rs6000-string.c (do_load_for_compare_from_addr): New function.
2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com> * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New function. (do_ifelse): New function. (do_isel): New function. (do_sub3): New function. (do_add3): New function. (do_load_mask_compare): New function. (do_overlap_load_compare): New function. (expand_compare_loop): New function. (expand_block_compare): Call expand_compare_loop() when appropriate. * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change option description. (-mblock-compare-inline-loop-limit): New option. From-SVN: r256351
This commit is contained in:
Aaron Sawdey
parent
5a2a87e1e2
commit
5ec3397e38
4 changed files with 1118 additions and 34 deletions
File diff suppressed because it is too large
Load diff
|
|
@ -331,8 +331,12 @@ Target Report Var(rs6000_block_move_inline_limit) Init(0) RejectNegative Joined
|
|||
Specify how many bytes should be moved inline before calling out to memcpy/memmove.
|
||||
|
||||
mblock-compare-inline-limit=
|
||||
Target Report Var(rs6000_block_compare_inline_limit) Init(5) RejectNegative Joined UInteger Save
|
||||
Specify the maximum number pairs of load instructions that should be generated inline for the compare. If the number needed exceeds the limit, a call to memcmp will be generated instead.
|
||||
Target Report Var(rs6000_block_compare_inline_limit) Init(31) RejectNegative Joined UInteger Save
|
||||
Specify the maximum number of bytes to compare inline with non-looping code. If this is set to 0, all inline expansion (non-loop and loop) of memcmp is disabled.
|
||||
|
||||
mblock-compare-inline-loop-limit=
|
||||
Target Report Var(rs6000_block_compare_inline_loop_limit) Init(-1) RejectNegative Joined UInteger Save
|
||||
Specify the maximum number of bytes to compare inline with loop code generation. If the length is not known at compile time, memcmp will be called after this many bytes are compared. By default, a length will be picked depending on the tuning target.
|
||||
|
||||
mstring-compare-inline-limit=
|
||||
Target Report Var(rs6000_string_compare_inline_limit) Init(8) RejectNegative Joined UInteger Save
|
||||
|
|
|
|||
|
|
@ -14,11 +14,80 @@ int lib_strncmp(const char *a, const char *b, size_t n) asm("strncmp");
|
|||
#ifndef NRAND
|
||||
#define NRAND 10000
|
||||
#endif
|
||||
#define MAX_SZ 200
|
||||
#define MAX_SZ 600
|
||||
|
||||
#define DEF_RS(ALIGN) \
|
||||
static void test_memcmp_runtime_size_ ## ALIGN (const char *str1, \
|
||||
const char *str2, \
|
||||
size_t sz, int expect) \
|
||||
{ \
|
||||
char three[8192] __attribute__ ((aligned (4096))); \
|
||||
char four[8192] __attribute__ ((aligned (4096))); \
|
||||
char *a, *b; \
|
||||
int i,j,a1,a2,r; \
|
||||
for (j = 0; j < 2; j++) \
|
||||
{ \
|
||||
for (i = 0; i < 2; i++) \
|
||||
{ \
|
||||
a = three+i*ALIGN+j*(4096-2*i*ALIGN); \
|
||||
b = four+i*ALIGN+j*(4096-2*i*ALIGN); \
|
||||
memcpy(a,str1,sz); \
|
||||
memcpy(b,str2,sz); \
|
||||
asm(" "); \
|
||||
r = memcmp(a,b,sz); \
|
||||
asm(" "); \
|
||||
if ( r < 0 && !(expect < 0) ) abort(); \
|
||||
if ( r > 0 && !(expect > 0) ) abort(); \
|
||||
if ( r == 0 && !(expect == 0) ) abort(); \
|
||||
} \
|
||||
} \
|
||||
}
|
||||
|
||||
DEF_RS(1)
|
||||
DEF_RS(2)
|
||||
DEF_RS(4)
|
||||
DEF_RS(8)
|
||||
DEF_RS(16)
|
||||
|
||||
static void test_memcmp_runtime_size (const char *str1, const char *str2,
|
||||
size_t sz, int expect)
|
||||
{
|
||||
char three[8192] __attribute__ ((aligned (4096)));
|
||||
char four[8192] __attribute__ ((aligned (4096)));
|
||||
char *a, *b;
|
||||
int i,j,a1,a2,r;
|
||||
test_memcmp_runtime_size_1 (str1,str2,sz,expect);
|
||||
test_memcmp_runtime_size_2 (str1,str2,sz,expect);
|
||||
test_memcmp_runtime_size_4 (str1,str2,sz,expect);
|
||||
test_memcmp_runtime_size_8 (str1,str2,sz,expect);
|
||||
test_memcmp_runtime_size_16 (str1,str2,sz,expect);
|
||||
for (j = 0; j < 2; j++)
|
||||
{
|
||||
for (i = 0; i < 2; i++)
|
||||
{
|
||||
for (a1=0; a1 < 2*sizeof(void *); a1++)
|
||||
{
|
||||
for (a2=0; a2 < 2*sizeof(void *); a2++)
|
||||
{
|
||||
a = three+i*a1+j*(4096-2*i*a1);
|
||||
b = four+i*a2+j*(4096-2*i*a2);
|
||||
memcpy(a,str1,sz);
|
||||
memcpy(b,str2,sz);
|
||||
asm(" ");
|
||||
r = memcmp(a,b,sz);
|
||||
asm(" ");
|
||||
if ( r < 0 && !(expect < 0) ) abort();
|
||||
if ( r > 0 && !(expect > 0) ) abort();
|
||||
if ( r == 0 && !(expect == 0) ) abort();
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
static void test_driver_memcmp (void (test_memcmp)(const char *, const char *, int),
|
||||
void (test_strncmp)(const char *, const char *, int),
|
||||
size_t sz, int align)
|
||||
size_t sz, int align)
|
||||
{
|
||||
char buf1[MAX_SZ*2+10],buf2[MAX_SZ*2+10];
|
||||
size_t test_sz = (sz<MAX_SZ)?sz:MAX_SZ;
|
||||
|
|
@ -35,11 +104,12 @@ static void test_driver_memcmp (void (test_memcmp)(const char *, const char *, i
|
|||
buf1[j] = rand() & 0xff;
|
||||
buf2[j] = rand() & 0xff;
|
||||
}
|
||||
e = lib_memcmp(buf1,buf2,sz);
|
||||
(*test_memcmp)(buf1,buf2,e);
|
||||
test_memcmp_runtime_size (buf1, buf2, sz, e);
|
||||
e = lib_strncmp(buf1,buf2,sz);
|
||||
(*test_strncmp)(buf1,buf2,e);
|
||||
}
|
||||
e = lib_memcmp(buf1,buf2,sz);
|
||||
(*test_memcmp)(buf1,buf2,e);
|
||||
e = lib_strncmp(buf1,buf2,sz);
|
||||
(*test_strncmp)(buf1,buf2,e);
|
||||
}
|
||||
for(diff_pos = ((test_sz>10)?(test_sz-10):0); diff_pos < test_sz+10; diff_pos++)
|
||||
for(zero_pos = ((test_sz>10)?(test_sz-10):0); zero_pos < test_sz+10; zero_pos++)
|
||||
|
|
@ -53,6 +123,9 @@ static void test_driver_memcmp (void (test_memcmp)(const char *, const char *, i
|
|||
(*test_memcmp)(buf1,buf2,e);
|
||||
(*test_memcmp)(buf2,buf1,-e);
|
||||
(*test_memcmp)(buf2,buf2,0);
|
||||
test_memcmp_runtime_size (buf1, buf2, sz, e);
|
||||
test_memcmp_runtime_size (buf2, buf1, sz, -e);
|
||||
test_memcmp_runtime_size (buf2, buf2, sz, 0);
|
||||
e = lib_strncmp(buf1,buf2,sz);
|
||||
(*test_strncmp)(buf1,buf2,e);
|
||||
(*test_strncmp)(buf2,buf1,-e);
|
||||
|
|
@ -61,6 +134,7 @@ static void test_driver_memcmp (void (test_memcmp)(const char *, const char *, i
|
|||
buf2[diff_pos] = 0;
|
||||
e = lib_memcmp(buf1,buf2,sz);
|
||||
(*test_memcmp)(buf1,buf2,e);
|
||||
test_memcmp_runtime_size (buf1, buf2, sz, e);
|
||||
e = lib_strncmp(buf1,buf2,sz);
|
||||
(*test_strncmp)(buf1,buf2,e);
|
||||
memset(buf2+diff_pos,'B',sizeof(buf2)-diff_pos);
|
||||
|
|
@ -68,6 +142,8 @@ static void test_driver_memcmp (void (test_memcmp)(const char *, const char *, i
|
|||
e = lib_memcmp(buf1,buf2,sz);
|
||||
(*test_memcmp)(buf1,buf2,e);
|
||||
(*test_memcmp)(buf2,buf1,-e);
|
||||
test_memcmp_runtime_size (buf1, buf2, sz, e);
|
||||
test_memcmp_runtime_size (buf2, buf1, sz, -e);
|
||||
e = lib_strncmp(buf1,buf2,sz);
|
||||
(*test_strncmp)(buf1,buf2,e);
|
||||
(*test_strncmp)(buf2,buf1,-e);
|
||||
|
|
@ -371,7 +447,14 @@ DEF_TEST(100,2)
|
|||
DEF_TEST(100,4)
|
||||
DEF_TEST(100,8)
|
||||
DEF_TEST(100,16)
|
||||
DEF_TEST(191,1)
|
||||
DEF_TEST(192,1)
|
||||
DEF_TEST(193,1)
|
||||
DEF_TEST(200,1)
|
||||
DEF_TEST(400,1)
|
||||
#else
|
||||
DEF_TEST(1,1)
|
||||
DEF_TEST(2,1)
|
||||
DEF_TEST(3,1)
|
||||
DEF_TEST(4,1)
|
||||
DEF_TEST(5,1)
|
||||
|
|
@ -389,13 +472,15 @@ DEF_TEST(16,1)
|
|||
DEF_TEST(32,1)
|
||||
DEF_TEST(100,1)
|
||||
DEF_TEST(100,8)
|
||||
DEF_TEST(180,1)
|
||||
DEF_TEST(180,8)
|
||||
#endif
|
||||
|
||||
int
|
||||
main(int argc, char **argv)
|
||||
{
|
||||
#ifdef TEST_ALL
|
||||
RUN_TEST(1,1)
|
||||
RUN_TEST(1,1)
|
||||
RUN_TEST(1,2)
|
||||
RUN_TEST(1,4)
|
||||
RUN_TEST(1,8)
|
||||
|
|
@ -645,7 +730,14 @@ main(int argc, char **argv)
|
|||
RUN_TEST(100,4)
|
||||
RUN_TEST(100,8)
|
||||
RUN_TEST(100,16)
|
||||
RUN_TEST(191,1)
|
||||
RUN_TEST(192,1)
|
||||
RUN_TEST(193,1)
|
||||
RUN_TEST(200,1)
|
||||
RUN_TEST(400,1)
|
||||
#else
|
||||
RUN_TEST(1,1)
|
||||
RUN_TEST(2,1)
|
||||
RUN_TEST(3,1)
|
||||
RUN_TEST(4,1)
|
||||
RUN_TEST(5,1)
|
||||
|
|
@ -663,5 +755,7 @@ main(int argc, char **argv)
|
|||
RUN_TEST(32,1)
|
||||
RUN_TEST(100,1)
|
||||
RUN_TEST(100,8)
|
||||
RUN_TEST(180,1)
|
||||
RUN_TEST(180,8)
|
||||
#endif
|
||||
}
|
||||
|
|
|
|||
|
|
@ -81,6 +81,15 @@ DEF_TEST(13)
|
|||
DEF_TEST(14)
|
||||
DEF_TEST(15)
|
||||
DEF_TEST(16)
|
||||
DEF_TEST(32)
|
||||
DEF_TEST(64)
|
||||
DEF_TEST(65)
|
||||
DEF_TEST(66)
|
||||
DEF_TEST(67)
|
||||
DEF_TEST(68)
|
||||
DEF_TEST(69)
|
||||
DEF_TEST(70)
|
||||
DEF_TEST(71)
|
||||
|
||||
int
|
||||
main(int argc, char **argv)
|
||||
|
|
@ -101,5 +110,14 @@ main(int argc, char **argv)
|
|||
RUN_TEST(14);
|
||||
RUN_TEST(15);
|
||||
RUN_TEST(16);
|
||||
RUN_TEST(32);
|
||||
RUN_TEST(64);
|
||||
RUN_TEST(65);
|
||||
RUN_TEST(66);
|
||||
RUN_TEST(67);
|
||||
RUN_TEST(68);
|
||||
RUN_TEST(69);
|
||||
RUN_TEST(70);
|
||||
RUN_TEST(71);
|
||||
return 0;
|
||||
}
|
||||
|
|
|
|||
Loading…
Add table
Reference in a new issue