[v1] add intel restricted transactional memory test case.

Add test case for intel restrict transactional memory. compiled with
Intel ICC 15.0 as well as GCC 4.8. This test case can be used to test
intel RTM support in the target-i386 frontend.

Signed-off-by: Xin Tong <trent.tong@gmail.com>

I am planning to implement support for Intel RTM. similar to what is
done for PowerPC. we can default to fault (transaction abort) to the
fallback code path. Would like to check in this test case first.


On Thu, Feb 5, 2015 at 1:56 PM,  <trent.tong@gmail.com> wrote:
> Add test case for intel restrict transactional memory. compiled with
> Intel ICC 15.0 as well as GCC 4.8. This test case can be used to test
> intel RTM support in the target-i386 frontend.
>
> Signed-off-by: Xin Tong <trent.tong@gmail.com>
>
> diff --git a/tests/tcg/test-intelrtm.c b/tests/tcg/test-intelrtm.c
> new file mode 100644
> index 0000000..ca1ce16
> --- /dev/null
> +++ b/tests/tcg/test-intelrtm.c
> @@ -0,0 +1,86 @@
> +/****************************************************************************
> +This file contains an example of Intel Restricted Transactional Memory (RTM).
> +In the example multiple threads are created to simulatenously increment
> +elements in a shared array.
> +
> +This test case should be compiled with Intel ICC compiler as it uses ICC RTM
> +intrnsics.
> +****************************************************************************/
> +#include <stdio.h>
> +#include <pthread.h>
> +#include <immintrin.h>  /* _XBEGIN_START */
> +
> +#define ARRAY_SIZE 1024
> +#define TESTTHREAD_COUNT 32
> +/* lock for atomic access to the testarray */
> +static pthread_mutex_t arraylock;
> +static pthread_t *threadarray;
> +static int testarray[ARRAY_SIZE];
> +
> +static void increment_func(void *arg)
> +{
> +     int i;
> +     long int threadnum = (long int)arg;
> +     unsigned int status = _xbegin();
> +     if (status == _XBEGIN_STARTED) {
> +        /* test _xtest() and _xabort(). fallback path for odd-num thread */
> +        if (_xtest() && (threadnum % 2)) {
> +            _xabort(_XABORT_EXPLICIT);
> +        }
> +        for (i = 0; i < ARRAY_SIZE; ++i) {
> +            testarray[i]++;
> +        }
> +        _xend();
> +     } else {
> +         /* fallback path. hold array lock */
> +         pthread_mutex_lock(&arraylock);
> +         for (i = 0; i < ARRAY_SIZE; ++i) {
> +            testarray[i]++;
> +         }
> +         /* release array lock */
> +         pthread_mutex_unlock(&arraylock);
> +    }
> +}
> +
> +static void run_test(int threadcount)
> +{
> +    int i;
> +    /* create the test threads */
> +    for (i = 0; i < threadcount; ++i) {
> +        pthread_create(&threadarray[i], NULL,
> +                      (void *(*)(void *))increment_func, (void*)i);
> +    }
> +    /* wait for the test threads to finish */
> +    for (i = 0; i < threadcount; ++i) {
> +        pthread_join(threadarray[i], NULL);
> +    }
> +    /* print the end results of the array */
> +    for (i = 0; i < ARRAY_SIZE; ++i) {
> +        printf("testarray[%d] is %d\n", i, testarray[i]);
> +    }
> +}
> +
> +static void run_setup(int threadcount)
> +{
> +    /* allocate the pthread_t structures */
> +    threadarray = (pthread_t *) malloc(sizeof(pthread_t)*threadcount);
> +    /* initialize the arraylock mutex */
> +    if (pthread_mutex_init(&arraylock, NULL) != 0) {
> +        printf("arraylock mutex init failed. exiting ...\n");
> +        exit(-1);
> +    }
> +}
> +
> +static void run_cleanup(void)
> +{
> +    free(threadarray);
> +    pthread_mutex_destroy(&arraylock);
> +}
> +
> +int main(void)
> +{
> +    run_setup(TESTTHREAD_COUNT);
> +    run_test(TESTTHREAD_COUNT);
> +    run_cleanup();
> +    return 0;
> +}

Please someone review this patch.

Thanks,
Xin

On Thu, Feb 5, 2015 at 2:04 PM, Xin Tong <trent.tong@gmail.com> wrote:
> I am planning to implement support for Intel RTM. similar to what is
> done for PowerPC. we can default to fault (transaction abort) to the
> fallback code path. Would like to check in this test case first.
>
>
> On Thu, Feb 5, 2015 at 1:56 PM,  <trent.tong@gmail.com> wrote:
>> Add test case for intel restrict transactional memory. compiled with
>> Intel ICC 15.0 as well as GCC 4.8. This test case can be used to test
>> intel RTM support in the target-i386 frontend.
>>
>> Signed-off-by: Xin Tong <trent.tong@gmail.com>
>>
>> diff --git a/tests/tcg/test-intelrtm.c b/tests/tcg/test-intelrtm.c
>> new file mode 100644
>> index 0000000..ca1ce16
>> --- /dev/null
>> +++ b/tests/tcg/test-intelrtm.c
>> @@ -0,0 +1,86 @@
>> +/****************************************************************************
>> +This file contains an example of Intel Restricted Transactional Memory (RTM).
>> +In the example multiple threads are created to simulatenously increment
>> +elements in a shared array.
>> +
>> +This test case should be compiled with Intel ICC compiler as it uses ICC RTM
>> +intrnsics.
>> +****************************************************************************/
>> +#include <stdio.h>
>> +#include <pthread.h>
>> +#include <immintrin.h>  /* _XBEGIN_START */
>> +
>> +#define ARRAY_SIZE 1024
>> +#define TESTTHREAD_COUNT 32
>> +/* lock for atomic access to the testarray */
>> +static pthread_mutex_t arraylock;
>> +static pthread_t *threadarray;
>> +static int testarray[ARRAY_SIZE];
>> +
>> +static void increment_func(void *arg)
>> +{
>> +     int i;
>> +     long int threadnum = (long int)arg;
>> +     unsigned int status = _xbegin();
>> +     if (status == _XBEGIN_STARTED) {
>> +        /* test _xtest() and _xabort(). fallback path for odd-num thread */
>> +        if (_xtest() && (threadnum % 2)) {
>> +            _xabort(_XABORT_EXPLICIT);
>> +        }
>> +        for (i = 0; i < ARRAY_SIZE; ++i) {
>> +            testarray[i]++;
>> +        }
>> +        _xend();
>> +     } else {
>> +         /* fallback path. hold array lock */
>> +         pthread_mutex_lock(&arraylock);
>> +         for (i = 0; i < ARRAY_SIZE; ++i) {
>> +            testarray[i]++;
>> +         }
>> +         /* release array lock */
>> +         pthread_mutex_unlock(&arraylock);
>> +    }
>> +}
>> +
>> +static void run_test(int threadcount)
>> +{
>> +    int i;
>> +    /* create the test threads */
>> +    for (i = 0; i < threadcount; ++i) {
>> +        pthread_create(&threadarray[i], NULL,
>> +                      (void *(*)(void *))increment_func, (void*)i);
>> +    }
>> +    /* wait for the test threads to finish */
>> +    for (i = 0; i < threadcount; ++i) {
>> +        pthread_join(threadarray[i], NULL);
>> +    }
>> +    /* print the end results of the array */
>> +    for (i = 0; i < ARRAY_SIZE; ++i) {
>> +        printf("testarray[%d] is %d\n", i, testarray[i]);
>> +    }
>> +}
>> +
>> +static void run_setup(int threadcount)
>> +{
>> +    /* allocate the pthread_t structures */
>> +    threadarray = (pthread_t *) malloc(sizeof(pthread_t)*threadcount);
>> +    /* initialize the arraylock mutex */
>> +    if (pthread_mutex_init(&arraylock, NULL) != 0) {
>> +        printf("arraylock mutex init failed. exiting ...\n");
>> +        exit(-1);
>> +    }
>> +}
>> +
>> +static void run_cleanup(void)
>> +{
>> +    free(threadarray);
>> +    pthread_mutex_destroy(&arraylock);
>> +}
>> +
>> +int main(void)
>> +{
>> +    run_setup(TESTTHREAD_COUNT);
>> +    run_test(TESTTHREAD_COUNT);
>> +    run_cleanup();
>> +    return 0;
>> +}