Skip to content

Commit

Permalink
Handle page boundaries in x86 SSE4.2 strncmp.
Browse files Browse the repository at this point in the history
  • Loading branch information
H.J. Lu authored and Ulrich Drepper committed Mar 21, 2011
1 parent 2a11560 commit c97a128
Show file tree
Hide file tree
Showing 4 changed files with 81 additions and 17 deletions.
8 changes: 8 additions & 0 deletions ChangeLog
Original file line number Diff line number Diff line change
@@ -1,3 +1,11 @@
2011-03-20 H.J. Lu <hongjiu.lu@intel.com>

[BZ #12597]
* string/test-strncmp.c (do_page_test): New function.
(check2): Likewise.
(test_main): Call check2.
* sysdeps/x86_64/multiarch/strcmp.S: Properly cross page boundary.

2011-03-20 Ulrich Drepper <drepper@gmail.com>

[BZ #12587]
Expand Down
4 changes: 2 additions & 2 deletions NEWS
Original file line number Diff line number Diff line change
@@ -1,4 +1,4 @@
GNU C Library NEWS -- history of user-visible changes. 2011-3-20
GNU C Library NEWS -- history of user-visible changes. 2011-3-21
Copyright (C) 1992-2009, 2010, 2011 Free Software Foundation, Inc.
See the end for copying conditions.

Expand All @@ -9,7 +9,7 @@ Version 2.14

* The following bugs are resolved with this release:

11724, 12445, 12454, 12460, 12469, 12489, 12509, 12510, 12583, 12587
11724, 12445, 12454, 12460, 12469, 12489, 12509, 12510, 12583, 12587, 12597

Version 2.13

Expand Down
41 changes: 41 additions & 0 deletions string/test-strncmp.c
Original file line number Diff line number Diff line change
Expand Up @@ -199,6 +199,27 @@ do_test (size_t align1, size_t align2, size_t len, size_t n, int max_char,
putchar ('\n');
}

static void
do_page_test (size_t offset1, size_t offset2, char *s2)
{
char *s1;
int exp_result;

if (offset1 >= page_size || offset2 >= page_size)
return;

s1 = (char *) (buf1 + offset1);
s2 += offset2;

exp_result= *s1;

FOR_EACH_IMPL (impl, 0)
{
check_result (impl, s1, s2, page_size, -exp_result);
check_result (impl, s2, s1, page_size, exp_result);
}
}

static void
do_random_tests (void)
{
Expand Down Expand Up @@ -312,6 +333,25 @@ check1 (void)
}
}

static void
check2 (void)
{
size_t i;
char *s1, *s2;

s1 = (char *) buf1;
for (i = 0; i < page_size - 1; i++)
s1[i] = 23;
s1[i] = 0;

s2 = strdup (s1);

for (i = 0; i < 64; ++i)
do_page_test (3990 + i, 2635, s2);

free (s2);
}

int
test_main (void)
{
Expand All @@ -320,6 +360,7 @@ test_main (void)
test_init ();

check1 ();
check2 ();

printf ("%23s", "");
FOR_EACH_IMPL (impl, 0)
Expand Down
45 changes: 30 additions & 15 deletions sysdeps/x86_64/multiarch/strcmp.S
Original file line number Diff line number Diff line change
Expand Up @@ -452,6 +452,7 @@ LABEL(loop_ashr_1_use_sse4_2):
add $16, %r10
jg LABEL(nibble_ashr_1_use_sse4_2)

LABEL(nibble_ashr_1_use_sse4_2_restart):
movdqa (%rdi, %rdx), %xmm0
palignr $1, -16(%rdi, %rdx), %xmm0
# if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
Expand Down Expand Up @@ -499,7 +500,7 @@ LABEL(nibble_ashr_1_use_sse4_2):
jae LABEL(nibble_ashr_use_sse4_2_exit)
# endif
cmp $14, %ecx
ja LABEL(loop_ashr_1_use_sse4_2)
ja LABEL(nibble_ashr_1_use_sse4_2_restart)

jmp LABEL(nibble_ashr_use_sse4_2_exit)

Expand Down Expand Up @@ -544,6 +545,7 @@ LABEL(loop_ashr_2_use_sse4_2):
add $16, %r10
jg LABEL(nibble_ashr_2_use_sse4_2)

LABEL(nibble_ashr_2_use_sse4_2_restart):
movdqa (%rdi, %rdx), %xmm0
palignr $2, -16(%rdi, %rdx), %xmm0
# if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
Expand Down Expand Up @@ -591,7 +593,7 @@ LABEL(nibble_ashr_2_use_sse4_2):
jae LABEL(nibble_ashr_use_sse4_2_exit)
# endif
cmp $13, %ecx
ja LABEL(loop_ashr_2_use_sse4_2)
ja LABEL(nibble_ashr_2_use_sse4_2_restart)

jmp LABEL(nibble_ashr_use_sse4_2_exit)

Expand Down Expand Up @@ -636,6 +638,7 @@ LABEL(loop_ashr_3_use_sse4_2):
add $16, %r10
jg LABEL(nibble_ashr_3_use_sse4_2)

LABEL(nibble_ashr_3_use_sse4_2_restart):
movdqa (%rdi, %rdx), %xmm0
palignr $3, -16(%rdi, %rdx), %xmm0
# if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
Expand Down Expand Up @@ -683,7 +686,7 @@ LABEL(nibble_ashr_3_use_sse4_2):
jae LABEL(nibble_ashr_use_sse4_2_exit)
# endif
cmp $12, %ecx
ja LABEL(loop_ashr_3_use_sse4_2)
ja LABEL(nibble_ashr_3_use_sse4_2_restart)

jmp LABEL(nibble_ashr_use_sse4_2_exit)

Expand Down Expand Up @@ -729,6 +732,7 @@ LABEL(loop_ashr_4_use_sse4_2):
add $16, %r10
jg LABEL(nibble_ashr_4_use_sse4_2)

LABEL(nibble_ashr_4_use_sse4_2_restart):
movdqa (%rdi, %rdx), %xmm0
palignr $4, -16(%rdi, %rdx), %xmm0
# if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
Expand Down Expand Up @@ -776,7 +780,7 @@ LABEL(nibble_ashr_4_use_sse4_2):
jae LABEL(nibble_ashr_use_sse4_2_exit)
# endif
cmp $11, %ecx
ja LABEL(loop_ashr_4_use_sse4_2)
ja LABEL(nibble_ashr_4_use_sse4_2_restart)

jmp LABEL(nibble_ashr_use_sse4_2_exit)

Expand Down Expand Up @@ -822,6 +826,7 @@ LABEL(loop_ashr_5_use_sse4_2):
add $16, %r10
jg LABEL(nibble_ashr_5_use_sse4_2)

LABEL(nibble_ashr_5_use_sse4_2_restart):
movdqa (%rdi, %rdx), %xmm0
palignr $5, -16(%rdi, %rdx), %xmm0
# if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
Expand Down Expand Up @@ -870,7 +875,7 @@ LABEL(nibble_ashr_5_use_sse4_2):
jae LABEL(nibble_ashr_use_sse4_2_exit)
# endif
cmp $10, %ecx
ja LABEL(loop_ashr_5_use_sse4_2)
ja LABEL(nibble_ashr_5_use_sse4_2_restart)

jmp LABEL(nibble_ashr_use_sse4_2_exit)

Expand Down Expand Up @@ -916,6 +921,7 @@ LABEL(loop_ashr_6_use_sse4_2):
add $16, %r10
jg LABEL(nibble_ashr_6_use_sse4_2)

LABEL(nibble_ashr_6_use_sse4_2_restart):
movdqa (%rdi, %rdx), %xmm0
palignr $6, -16(%rdi, %rdx), %xmm0
# if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
Expand Down Expand Up @@ -963,7 +969,7 @@ LABEL(nibble_ashr_6_use_sse4_2):
jae LABEL(nibble_ashr_use_sse4_2_exit)
# endif
cmp $9, %ecx
ja LABEL(loop_ashr_6_use_sse4_2)
ja LABEL(nibble_ashr_6_use_sse4_2_restart)

jmp LABEL(nibble_ashr_use_sse4_2_exit)

Expand Down Expand Up @@ -1009,6 +1015,7 @@ LABEL(loop_ashr_7_use_sse4_2):
add $16, %r10
jg LABEL(nibble_ashr_7_use_sse4_2)

LABEL(nibble_ashr_7_use_sse4_2_restart):
movdqa (%rdi, %rdx), %xmm0
palignr $7, -16(%rdi, %rdx), %xmm0
# if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
Expand Down Expand Up @@ -1056,7 +1063,7 @@ LABEL(nibble_ashr_7_use_sse4_2):
jae LABEL(nibble_ashr_use_sse4_2_exit)
# endif
cmp $8, %ecx
ja LABEL(loop_ashr_7_use_sse4_2)
ja LABEL(nibble_ashr_7_use_sse4_2_restart)

jmp LABEL(nibble_ashr_use_sse4_2_exit)

Expand Down Expand Up @@ -1102,6 +1109,7 @@ LABEL(loop_ashr_8_use_sse4_2):
add $16, %r10
jg LABEL(nibble_ashr_8_use_sse4_2)

LABEL(nibble_ashr_8_use_sse4_2_restart):
movdqa (%rdi, %rdx), %xmm0
palignr $8, -16(%rdi, %rdx), %xmm0
# if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
Expand Down Expand Up @@ -1149,7 +1157,7 @@ LABEL(nibble_ashr_8_use_sse4_2):
jae LABEL(nibble_ashr_use_sse4_2_exit)
# endif
cmp $7, %ecx
ja LABEL(loop_ashr_8_use_sse4_2)
ja LABEL(nibble_ashr_8_use_sse4_2_restart)

jmp LABEL(nibble_ashr_use_sse4_2_exit)

Expand Down Expand Up @@ -1195,6 +1203,7 @@ LABEL(loop_ashr_9_use_sse4_2):
add $16, %r10
jg LABEL(nibble_ashr_9_use_sse4_2)

LABEL(nibble_ashr_9_use_sse4_2_restart):
movdqa (%rdi, %rdx), %xmm0

palignr $9, -16(%rdi, %rdx), %xmm0
Expand Down Expand Up @@ -1243,7 +1252,7 @@ LABEL(nibble_ashr_9_use_sse4_2):
jae LABEL(nibble_ashr_use_sse4_2_exit)
# endif
cmp $6, %ecx
ja LABEL(loop_ashr_9_use_sse4_2)
ja LABEL(nibble_ashr_9_use_sse4_2_restart)

jmp LABEL(nibble_ashr_use_sse4_2_exit)

Expand Down Expand Up @@ -1289,6 +1298,7 @@ LABEL(loop_ashr_10_use_sse4_2):
add $16, %r10
jg LABEL(nibble_ashr_10_use_sse4_2)

LABEL(nibble_ashr_10_use_sse4_2_restart):
movdqa (%rdi, %rdx), %xmm0
palignr $10, -16(%rdi, %rdx), %xmm0
# if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
Expand Down Expand Up @@ -1336,7 +1346,7 @@ LABEL(nibble_ashr_10_use_sse4_2):
jae LABEL(nibble_ashr_use_sse4_2_exit)
# endif
cmp $5, %ecx
ja LABEL(loop_ashr_10_use_sse4_2)
ja LABEL(nibble_ashr_10_use_sse4_2_restart)

jmp LABEL(nibble_ashr_use_sse4_2_exit)

Expand Down Expand Up @@ -1382,6 +1392,7 @@ LABEL(loop_ashr_11_use_sse4_2):
add $16, %r10
jg LABEL(nibble_ashr_11_use_sse4_2)

LABEL(nibble_ashr_11_use_sse4_2_restart):
movdqa (%rdi, %rdx), %xmm0
palignr $11, -16(%rdi, %rdx), %xmm0
# if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
Expand Down Expand Up @@ -1429,7 +1440,7 @@ LABEL(nibble_ashr_11_use_sse4_2):
jae LABEL(nibble_ashr_use_sse4_2_exit)
# endif
cmp $4, %ecx
ja LABEL(loop_ashr_11_use_sse4_2)
ja LABEL(nibble_ashr_11_use_sse4_2_restart)

jmp LABEL(nibble_ashr_use_sse4_2_exit)

Expand Down Expand Up @@ -1475,6 +1486,7 @@ LABEL(loop_ashr_12_use_sse4_2):
add $16, %r10
jg LABEL(nibble_ashr_12_use_sse4_2)

LABEL(nibble_ashr_12_use_sse4_2_restart):
movdqa (%rdi, %rdx), %xmm0
palignr $12, -16(%rdi, %rdx), %xmm0
# if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
Expand Down Expand Up @@ -1522,7 +1534,7 @@ LABEL(nibble_ashr_12_use_sse4_2):
jae LABEL(nibble_ashr_use_sse4_2_exit)
# endif
cmp $3, %ecx
ja LABEL(loop_ashr_12_use_sse4_2)
ja LABEL(nibble_ashr_12_use_sse4_2_restart)

jmp LABEL(nibble_ashr_use_sse4_2_exit)

Expand Down Expand Up @@ -1569,6 +1581,7 @@ LABEL(loop_ashr_13_use_sse4_2):
add $16, %r10
jg LABEL(nibble_ashr_13_use_sse4_2)

LABEL(nibble_ashr_13_use_sse4_2_restart):
movdqa (%rdi, %rdx), %xmm0
palignr $13, -16(%rdi, %rdx), %xmm0
# if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
Expand Down Expand Up @@ -1616,7 +1629,7 @@ LABEL(nibble_ashr_13_use_sse4_2):
jae LABEL(nibble_ashr_use_sse4_2_exit)
# endif
cmp $2, %ecx
ja LABEL(loop_ashr_13_use_sse4_2)
ja LABEL(nibble_ashr_13_use_sse4_2_restart)

jmp LABEL(nibble_ashr_use_sse4_2_exit)

Expand Down Expand Up @@ -1663,6 +1676,7 @@ LABEL(loop_ashr_14_use_sse4_2):
add $16, %r10
jg LABEL(nibble_ashr_14_use_sse4_2)

LABEL(nibble_ashr_14_use_sse4_2_restart):
movdqa (%rdi, %rdx), %xmm0
palignr $14, -16(%rdi, %rdx), %xmm0
# if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
Expand Down Expand Up @@ -1710,7 +1724,7 @@ LABEL(nibble_ashr_14_use_sse4_2):
jae LABEL(nibble_ashr_use_sse4_2_exit)
# endif
cmp $1, %ecx
ja LABEL(loop_ashr_14_use_sse4_2)
ja LABEL(nibble_ashr_14_use_sse4_2_restart)

jmp LABEL(nibble_ashr_use_sse4_2_exit)

Expand Down Expand Up @@ -1759,6 +1773,7 @@ LABEL(loop_ashr_15_use_sse4_2):
add $16, %r10
jg LABEL(nibble_ashr_15_use_sse4_2)

LABEL(nibble_ashr_15_use_sse4_2_restart):
movdqa (%rdi, %rdx), %xmm0
palignr $15, -16(%rdi, %rdx), %xmm0
# if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
Expand Down Expand Up @@ -1806,7 +1821,7 @@ LABEL(nibble_ashr_15_use_sse4_2):
jae LABEL(nibble_ashr_use_sse4_2_exit)
# endif
cmp $0, %ecx
ja LABEL(loop_ashr_15_use_sse4_2)
ja LABEL(nibble_ashr_15_use_sse4_2_restart)

LABEL(nibble_ashr_use_sse4_2_exit):
# if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
Expand Down

0 comments on commit c97a128

Please sign in to comment.