is non-zero iff a byte is zero, and
+ can be done in parallel across the entire word.
+ Since carry propagation makes 0x1 bytes before a NUL byte appear
+ NUL too in big-endian, byte-reverse the data before the NUL check. */
+
+
-SYM_FUNC_START_WEAK_PI(strcmp)
+SYM_FUNC_START(__pi_strcmp)
- eor tmp1, src1, src2
- mov zeroones, #REP8_01
- tst tmp1, #7
+ sub off2, src2, src1
+ mov zeroones, REP8_01
+ and tmp, src1, 7
+ tst off2, 7
b.ne L(misaligned8)
- ands tmp1, src1, #7
- b.ne L(mutual_align)
- /* NUL detection works on the principle that (X - 1) & (~X) & 0x80
- (=> (X - 1) & ~(X | 0x7f)) is non-zero iff a byte is zero, and
- can be done in parallel across the entire word. */
+ cbnz tmp, L(mutual_align)
+
+ .p2align 4
+
L(loop_aligned):
- ldr data1, [src1], #8
- ldr data2, [src2], #8
+ ldr data2, [src1, off2]
+ ldr data1, [src1], 8
L(start_realigned):
- sub tmp1, data1, zeroones
- orr tmp2, data1, #REP8_7f
- eor diff, data1, data2 /* Non-zero if differences found. */
- bic has_nul, tmp1, tmp2 /* Non-zero if NUL terminator. */
+ #ifdef __AARCH64EB__
+ rev tmp, data1
+ sub has_nul, tmp, zeroones
+ orr tmp, tmp, REP8_7f
+ #else
+ sub has_nul, data1, zeroones
+ orr tmp, data1, REP8_7f
+ #endif
+ bics has_nul, has_nul, tmp /* Non-zero if NUL terminator. */
+ ccmp data1, data2, 0, eq
+ b.eq L(loop_aligned)
+ #ifdef __AARCH64EB__
+ rev has_nul, has_nul
+ #endif
+ eor diff, data1, data2
orr syndrome, diff, has_nul
- cbz syndrome, L(loop_aligned)
- /* End of performance-critical section -- one 64B cache line. */
-
L(end):
- #ifndef __AARCH64EB__
+ #ifndef __AARCH64EB__
rev syndrome, syndrome
rev data1, data1
- /* The MS-non-zero bit of the syndrome marks either the first bit
- that is different, or the top bit of the first zero byte.
- Shifting left now will bring the critical information into the
- top bits. */
- clz pos, syndrome
rev data2, data2
- lsl data1, data1, pos
- lsl data2, data2, pos
- /* But we need to zero-extend (char is unsigned) the value and then
- perform a signed 32-bit subtraction. */
- lsr data1, data1, #56
- sub result, data1, data2, lsr #56
- ret
- #else
- /* For big-endian we cannot use the trick with the syndrome value
- as carry-propagation can corrupt the upper bits if the trailing
- bytes in the string contain 0x01. */
- /* However, if there is no NUL byte in the dword, we can generate
- the result directly. We can't just subtract the bytes as the
- MSB might be significant. */
- cbnz has_nul, 1f
- cmp data1, data2
- cset result, ne
- cneg result, result, lo
- ret
- 1:
- /* Re-compute the NUL-byte detection, using a byte-reversed value. */
- rev tmp3, data1
- sub tmp1, tmp3, zeroones
- orr tmp2, tmp3, #REP8_7f
- bic has_nul, tmp1, tmp2
- rev has_nul, has_nul
- orr syndrome, diff, has_nul
- clz pos, syndrome
- /* The MS-non-zero bit of the syndrome marks either the first bit
- that is different, or the top bit of the first zero byte.
+ #endif
+ clz shift, syndrome
+ /* The most-significant-non-zero bit of the syndrome marks either the
+ first bit that is different, or the top bit of the first zero byte.
Shifting left now will bring the critical information into the
top bits. */
- lsl data1, data1, pos
- lsl data2, data2, pos
+ lsl data1, data1, shift
+ lsl data2, data2, shift
/* But we need to zero-extend (char is unsigned) the value and then
perform a signed 32-bit subtraction. */
- lsr data1, data1, #56
- sub result, data1, data2, lsr #56
+ lsr data1, data1, 56
+ sub result, data1, data2, lsr 56
ret
- #endif
+
+ .p2align 4
L(mutual_align):
/* Sources are mutually aligned, but are not currently at an
L(done):
sub result, data1, data2
ret
-
-SYM_FUNC_END_PI(strcmp)
+SYM_FUNC_END(__pi_strcmp)
+SYM_FUNC_ALIAS_WEAK(strcmp, __pi_strcmp)
- EXPORT_SYMBOL_NOHWKASAN(strcmp)
+ EXPORT_SYMBOL_NOKASAN(strcmp)