Re: i387 floating-point test program/benchmark

2005-08-03 Thread Jan Engelhardt 
>
> Among other issues.
>
> Therefore: add_tail(spare_time_Q);
>
> oddball:davidsen> cc -o i387bench -Os i387_bench.c
> i387_bench.c:27: parse error before `cpuset'
> i387_bench.c:27: warning: data definition has no type or storage class

This looks like an "old" glibc that does not have cpuset yet.

> i387_bench.c:34: unknown field `sa_handler' specified in initializer

Maybe a missing #include ?


Jan Engelhardt
-- 
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Re: i387 floating-point test program/benchmark

2005-08-03 Thread Jan Engelhardt 

 Among other issues.

 Therefore: add_tail(spare_time_Q);

 oddball:davidsen cc -o i387bench -Os i387_bench.c
 i387_bench.c:27: parse error before `cpuset'
 i387_bench.c:27: warning: data definition has no type or storage class

This looks like an old glibc that does not have cpuset yet.

 i387_bench.c:34: unknown field `sa_handler' specified in initializer

Maybe a missing #include signal.h?


Jan Engelhardt
-- 
-
To unsubscribe from this list: send the line unsubscribe linux-kernel in
the body of a message to [EMAIL PROTECTED]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Re: i387 floating-point test program/benchmark

2005-08-02 Thread Bill Davidsen 

Dave Jones wrote:

On Fri, Jul 29, 2005 at 04:36:05PM -0400, Chuck Ebbert wrote:
 >   memset(, sizeof(cpuset), 0);

This bug is like a disease, I swear.
(swapped args)


Among other issues.

Therefore: add_tail(spare_time_Q);


oddball:davidsen> cc -o i387bench -Os i387_bench.c
i387_bench.c:27: parse error before `cpuset'
i387_bench.c:27: warning: data definition has no type or storage class
i387_bench.c:34: unknown field `sa_handler' specified in initializer


--
   -bill davidsen ([EMAIL PROTECTED])
"The secret to procrastination is to put things off until the
 last possible moment - but no longer"  -me
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Re: i387 floating-point test program/benchmark

2005-08-02 Thread Bill Davidsen 

Dave Jones wrote:

On Fri, Jul 29, 2005 at 04:36:05PM -0400, Chuck Ebbert wrote:
memset(cpuset, sizeof(cpuset), 0);

This bug is like a disease, I swear.
(swapped args)


Among other issues.

Therefore: add_tail(spare_time_Q);


oddball:davidsen cc -o i387bench -Os i387_bench.c
i387_bench.c:27: parse error before `cpuset'
i387_bench.c:27: warning: data definition has no type or storage class
i387_bench.c:34: unknown field `sa_handler' specified in initializer


--
   -bill davidsen ([EMAIL PROTECTED])
The secret to procrastination is to put things off until the
 last possible moment - but no longer  -me
-
To unsubscribe from this list: send the line unsubscribe linux-kernel in
the body of a message to [EMAIL PROTECTED]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Re: i387 floating-point test program/benchmark

2005-07-29 Thread Con Kolivas 
On Sat, 30 Jul 2005 10:15, Puneet Vyas wrote:
> Chuck Ebbert wrote:
> >/* fp.c: i387 benchmark/test program */
>
> [EMAIL PROTECTED] C]$ cc FPUtest.c -o FPUtest
>
> FPUtest.c: In function `main':
>
> FPUtest.c:103: warning: passing arg 2 of `sched_setaffinity' makes
> integer from pointer without a cast
>
> FPUtest.c:103: error: too few arguments to function `sched_setaffinity'

There are half a dozen different versions of sched_setaffinity in glibc and 
they vary in argument type and even argument number across architectures.

Cheers,
Con


pgpuSk08N49hq.pgp
Description: PGP signature

Re: i387 floating-point test program/benchmark

2005-07-29 Thread Puneet Vyas 

Chuck Ebbert wrote:


/* fp.c: i387 benchmark/test program */
 



[EMAIL PROTECTED] C]$ cc FPUtest.c -o FPUtest

FPUtest.c: In function `main':

FPUtest.c:103: warning: passing arg 2 of `sched_setaffinity' makes 
integer from pointer without a cast


FPUtest.c:103: error: too few arguments to function `sched_setaffinity'


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Re: i387 floating-point test program/benchmark

2005-07-29 Thread Dave Jones 
On Fri, Jul 29, 2005 at 04:36:05PM -0400, Chuck Ebbert wrote:
 >  memset(, sizeof(cpuset), 0);

This bug is like a disease, I swear.
(swapped args)

Dave

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

i387 floating-point test program/benchmark

2005-07-29 Thread Chuck Ebbert 
/* fp.c: i387 benchmark/test program */

#define FP_VERSION "0.10"
#define _GNU_SOURCE
#include 
#include 
#include 
#include 
#include 
#include 
#include 
#include 

#define COND_YIELD(iters, ctr)  \
if ((iters) > 0 && ++(ctr) > (iters)) { \
(ctr) = 0;  \
sched_yield();  \
}
#define RDTSCLL(r) __asm__ __volatile__("rdtsc" : "=A" (r))

int fp = 0, in = 0, yi = 0;
volatile int lo = 0, lo2 = 0;
int p_ctr = 0, c_ctr = 0, get;
long long i, c_iters, p_iters = 0; 
long double c_res, p_res;
unsigned long long tsc1, tsc2;
cpu_set_t cpuset;

static void handler(int sig)
{
lo = 0;  /* child exited -- stop looping */
}
struct sigaction sa = {
.sa_handler = handler,
};

void usage(char *prog)
{
printf("\n i387 floating point benchmark/test program v" FP_VERSION);
printf("\n\n Usage:\n");
printf("\t%s [-f] [-i] [-a] [-y count] loops\n\n", prog);
printf("\t-f : loop in parent process waiting for child to exit\n");
printf("\t-i : do integer math while looping instead of FP math\n");
printf("\t (-f and -i: do one FP operation, then use integer 
math)\n");
printf("\t-a : run parent and child on single cpu\n");
printf("\t-y : yield CPU every 'count' loops in parent and child\n\n");
exit(1);
}

static void do_parent()
{
if (fp)
__asm__ __volatile__("fld1 ; fldz");

while (lo) {
if (fp & !in)
__asm__ __volatile__("fadd %st(1), %st(0)");
else
p_iters++;

COND_YIELD(yi, p_ctr);
}

if (fp)
__asm__ __volatile__("fxch ; fstpt p_res ; fstpt p_res");
if (fp & !in)
p_iters = (long long)p_res;

printf("Parent did: %12lld loops\n", p_iters);
}

static void do_child()
{
RDTSCLL(tsc1);
__asm__ __volatile__("fld1 ; fldz");
for (i = 0; i < c_iters; i++) {
__asm__ __volatile__("fadd %st(1), %st(0)");
COND_YIELD(yi, c_ctr);
}
__asm__ __volatile__("fxch ; fstpt c_res ; fstpt c_res");
RDTSCLL(tsc2);

if (c_res != (long double)c_iters)
printf("FP error! Result was %Lg; expected %lld\n", c_res, 
c_iters);

printf("CPU clocks: %12llu\n", tsc2 - tsc1);
}

int main(int argc, char * argv[])
{
do {
get = getopt(argc, argv, "fiay:");
switch (get) {
case 'f':
fp = 1;
break;
case 'i':
in = 1;
break;
case 'a':
memset(, sizeof(cpuset), 0);
CPU_SET(1, );
sched_setaffinity(0, );
break;
case 'y':
yi = atoi(optarg);
if (yi == 0)
get = '?';
break;
default:
break;
}

} while (get != -1 && get != '?');

if (get == '?' || optind >= argc)
usage(argv[0]);

c_iters = atoll(argv[optind]);

if (fp | in) {
sigaction(SIGCHLD, , NULL);
lo2 = lo = 1;
}

if (fork())
if (lo2)
do_parent();
else
wait(NULL);
else
do_child();

return 0;
}
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

i387 floating-point test program/benchmark

2005-07-29 Thread Chuck Ebbert 
/* fp.c: i387 benchmark/test program */

#define FP_VERSION 0.10
#define _GNU_SOURCE
#include sys/types.h
#include sys/wait.h
#include string.h
#include stdlib.h
#include unistd.h
#include stdio.h
#include signal.h
#include sched.h

#define COND_YIELD(iters, ctr)  \
if ((iters)  0  ++(ctr)  (iters)) { \
(ctr) = 0;  \
sched_yield();  \
}
#define RDTSCLL(r) __asm__ __volatile__(rdtsc : =A (r))

int fp = 0, in = 0, yi = 0;
volatile int lo = 0, lo2 = 0;
int p_ctr = 0, c_ctr = 0, get;
long long i, c_iters, p_iters = 0; 
long double c_res, p_res;
unsigned long long tsc1, tsc2;
cpu_set_t cpuset;

static void handler(int sig)
{
lo = 0;  /* child exited -- stop looping */
}
struct sigaction sa = {
.sa_handler = handler,
};

void usage(char *prog)
{
printf(\n i387 floating point benchmark/test program v FP_VERSION);
printf(\n\n Usage:\n);
printf(\t%s [-f] [-i] [-a] [-y count] loops\n\n, prog);
printf(\t-f : loop in parent process waiting for child to exit\n);
printf(\t-i : do integer math while looping instead of FP math\n);
printf(\t (-f and -i: do one FP operation, then use integer 
math)\n);
printf(\t-a : run parent and child on single cpu\n);
printf(\t-y : yield CPU every 'count' loops in parent and child\n\n);
exit(1);
}

static void do_parent()
{
if (fp)
__asm__ __volatile__(fld1 ; fldz);

while (lo) {
if (fp  !in)
__asm__ __volatile__(fadd %st(1), %st(0));
else
p_iters++;

COND_YIELD(yi, p_ctr);
}

if (fp)
__asm__ __volatile__(fxch ; fstpt p_res ; fstpt p_res);
if (fp  !in)
p_iters = (long long)p_res;

printf(Parent did: %12lld loops\n, p_iters);
}

static void do_child()
{
RDTSCLL(tsc1);
__asm__ __volatile__(fld1 ; fldz);
for (i = 0; i  c_iters; i++) {
__asm__ __volatile__(fadd %st(1), %st(0));
COND_YIELD(yi, c_ctr);
}
__asm__ __volatile__(fxch ; fstpt c_res ; fstpt c_res);
RDTSCLL(tsc2);

if (c_res != (long double)c_iters)
printf(FP error! Result was %Lg; expected %lld\n, c_res, 
c_iters);

printf(CPU clocks: %12llu\n, tsc2 - tsc1);
}

int main(int argc, char * argv[])
{
do {
get = getopt(argc, argv, fiay:);
switch (get) {
case 'f':
fp = 1;
break;
case 'i':
in = 1;
break;
case 'a':
memset(cpuset, sizeof(cpuset), 0);
CPU_SET(1, cpuset);
sched_setaffinity(0, cpuset);
break;
case 'y':
yi = atoi(optarg);
if (yi == 0)
get = '?';
break;
default:
break;
}

} while (get != -1  get != '?');

if (get == '?' || optind = argc)
usage(argv[0]);

c_iters = atoll(argv[optind]);

if (fp | in) {
sigaction(SIGCHLD, sa, NULL);
lo2 = lo = 1;
}

if (fork())
if (lo2)
do_parent();
else
wait(NULL);
else
do_child();

return 0;
}
-
To unsubscribe from this list: send the line unsubscribe linux-kernel in
the body of a message to [EMAIL PROTECTED]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Re: i387 floating-point test program/benchmark

2005-07-29 Thread Dave Jones 
On Fri, Jul 29, 2005 at 04:36:05PM -0400, Chuck Ebbert wrote:
   memset(cpuset, sizeof(cpuset), 0);

This bug is like a disease, I swear.
(swapped args)

Dave

-
To unsubscribe from this list: send the line unsubscribe linux-kernel in
the body of a message to [EMAIL PROTECTED]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Re: i387 floating-point test program/benchmark

2005-07-29 Thread Con Kolivas 
On Sat, 30 Jul 2005 10:15, Puneet Vyas wrote:
 Chuck Ebbert wrote:
 /* fp.c: i387 benchmark/test program */

 [EMAIL PROTECTED] C]$ cc FPUtest.c -o FPUtest

 FPUtest.c: In function `main':

 FPUtest.c:103: warning: passing arg 2 of `sched_setaffinity' makes
 integer from pointer without a cast

 FPUtest.c:103: error: too few arguments to function `sched_setaffinity'

There are half a dozen different versions of sched_setaffinity in glibc and 
they vary in argument type and even argument number across architectures.

Cheers,
Con


pgpuSk08N49hq.pgp
Description: PGP signature