usarmyresearchlab / openshmem-epiphany Goto Github PK

Hi there
Once I intend to broadcast a variable in attached program called Td0 which is an unsigned int array containing 256 elements, I have faced a problem in the output (after broadcasting ). The last 8 elements of mentioned array did copy wrongly, the only solution which I found is to pass the number of elements more than 256 ( actual size) to 256+8!! I can not understand what is happened.
Kindly find the enclosed.
aesd.c.txt
aesd_epi_helper.c.txt
aesd_epi_helper.h.txt
device.c.txt
Makefile.txt

Hi there,
I am trying to use shmem_put() to realize the same functionality with shmem_broadcast(), because I'm interested in the time difference in between.

But the code I use below doesn't seem to get a correct answer, for core0, the rk[odd] are all right, but the rk[even] are not right, do I miss something?

#include <shmem.h>
long pSyncA[SHMEM_REDUCE_SYNC_SIZE] = { SHMEM_SYNC_VALUE };

int main(void)
{
	shmem_init(); 
	int my_pe = shmem_my_pe(); 
	int n_pes = shmem_n_pes(); 

     // vector for broadcast 
	int* rk = (int*) 0x5000;

    // init vector value for every core
	for (i=0; i < 16; i++){
		if (my_pe == i) rk[i] = i; 
		else rk[i] = 88;
	}

    shmem_barrier_all();

    // broadcast use shmem_put32()   $# can't get crrect answer for all core's rk array
    for(unsigned j = 0; j < 16; j++){
		if(my_pe == j){
			for(unsigned i = 0; i < 16; i++){
				if(i != j) shmem_put32(&rk[j], &rk[j], 1, i);	
			}
		}
    }
    shmem_barrier_all();

    shmem_finalize();
}

And by the way, I am using Parallella Board, Epiphany III 16 core device with the Parabuntu 2016.11 release.

Hi! I'm using this project in a research. It will be cool if Makefiles use $(gcc) instead of gcc. This will allow to setup any gcc compiler.
Changing this:
$(gcc) -O3 $< -o $@ $(EINCS) $(ELIBS) -lpthread
To:
gcc -O3 $< -o $@ $(EINCS) $(ELIBS) -lpthread

In every makefile.

EDIT:
This was done in every example but in e_hello.

Thanks!

Dear Admin

I am trying to do all to all communication in my device program written for Epiphany.
Following is the initialization code:
//id stands for the thread ID.
for (j = 0; j < NUM_ROWS; ++j)
{
offsetj = id*NUM_ROWS + j;
local_dataIn[offsetj].real = j;
local_dataIn[offsetj].img = offsetj;
}
shmem_alltoall64(local_dataIn, local_dataIn, NUM_ROWS, 0, 0, npes, pSyncA);

While doing this only the first NUM_ROWS indices are copied correctly. Is there any thing wrong in the way I am using this function?

Any help appreciated!