SHMEM_FLOAT_IPUT(3)
okay bestie, sit down, hydrate, we're doing SHMEM_FLOAT_IPUT(3) β yes that's the Open MPI manual, section 3, and she is everything π
This page is basically the group chat for a whole squad of strided-put routines, and every single one of them is invited:
shmem_complex_iput(3)shmem_double_iput(3)shmem_float_iput(3)shmem_int_iput(3)shmem_integer_iput(3)shmem_iput4(3)shmem_iput8(3)shmem_iput32(3)shmem_iput64(3)shmem_iput128(3)shmem_logical_iput(3)shmem_long_iput(3)shmem_longdouble_iput(3)shmem_longlong_iput(3)shmem_real_iput(3)shmem_short_iput(3)
and what do they collectively do?? Transfer strided data to a specified processing element (PE). That's the whole personality. Iconic, honestly. π
β¨ SYNOPSIS β¨
π©΅ C or C++
#include <mpp/shmem.h>
void shmem_double_iput(double *target, const double *source,
ptrdiff_t tst, ptrdiff_t sst, size_t len, int pe)
void shmem_float_iput(float *target, const float *source,
ptrdiff_t tst, ptrdiff_t sst, size_t len, int pe)
void shmem_int_iput(int *target, const int *source,
ptrdiff_t tst, ptrdiff_t sst, size_t len, int pe)
void shmem_iput32(void *target, const void *source,
ptrdiff_t tst, ptrdiff_t sst, size_t len, int pe)
void shmem_iput64(void *target, const void *source,
ptrdiff_t tst, ptrdiff_t sst, size_t len, int pe)
void shmem_iput128(void *target, const void *source,
ptrdiff_t tst, ptrdiff_t sst, size_t len, int pe)
void shmem_long_iput(long *target, const long *source,
ptrdiff_t tst, ptrdiff_t sst, size_t len, int pe)
void shmem_longdouble_iput(long double *target,
const long double *source, ptrdiff_t tst, ptrdiff_t sst,
size_t len, int pe)
void shmem_longlong_iput(long long *target,
const long long *source, ptrdiff_t tst, ptrdiff_t sst,
size_t len, int pe)
void shmem_short_iput(short *target, const short *source,
ptrdiff_t tst, ptrdiff_t sst, size_t len, int pe)
π Fortran
INCLUDE "mpp/shmem.fh"
INTEGER tst, sst, len, pe
CALL SHMEM_COMPLEX_IPUT(target, source, tst, sst, len,
& pe)
CALL SHMEM_DOUBLE_IPUT(target, source, tst, sst, len,
& pe)
CALL SHMEM_INTEGER_IPUT(target, source, tst, sst, len,
& pe)
CALL SHMEM_IPUT4(target, source, tst, sst, len, pe)
CALL SHMEM_IPUT8(target, source, tst, sst, len, pe)
CALL SHMEM_IPUT32(target, source, tst, sst, len, pe)
CALL SHMEM_IPUT64(target, source, tst, sst, len, pe)
CALL SHMEM_IPUT128(target, source, tst, sst, len, pe)
CALL SHMEM_LOGICAL_IPUT(target, source, tst, sst, len,
& pe)
CALL SHMEM_REAL_IPUT(target, source, tst, sst, len, pe)
π DESCRIPTION π
so here's the tea: the shmem_iput routines read the elements of a local array (source) and write them to a remote array (target) on the PE indicated by pe. That's the slay. π¦
But listen, bestie, the timing is a whole thing: these routines return when the data has been copied out of the source array on the local PE, but not necessarily before the data has been delivered to the remote data object. So likeβ¦ your local end is done, but the remote end might still be getting ready. Don't assume otherwise, I'm begging. π
πΈ the arguments (aka the lore)
targetβ the array that gets updated on the remote PE. This data object must be remotely accessible. Non-negotiable.sourceβ the array holding the data being copied. Your local stash.tstβ the stride between consecutive elements of thetargetarray. It gets scaled by the element size of thetargetarray. A value of1means contiguous data.tstmust be of type integer, and if you're in Fortran it must be a default integer value.sstβ the stride between consecutive elements of thesourcearray. It gets scaled by the element size of thesourcearray. A value of1means contiguous data.sstmust be of type integer, and if you're in Fortran it must be a default integer value.lenβ number of elements in thetargetandsourcearrays.lenmust be of type integer, and if you're in Fortran it must be a constant, variable, or array element of default integer type.peβ PE number of the remote PE.pemust be of type integer, and if you're in Fortran it must be a constant, variable, or array element of default integer type.
π typing constraints (the vibe check)
The target and source data objects must conform to typing constraints, and they go a lil something like this:
shmem_iput32,shmem_iput4: any noncharacter type that has a storage size equal to 32 bits.shmem_iput64,shmem_iput8: any noncharacter type that has a storage size equal to 64 bits.shmem_iput128: any noncharacter type that has a storage size equal to 128 bits.shmem_short_iput: elements of typeshort.shmem_int_iput: elements of typeint.shmem_long_iput: elements of typelong.shmem_longlong_iput: elements of typelong long.shmem_float_iput: elements of typefloat.shmem_double_iput: elements of typedouble.shmem_longdouble_iput: elements of typelong double.SHMEM_COMPLEX_IPUT: elements of typecomplexof default size.SHMEM_DOUBLE_IPUT: (Fortran) elements of typedouble precision.SHMEM_INTEGER_IPUT: elements of typeinteger.SHMEM_LOGICAL_IPUT: elements of typelogical.SHMEM_REAL_IPUT: elements of typereal.SHMEM_LOGICAL_IPUT: elements of typelogical.SHMEM_REAL_IPUT: elements of typereal.
yes, SHMEM_LOGICAL_IPUT and SHMEM_REAL_IPUT show up twice in the source. we're keeping the receipts exactly as written, fr fr π«Ά
And if you're using Fortran, data types must be of default size. For example, a real variable must be declared as REAL, REAL*4 or REAL(KIND=4). No custom-sized chaos, bestie. π«
π«Ά NOTES π«Ά
Go read intro_shmem(3) for a definition of the term remotely accessible. That's the homework. It's giving required reading, but make it cute. πβ¨
π EXAMPLES π
okay so picture this β a simple shmem_long_iput example for C/C++ programs. (main character energy, let's go) π
#include <mpp/shmem.h>
main()
{
short source[10] = { 1, 2, 3, 4, 5,
6, 7, 8, 9, 10 };
static short target[10];
shmem_init();
if (shmem_my_pe() == 0) {
/* put 10 words into target on PE 1 */
shmem_short_iput(target, source, 1, 2, 5, 1);
}
shmem_barrier_all(); /* sync sender and receiver */
if (shmem_my_pe() == 1) {
shmem_udcflush(); /* not required on IRIX systems */
printf("target on PE %d is %d %d %d %d %d0, shmem_my_pe(),
(int)target[0], (int)target[1], (int)target[2],
(int)target[3], (int)target[4] );
}
shmem_barrier_all(); /* sync before exiting */
}
See also:
intro_shmem(3)shmem_iget(3)shmem_put(3)shmem_quiet(3)
π and that's the whole page, bestie! Copyright 2003-2026, The Open MPI Community β go off kings π β and this edition of the docs is dated Sep 10, 2026. stay remotely accessible, stay iconic, and close your transfers responsibly ππ¦β¨
π¦ where this page comes from
- ubuntu stonking: openmpi-doc 5.0.10-6 Β· other
- π girlified by deepseek-ai/DeepSeek-V4.1-Flash on : a language model's rewrite of the original, under its terms