JehanneOS/newlib
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

2013-01-17 Marcus Shawcroft <marcus.shawcroft@linaro.org>

Browse Source 
* libc/machine/aarch64/strncmp.S: Correct arithmetic for
	    argument N values close to the maximum representable
	    value in an unsigned 64 bit value.
This commit is contained in:
Marcus Shawcroft 2013-01-17 14:53:32 +00:00
parent 78f66de6ce
commit 450fe1bfa3
2 changed files with 25 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
newlib/ChangeLog
View File

@ -1,3 +1,9 @@
2013-01-17 Marcus Shawcroft <marcus.shawcroft@linaro.org>
* libc/machine/aarch64/strncmp.S: Correct arithmetic for
argument N values close to the maximum representable
value in an unsigned 64 bit value.
2013-01-17 Marcus Shawcroft <marcus.shawcroft@linaro.org> 2013-01-17 Marcus Shawcroft <marcus.shawcroft@linaro.org>
* libc/machine/aarch64/strnlen.S: Correct arithmetic for * libc/machine/aarch64/strnlen.S: Correct arithmetic for

31
newlib/libc/machine/aarch64/strncmp.S
View File

@ -81,8 +81,10 @@ def_fn strncmp
b.ne .Lmisaligned8 b.ne .Lmisaligned8
ands tmp1, src1, #7 ands tmp1, src1, #7
b.ne .Lmutual_align b.ne .Lmutual_align
add limit_wd, limit, #7 /* Calculate the number of full and partial words -1. */
lsr limit_wd, limit_wd, #3 sub limit_wd, limit, #1 /* limit != 0, so no underflow. */
lsr limit_wd, limit_wd, #3 /* Convert to Dwords. */
/* NUL detection works on the principle that (X - 1) & (~X) & 0x80 /* NUL detection works on the principle that (X - 1) & (~X) & 0x80
(=> (X - 1) & ~(X | 0x7f)) is non-zero iff a byte is zero, and (=> (X - 1) & ~(X | 0x7f)) is non-zero iff a byte is zero, and
can be done in parallel across the entire word. */ can be done in parallel across the entire word. */
@ -95,14 +97,14 @@ def_fn strncmp
sub tmp1, data1, zeroones sub tmp1, data1, zeroones
orr tmp2, data1, #REP8_7f orr tmp2, data1, #REP8_7f
eor diff, data1, data2 /* Non-zero if differences found. */ eor diff, data1, data2 /* Non-zero if differences found. */
csinv endloop, diff, xzr, ne /* Last Dword or differences. */ csinv endloop, diff, xzr, pl /* Last Dword or differences. */
bics has_nul, tmp1, tmp2 /* Non-zero if NUL terminator. */ bics has_nul, tmp1, tmp2 /* Non-zero if NUL terminator. */
ccmp endloop, #0, #0, eq ccmp endloop, #0, #0, eq
b.eq .Lloop_aligned b.eq .Lloop_aligned
/* End of performance-critical section -- one 64B cache line. */ /* End of performance-critical section -- one 64B cache line. */
/* Not reached the limit, must have found the end or a diff. */ /* Not reached the limit, must have found the end or a diff. */
cbnz limit_wd, .Lnot_limit tbz limit_wd, #63, .Lnot_limit
/* Limit % 8 == 0 => all bytes significant. */ /* Limit % 8 == 0 => all bytes significant. */
ands limit, limit, #7 ands limit, limit, #7
@ -177,26 +179,31 @@ def_fn strncmp
.Lmutual_align: .Lmutual_align:
/* Sources are mutually aligned, but are not currently at an /* Sources are mutually aligned, but are not currently at an
alignment boundary. Round down the addresses and then mask off alignment boundary. Round down the addresses and then mask off
the bytes that precede the start point. */ the bytes that precede the start point.
We also need to adjust the limit calculations, but without
overflowing if the limit is near ULONG_MAX. */
bic src1, src1, #7 bic src1, src1, #7
bic src2, src2, #7 bic src2, src2, #7
add limit, limit, tmp1 /* Adjust the limit for the extra. */
lsl tmp1, tmp1, #3 /* Bytes beyond alignment -> bits. */
ldr data1, [src1], #8 ldr data1, [src1], #8
neg tmp1, tmp1 /* Bits to alignment -64. */ neg tmp3, tmp1, lsl #3 /* 64 - bits(bytes beyond align). */
ldr data2, [src2], #8 ldr data2, [src2], #8
mov tmp2, #~0 mov tmp2, #~0
sub limit_wd, limit, #1 /* limit != 0, so no underflow. */
#ifdef __AARCH64EB__ #ifdef __AARCH64EB__
/* Big-endian. Early bytes are at MSB. */ /* Big-endian. Early bytes are at MSB. */
lsl tmp2, tmp2, tmp1 /* Shift (tmp1 & 63). */ lsl tmp2, tmp2, tmp3 /* Shift (tmp1 & 63). */
#else #else
/* Little-endian. Early bytes are at LSB. */ /* Little-endian. Early bytes are at LSB. */
lsr tmp2, tmp2, tmp1 /* Shift (tmp1 & 63). */ lsr tmp2, tmp2, tmp3 /* Shift (tmp1 & 63). */
#endif #endif
add limit_wd, limit, #7 and tmp3, limit_wd, #7
lsr limit_wd, limit_wd, #3
/* Adjust the limit. Only low 3 bits used, so overflow irrelevant. */
add limit, limit, tmp1
add tmp3, tmp3, tmp1
orr data1, data1, tmp2 orr data1, data1, tmp2
orr data2, data2, tmp2 orr data2, data2, tmp2
lsr limit_wd, limit_wd, #3 add limit_wd, limit_wd, tmp3, lsr #3
b .Lstart_realigned b .Lstart_realigned
.Lret0: .Lret0: