AI/rocm-systems
2
0
Derivar 0
Código Questões Pedidos de integração Operações Pacotes Planeamentos Lançamentos Wiki Trabalho
Ficheiros
801d2c5012792d6f0886bb4349ec88fdcc12314d
rocm-systems/src/rocshmem_gpu.cpp
T
Aurelien Bouteiller 801d2c5012 Enable GDA+IPC (#249) 
* Enable GDA+IPC
Fix ROCSHMEM_DISABLE_IPC for both RO and GDA

* add more functionality to bootstrap class

we need a few more functions in the boostrap class to be able to fully
handle the rocshmem requirements:
 - add a function to return the list of local ranks
 - provide a groupAllgather operation which takes a vector of ranks
   participating
 - provide a groupAlltoall operation which takes a vector of ranks
   participating

Also, update the functionality of the gda-Alltoall and gda-Allreduce
operations to take advantage of these functions.

* ipc_policy adapted to use bootstrap groupallgather

* bugfix: there was a mistake in computing sendto in groupallgather

* bugfix: shm_size and shm_rank were set in a local variable rather than
the class member

* mpi-bootstrap: remove an unecessary allgather

---------

Co-authored-by: Edgar Gabriel <Edgar.Gabriel@amd.com>
2025-09-16 11:54:53 -04:00

1821 linhas
88 KiB
C++
Em bruto Responsabilidade Histórico

/******************************************************************************
* Copyright (c) Advanced Micro Devices, Inc. All rights reserved.
*
* SPDX-License-Identifier: MIT
*
* Permission is hereby granted, free of charge, to any person obtaining a copy
* of this software and associated documentation files (the "Software"), to
* deal in the Software without restriction, including without limitation the
* rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
* sell copies of the Software, and to permit persons to whom the Software is
* furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
* IN THE SOFTWARE.
*****************************************************************************/
/**
* @file rocshmem.cpp
* @brief Public header for rocSHMEM device and host libraries.
*
* This is the implementation for the public rocshmem.hpp header file. This
* guy just extracts the transport from the opaque public handles and delegates
* to the appropriate backend.
*
* The device-side delegation is nasty because we can't use polymorphism with
* our current shader compiler stack. Maybe one day.....
*
* TODO: Could probably autogenerate many of these functions from macros.
*
* TODO: Support runtime backend detection.
*
*/
#include <hip/hip_runtime.h>
#include <cstdlib>
#include "rocshmem/rocshmem_config.h" // NOLINT(build/include_subdir)
#include "rocshmem/rocshmem.hpp"
#include "backend_bc.hpp"
#include "context_incl.hpp"
#include "team.hpp"
#include "templates.hpp"
#include "util.hpp"
#if defined(USE_GDA)
#include "gda/context_gda_tmpl_device.hpp"
#elif defined(USE_RO)
#include "reverse_offload/context_ro_tmpl_device.hpp"
#elif defined(USE_IPC)
# if defined(ENABLE_IPC_BITCODE)
# include "ipc/backend_ipc.hpp"
# endif
#include "ipc/context_ipc_tmpl_device.hpp"
#else
#error "Select one backend among USE_RO, USE_IPC, USE_GDA"
#endif
/******************************************************************************
**************************** Device Vars And Init ****************************
*****************************************************************************/
namespace rocshmem {
__device__ rocshmem_ctx_t __attribute__((visibility("default"))) ROCSHMEM_CTX_DEFAULT{};
__constant__ Backend *device_backend_proxy;
#if defined(ENABLE_IPC_BITCODE)
typedef IPCContext ContextTy;
#else
typedef Context ContextTy;
#endif
__device__ void rocshmem_wg_init() {
int provided;
/*
* Non-threaded init is allowed to select any thread mode, so don't worry
* if provided is different.
*/
rocshmem_wg_init_thread(ROCSHMEM_THREAD_WG_FUNNELED, &provided);
}
__device__ void rocshmem_wg_init_thread([[maybe_unused]] int requested,
int *provided) {
rocshmem_query_thread(provided);
}
__device__ void rocshmem_query_thread(int *provided) {
#ifdef USE_THREADS
*provided = ROCSHMEM_THREAD_MULTIPLE;
#else
*provided = ROCSHMEM_THREAD_WG_FUNNELED;
#endif
}
__device__ void rocshmem_wg_finalize() {}
/******************************************************************************
* These host APIs use Device side symbol - ROCSHMEM_CTX_DEFAULT so it needs
* to stay here to avoid getting pulled into other places in compilation
******************************************************************************/
__host__ void * rocshmem_get_device_ctx() {
void *ctx = nullptr;
CHECK_HIP(hipMemcpyFromSymbol(&ctx, HIP_SYMBOL(ROCSHMEM_CTX_DEFAULT),
sizeof(rocshmem_ctx_t)));
return ctx;
}
/******************************************************************************
************************** Default Context Wrappers **************************
*****************************************************************************/
__device__ void rocshmem_putmem(void *dest, const void *source, size_t nelems,
int pe) {
rocshmem_ctx_putmem(ROCSHMEM_CTX_DEFAULT, dest, source, nelems, pe);
}
template <typename T>
__device__ void rocshmem_put(T *dest, const T *source, size_t nelems, int pe) {
rocshmem_put(ROCSHMEM_CTX_DEFAULT, dest, source, nelems, pe);
}
template <typename T>
__device__ void rocshmem_p(T *dest, T value, int pe) {
rocshmem_p(ROCSHMEM_CTX_DEFAULT, dest, value, pe);
}
template <typename T>
__device__ T rocshmem_g(const T *source, int pe) {
return rocshmem_g(ROCSHMEM_CTX_DEFAULT, source, pe);
}
__device__ void rocshmem_getmem(void *dest, const void *source, size_t nelems,
int pe) {
rocshmem_ctx_getmem(ROCSHMEM_CTX_DEFAULT, dest, source, nelems, pe);
}
template <typename T>
__device__ void rocshmem_get(T *dest, const T *source, size_t nelems, int pe) {
rocshmem_get(ROCSHMEM_CTX_DEFAULT, dest, source, nelems, pe);
}
__device__ void rocshmem_putmem_nbi(void *dest, const void *source,
size_t nelems, int pe) {
rocshmem_ctx_putmem_nbi(ROCSHMEM_CTX_DEFAULT, dest, source, nelems, pe);
}
template <typename T>
__device__ void rocshmem_put_nbi(T *dest, const T *source, size_t nelems,
int pe) {
rocshmem_put_nbi(ROCSHMEM_CTX_DEFAULT, dest, source, nelems, pe);
}
__device__ void rocshmem_getmem_nbi(void *dest, const void *source,
size_t nelems, int pe) {
rocshmem_ctx_getmem_nbi(ROCSHMEM_CTX_DEFAULT, dest, source, nelems, pe);
}
template <typename T>
__device__ void rocshmem_get_nbi(T *dest, const T *source, size_t nelems,
int pe) {
rocshmem_get_nbi(ROCSHMEM_CTX_DEFAULT, dest, source, nelems, pe);
}
__device__ void rocshmem_fence() {
rocshmem_ctx_fence(ROCSHMEM_CTX_DEFAULT);
}
__device__ void rocshmem_fence(int pe) {
rocshmem_ctx_fence(ROCSHMEM_CTX_DEFAULT, pe);
}
__device__ void rocshmem_quiet() {
rocshmem_ctx_quiet(ROCSHMEM_CTX_DEFAULT);
}
__device__ void rocshmem_threadfence_system() {
rocshmem_ctx_threadfence_system(ROCSHMEM_CTX_DEFAULT);
}
template <typename T>
__device__ T rocshmem_atomic_fetch_add(T *dest, T val, int pe) {
return rocshmem_atomic_fetch_add(ROCSHMEM_CTX_DEFAULT, dest, val, pe);
}
template <typename T>
__device__ T rocshmem_atomic_compare_swap(T *dest, T cond, T val, int pe) {
return rocshmem_atomic_compare_swap(ROCSHMEM_CTX_DEFAULT, dest, cond, val,
pe);
}
template <typename T>
__device__ T rocshmem_atomic_fetch_inc(T *dest, int pe) {
return rocshmem_atomic_fetch_inc(ROCSHMEM_CTX_DEFAULT, dest, pe);
}
template <typename T>
__device__ T rocshmem_atomic_fetch(T *source, int pe) {
return rocshmem_atomic_fetch(ROCSHMEM_CTX_DEFAULT, source, pe);
}
template <typename T>
__device__ void rocshmem_atomic_add(T *dest, T val, int pe) {
rocshmem_atomic_add(ROCSHMEM_CTX_DEFAULT, dest, val, pe);
}
template <typename T>
__device__ void rocshmem_atomic_inc(T *dest, int pe) {
rocshmem_atomic_inc(ROCSHMEM_CTX_DEFAULT, dest, pe);
}
template <typename T>
__device__ void rocshmem_atomic_set(T *dest, T value, int pe) {
rocshmem_atomic_set(ROCSHMEM_CTX_DEFAULT, dest, value, pe);
}
template <typename T>
__device__ T rocshmem_atomic_swap(T *dest, T value, int pe) {
return rocshmem_atomic_swap(ROCSHMEM_CTX_DEFAULT, dest, value, pe);
}
template <typename T>
__device__ T rocshmem_atomic_fetch_and(T *dest, T value, int pe) {
return rocshmem_atomic_fetch_and(ROCSHMEM_CTX_DEFAULT, dest, value, pe);
}
template <typename T>
__device__ void rocshmem_atomic_and(T *dest, T value, int pe) {
rocshmem_atomic_and(ROCSHMEM_CTX_DEFAULT, dest, value, pe);
}
template <typename T>
__device__ T rocshmem_atomic_fetch_or(T *dest, T value, int pe) {
return rocshmem_atomic_fetch_or(ROCSHMEM_CTX_DEFAULT, dest, value, pe);
}
template <typename T>
__device__ void rocshmem_atomic_or(T *dest, T value, int pe) {
rocshmem_atomic_or(ROCSHMEM_CTX_DEFAULT, dest, value, pe);
}
template <typename T>
__device__ T rocshmem_atomic_fetch_xor(T *dest, T value, int pe) {
return rocshmem_atomic_fetch_xor(ROCSHMEM_CTX_DEFAULT, dest, value, pe);
}
template <typename T>
__device__ void rocshmem_atomic_xor(T *dest, T value, int pe) {
rocshmem_atomic_xor(ROCSHMEM_CTX_DEFAULT, dest, value, pe);
}
#define ROCSHMEM_PUTMEM_SIGNAL_DEF(SUFFIX) \
__device__ void rocshmem_putmem_signal##SUFFIX(void *dest, const void *source, size_t nelems, \
uint64_t *sig_addr, uint64_t signal, \
int sig_op, int pe) { \
rocshmem_ctx_putmem_signal##SUFFIX(ROCSHMEM_CTX_DEFAULT, \
dest, source, nelems, \
sig_addr, signal, sig_op, pe); \
} \
\
template <typename T> \
__device__ void rocshmem_put_signal##SUFFIX(T *dest, const T *source, size_t nelems, \
uint64_t *sig_addr, uint64_t signal, \
int sig_op, int pe) { \
rocshmem_ctx_put_signal##SUFFIX(ROCSHMEM_CTX_DEFAULT, \
dest, source, nelems, \
sig_addr, signal, sig_op, pe); \
}
ROCSHMEM_PUTMEM_SIGNAL_DEF()
ROCSHMEM_PUTMEM_SIGNAL_DEF(_wg)
ROCSHMEM_PUTMEM_SIGNAL_DEF(_wave)
ROCSHMEM_PUTMEM_SIGNAL_DEF(_nbi)
ROCSHMEM_PUTMEM_SIGNAL_DEF(_nbi_wg)
ROCSHMEM_PUTMEM_SIGNAL_DEF(_nbi_wave)
/******************************************************************************
************************* Private Context Interfaces *************************
*****************************************************************************/
__device__ int translate_pe(rocshmem_ctx_t ctx, int pe) {
if (ctx.team_opaque) {
TeamInfo *tinfo = reinterpret_cast<TeamInfo *>(ctx.team_opaque);
return (tinfo->pe_start + tinfo->stride * pe);
} else {
return pe;
}
}
__host__ void set_internal_ctx(rocshmem_ctx_t *ctx) {
CHECK_HIP(hipMemcpyToSymbol(HIP_SYMBOL(ROCSHMEM_CTX_DEFAULT), ctx,
sizeof(rocshmem_ctx_t), 0,
hipMemcpyHostToDevice));
}
__device__ ContextTy *get_internal_ctx(rocshmem_ctx_t ctx) {
return reinterpret_cast<ContextTy *>(ctx.ctx_opaque);
}
__device__ int rocshmem_wg_ctx_create(long options, rocshmem_ctx_t *ctx) {
GPU_DPRINTF("Function: rocshmem_wg_ctx_create (options=%ld)\n", options);
bool result{true};
if (get_flat_block_id() == 0) {
ctx->team_opaque = reinterpret_cast<TeamInfo *>(ROCSHMEM_CTX_DEFAULT.team_opaque);
result = device_backend_proxy->create_ctx(options, ctx);
if(result) {
reinterpret_cast<Context *>(ctx->ctx_opaque)->setFence(options);
}
}
__syncthreads();
return result == true ? 0 : -1;
}
__device__ int rocshmem_wg_team_create_ctx(rocshmem_team_t team, long options,
rocshmem_ctx_t *ctx) {
GPU_DPRINTF("Function: rocshmem_wg_team_create_ctx (team=%zd, options=%ld)\n",
(intptr_t)team, options);
if (team == ROCSHMEM_TEAM_INVALID) {
return -1;
}
bool result{true};
if (get_flat_block_id() == 0) {
Team *team_obj{get_internal_team(team)};
TeamInfo *info_wrt_world = team_obj->tinfo_wrt_world;
ctx->team_opaque = info_wrt_world;
result = device_backend_proxy->create_ctx(options, ctx);
if(result) {
reinterpret_cast<Context *>(ctx->ctx_opaque)->setFence(options);
}
}
__syncthreads();
return result == true ? 0 : -1;
}
__device__ void rocshmem_wg_ctx_destroy(
[[maybe_unused]] rocshmem_ctx_t *ctx) {
GPU_DPRINTF("Function: rocshmem_wg_ctx_destroy (ctx=%zd)\n",
ctx->ctx_opaque);
if (get_flat_block_id() == 0) {
device_backend_proxy->destroy_ctx(ctx);
}
}
__device__ void rocshmem_ctx_threadfence_system(rocshmem_ctx_t ctx) {
GPU_DPRINTF("Function: rocshmem_ctx_threadfence_system (ctx=%zd)\n",
ctx.ctx_opaque);
get_internal_ctx(ctx)->threadfence_system();
}
__device__ void rocshmem_ctx_putmem(rocshmem_ctx_t ctx, void *dest,
const void *source, size_t nelems,
int pe) {
int pe_in_world = translate_pe(ctx, pe);
GPU_DPRINTF("Function: rocshmem_ctx_putmem (ctx=%zd, dest=%p, source=%p, nelems=%zd, pe=%d w%d)\n",
ctx.ctx_opaque, dest, source, nelems, pe, pe_in_world);
get_internal_ctx(ctx)->putmem(dest, source, nelems, pe_in_world);
}
template <typename T>
__device__ void rocshmem_put(rocshmem_ctx_t ctx, T *dest, const T *source,
size_t nelems, int pe) {
int pe_in_world = translate_pe(ctx, pe);
GPU_DPRINTF("Function: rocshmem_put (ctx=%zd, dest=%p, source=%p, nelems=%zd, pe=%d w%d)\n",
ctx.ctx_opaque, dest, source, nelems, pe, pe_in_world);
get_internal_ctx(ctx)->put(dest, source, nelems, pe_in_world);
}
template <typename T>
__device__ void rocshmem_p(rocshmem_ctx_t ctx, T *dest, T value, int pe) {
int pe_in_world = translate_pe(ctx, pe);
GPU_DPRINTF("Function: rocshmem_p (ctx=%zd, dest=%p, value=%g, pe=%d w%d)\n",
ctx.ctx_opaque, dest, value, pe, pe_in_world);
get_internal_ctx(ctx)->p(dest, value, pe_in_world);
}
template <typename T>
__device__ T rocshmem_g(rocshmem_ctx_t ctx, const T *source, int pe) {
int pe_in_world = translate_pe(ctx, pe);
GPU_DPRINTF("Function: rocshmem_g (ctx=%zd, source=%p, pe=%d w%d)\n",
ctx.ctx_opaque, source, pe, pe_in_world);
return get_internal_ctx(ctx)->g(source, pe_in_world);
}
__device__ void rocshmem_ctx_getmem(rocshmem_ctx_t ctx, void *dest,
const void *source, size_t nelems,
int pe) {
int pe_in_world = translate_pe(ctx, pe);
GPU_DPRINTF("Function: rocshmem_ctx_getmem (ctx=%zd, dest=%p, source=%p, nelems=%zd, pe=%d w%d)\n",
ctx.ctx_opaque, dest, source, nelems, pe, pe_in_world);
get_internal_ctx(ctx)->getmem(dest, source, nelems, pe_in_world);
}
template <typename T>
__device__ void rocshmem_get(rocshmem_ctx_t ctx, T *dest, const T *source,
size_t nelems, int pe) {
int pe_in_world = translate_pe(ctx, pe);
GPU_DPRINTF("Function: rocshmem_get (ctx=%zd, dest=%p, source=%p, nelems=%zd, pe=%d w%d)\n",
ctx.ctx_opaque, dest, source, nelems, pe, pe_in_world);
get_internal_ctx(ctx)->get(dest, source, nelems, pe_in_world);
}
__device__ void rocshmem_ctx_putmem_nbi(rocshmem_ctx_t ctx, void *dest,
const void *source, size_t nelems,
int pe) {
int pe_in_world = translate_pe(ctx, pe);
GPU_DPRINTF("Function: rocshmem_ctx_putmem_nbi (ctx=%zd, dest=%p, source=%p, nelems=%zd, pe=%d w%d)\n",
ctx.ctx_opaque, dest, source, nelems, pe, pe_in_world);
get_internal_ctx(ctx)->putmem_nbi(dest, source, nelems, pe_in_world);
}
template <typename T>
__device__ void rocshmem_put_nbi(rocshmem_ctx_t ctx, T *dest, const T *source,
size_t nelems, int pe) {
int pe_in_world = translate_pe(ctx, pe);
GPU_DPRINTF("Function: rocshmem_put_nbi (ctx=%zd, dest=%p, source=%p, nelems=%zd, pe=%d w%d)\n",
ctx.ctx_opaque, dest, source, nelems, pe, pe_in_world);
get_internal_ctx(ctx)->put_nbi(dest, source, nelems, pe_in_world);
}
__device__ void rocshmem_ctx_getmem_nbi(rocshmem_ctx_t ctx, void *dest,
const void *source, size_t nelems,
int pe) {
int pe_in_world = translate_pe(ctx, pe);
GPU_DPRINTF("Function: rocshmem_ctx_getmem_nbi (ctx=%zd, dest=%p, source=%p, nelems=%zd, pe=%d w%d)\n",
ctx.ctx_opaque, dest, source, nelems, pe, pe_in_world);
get_internal_ctx(ctx)->getmem_nbi(dest, source, nelems, pe_in_world);
}
template <typename T>
__device__ void rocshmem_get_nbi(rocshmem_ctx_t ctx, T *dest, const T *source,
size_t nelems, int pe) {
int pe_in_world = translate_pe(ctx, pe);
GPU_DPRINTF("Function: rocshmem_get_nbi (ctx=%zd, dest=%p, source=%p, nelems=%zd, pe=%d w%d)\n",
ctx.ctx_opaque, dest, source, nelems, pe, pe_in_world);
get_internal_ctx(ctx)->get_nbi(dest, source, nelems, pe_in_world);
}
__device__ void rocshmem_ctx_fence(rocshmem_ctx_t ctx) {
GPU_DPRINTF("Function: rocshmem_ctx_fence (ctx=%zd)\n", ctx.ctx_opaque);
get_internal_ctx(ctx)->fence();
}
__device__ void rocshmem_ctx_fence(rocshmem_ctx_t ctx, int pe) {
int pe_in_world = translate_pe(ctx, pe);
GPU_DPRINTF("Function: rocshmem_ctx_fence (ctx=%zd, pe=%d w%d))\n",
ctx.ctx_opaque, pe, pe_in_world);
get_internal_ctx(ctx)->fence(pe_in_world);
}
__device__ void rocshmem_ctx_quiet(rocshmem_ctx_t ctx) {
GPU_DPRINTF("Function: rocshmem_ctx_quiet (ctx=%zd)\n",
ctx.ctx_opaque);
get_internal_ctx(ctx)->quiet();
}
__device__ void *rocshmem_ptr(const void *dest, int pe) {
GPU_DPRINTF("Function: rocshmem_ptr (dest=%p, pe=%d w%d\n",
dest, pe, pe);
return get_internal_ctx(ROCSHMEM_CTX_DEFAULT)->shmem_ptr(dest, pe);
}
template <typename T, ROCSHMEM_OP Op>
__device__ int rocshmem_reduce_wg(rocshmem_ctx_t ctx, rocshmem_team_t team,
T *dest, const T *source, int nreduce) {
GPU_DPRINTF("Function: rocshmem_reduce_wg (ctx=%zd, team=%zd, dest=%p, source=%p, nreduce=%d\n",
ctx.ctx_opaque, team, dest, source, nreduce);
return get_internal_ctx(ctx)->reduce<T, Op>(team, dest, source, nreduce);
}
template <typename T>
__device__ void rocshmem_broadcast_wg(rocshmem_ctx_t ctx,
rocshmem_team_t team, T *dest,
const T *source, int nelem,
int pe_root) {
GPU_DPRINTF("Function: Team-based rocshmem_broadcast_wg (ctx=%zd, team=%zd, dest=%p, source=%p, nelem=%d, root=%d)\n",
ctx.ctx_opaque, team, dest, source, nelem, pe_root);
get_internal_ctx(ctx)->broadcast<T>(team, dest, source, nelem, pe_root);
}
template <typename T>
__device__ void rocshmem_alltoall_wg(rocshmem_ctx_t ctx,
rocshmem_team_t team, T *dest,
const T *source, int nelem) {
GPU_DPRINTF("Function: rocshmem_alltoall_wg (ctx=%zd, team=%zd, dest=%p, source=%p, nelem=%d\n",
ctx.ctx_opaque, team, dest, source, nelem);
get_internal_ctx(ctx)->alltoall<T>(team, dest, source, nelem);
}
template <typename T>
__device__ void rocshmem_fcollect_wg(rocshmem_ctx_t ctx,
rocshmem_team_t team, T *dest,
const T *source, int nelem) {
GPU_DPRINTF("Function: rocshmem_fcollect_wg (ctx=%zd, team=%zd, dest=%p, source=%p, nelem=%d\n",
ctx.ctx_opaque, team, dest, source, nelem);
get_internal_ctx(ctx)->fcollect<T>(team, dest, source, nelem);
}
template <typename T>
__device__ void rocshmem_wait_until(T *ivars, int cmp, T val) {
GPU_DPRINTF("Function: rocshmem_wait_until (ivars=%p, cmp=%d, val=%g)\n",
ivars, cmp, (double)val);
Context *ctx_internal = get_internal_ctx(ROCSHMEM_CTX_DEFAULT);
ctx_internal->ctxStats.incStat(NUM_WAIT_UNTIL);
ctx_internal->wait_until(ivars, cmp, val);
}
template <typename T>
__device__ void rocshmem_wait_until_all(T *ivars, size_t nelems, const int* status,
int cmp, T val) {
GPU_DPRINTF("Function: rocshmem_wait_until_all (ivars=%p, nelems=%zd cmp=%d, val=%g)\n",
ivars, nelems, cmp, (double)val);
Context *ctx_internal = get_internal_ctx(ROCSHMEM_CTX_DEFAULT);
ctx_internal->ctxStats.incStat(NUM_WAIT_UNTIL_ALL);
ctx_internal->wait_until_all(ivars, nelems, status, cmp, val);
}
template <typename T>
__device__ size_t rocshmem_wait_until_any(T *ivars, size_t nelems, const int* status,
int cmp, T val) {
GPU_DPRINTF("Function: rocshmem_wait_until_any (ivars=%p, nelems=%zd cmp=%d, val=%g)\n",
ivars, nelems, cmp, (double)val);
Context *ctx_internal = get_internal_ctx(ROCSHMEM_CTX_DEFAULT);
ctx_internal->ctxStats.incStat(NUM_WAIT_UNTIL_ANY);
return ctx_internal->wait_until_any(ivars, nelems, status, cmp, val);
}
template <typename T>
__device__ size_t rocshmem_wait_until_some(T *ivars, size_t nelems, size_t* indices,
const int* status, int cmp,
T val) {
DPRINTF("Function: rocshmem_wait_until_some (ivars=%p, nelems=%zd cmp=%d, val=%g)\n",
ivars, nelems, cmp, (double)val);
Context *ctx_internal = get_internal_ctx(ROCSHMEM_CTX_DEFAULT);
ctx_internal->ctxStats.incStat(NUM_WAIT_UNTIL_SOME);
return ctx_internal->wait_until_some(ivars, nelems, indices, status, cmp, val);
}
template <typename T>
__device__ size_t rocshmem_wait_until_any_vector(T *ivars, size_t nelems, const int* status,
int cmp, T* vals) {
DPRINTF("Function: rocshmem_wait_until_any_vector (ivars=%p, nelems=%zd cmp=%d, vals=%p)\n",
ivars, nelems, cmp, vals);
Context *ctx_internal = get_internal_ctx(ROCSHMEM_CTX_DEFAULT);
ctx_internal->ctxStats.incStat(NUM_WAIT_UNTIL_ANY_VECTOR);
return ctx_internal->wait_until_any_vector(ivars, nelems, status, cmp, vals);
}
template <typename T>
__device__ void rocshmem_wait_until_all_vector(T *ivars, size_t nelems, const int* status,
int cmp, T* vals) {
DPRINTF("Function: rocshmem_wait_until_all_vector (ivars=%p, nelems=%zd cmp=%d, vals=%p)\n",
ivars, nelems, cmp, vals);
Context *ctx_internal = get_internal_ctx(ROCSHMEM_CTX_DEFAULT);
ctx_internal->ctxStats.incStat(NUM_WAIT_UNTIL_ALL_VECTOR);
ctx_internal->wait_until_all_vector(ivars, nelems, status, cmp, vals);
}
template <typename T>
__device__ size_t rocshmem_wait_until_some_vector(T *ivars, size_t nelems,
size_t* indices,
const int* status,
int cmp, T* vals) {
DPRINTF("Function: rocshmem_wait_until_some_vector (ivars=%p, nelems=%zd cmp=%d, vals=%p)\n",
ivars, nelems, cmp, vals);
Context *ctx_internal = get_internal_ctx(ROCSHMEM_CTX_DEFAULT);
ctx_internal->ctxStats.incStat(NUM_WAIT_UNTIL_SOME_VECTOR);
return ctx_internal->wait_until_some_vector(ivars, nelems, indices, status, cmp, vals);
}
template <typename T>
__device__ int rocshmem_test(T *ivars, int cmp, T val) {
GPU_DPRINTF("Function: rocshmem_test (ivars=%p, cmp=%d, val=%g)\n",
ivars, cmp, (double)val);
Context *ctx_internal = get_internal_ctx(ROCSHMEM_CTX_DEFAULT);
ctx_internal->ctxStats.incStat(NUM_TEST);
return ctx_internal->test(ivars, cmp, val);
}
__device__ void rocshmem_barrier_all() {
GPU_DPRINTF("Function: rocshmem_barrier_all (ctx=%zd)\n",
get_internal_ctx(ROCSHMEM_CTX_DEFAULT));
get_internal_ctx(ROCSHMEM_CTX_DEFAULT)->barrier_all();
}
__device__ void rocshmem_barrier_all_wave() {
GPU_DPRINTF("Function: rocshmem_barrier_all_wave (ctx=%zd)\n",
get_internal_ctx(ROCSHMEM_CTX_DEFAULT));
get_internal_ctx(ROCSHMEM_CTX_DEFAULT)->barrier_all_wave();
}
__device__ void rocshmem_barrier_all_wg() {
GPU_DPRINTF("Function: rocshmem_barrier_all_wg (ctx=%zd)\n",
get_internal_ctx(ROCSHMEM_CTX_DEFAULT));
get_internal_ctx(ROCSHMEM_CTX_DEFAULT)->barrier_all_wg();
}
__device__ void rocshmem_ctx_barrier(rocshmem_ctx_t ctx, rocshmem_team_t team) {
GPU_DPRINTF("Function: rocshmem_ctx_barrier (ctx=%zd, team=%zd)\n",
ctx.ctx_opaque, team);
get_internal_ctx(ctx)->barrier(team);
}
__device__ void rocshmem_ctx_barrier_wave(rocshmem_ctx_t ctx, rocshmem_team_t team) {
GPU_DPRINTF("Function: rocshmem_ctx_barrier_wave (ctx=%zd, team=%zd)\n",
ctx.ctx_opaque, team);
get_internal_ctx(ctx)->barrier_wave(team);
}
__device__ void rocshmem_ctx_barrier_wg(rocshmem_ctx_t ctx, rocshmem_team_t team) {
GPU_DPRINTF("Function: rocshmem_ctx_barrier_wg (ctx=%zd, team=%zd)\n",
ctx.ctx_opaque, team);
get_internal_ctx(ctx)->barrier_wg(team);
}
__device__ void rocshmem_sync_all() {
GPU_DPRINTF("Function: rocshmem_sync_all (ctx=%zd)\n",
get_internal_ctx(ROCSHMEM_CTX_DEFAULT));
get_internal_ctx(ROCSHMEM_CTX_DEFAULT)->sync_all();
}
__device__ void rocshmem_sync_all_wave() {
GPU_DPRINTF("Function: rocshmem_sync_all_wave (ctx=%zd)\n",
get_internal_ctx(ROCSHMEM_CTX_DEFAULT));
get_internal_ctx(ROCSHMEM_CTX_DEFAULT)->sync_all_wave();
}
__device__ void rocshmem_sync_all_wg() {
GPU_DPRINTF("Function: rocshmem_sync_all_wg (ctx=%zd)\n",
get_internal_ctx(ROCSHMEM_CTX_DEFAULT));
get_internal_ctx(ROCSHMEM_CTX_DEFAULT)->sync_all_wg();
}
__device__ void rocshmem_ctx_sync(rocshmem_ctx_t ctx,
rocshmem_team_t team) {
GPU_DPRINTF("Function: rocshmem_ctx_sync (ctx=%zd, team=%zd)\n",
ctx.ctx_opaque, team);
get_internal_ctx(ctx)->sync_wg(team);
}
__device__ void rocshmem_ctx_sync_wave(rocshmem_ctx_t ctx,
rocshmem_team_t team) {
GPU_DPRINTF("Function: rocshmem_ctx_sync_wave (ctx=%zd, team=%zd)\n",
ctx.ctx_opaque, team);
get_internal_ctx(ctx)->sync_wg(team);
}
__device__ void rocshmem_ctx_sync_wg(rocshmem_ctx_t ctx,
rocshmem_team_t team) {
GPU_DPRINTF("Function: rocshmem_ctx_sync_wg (ctx=%zd, team=%zd)\n",
ctx.ctx_opaque, team);
get_internal_ctx(ctx)->sync_wg(team);
}
__device__ int rocshmem_ctx_n_pes(rocshmem_ctx_t ctx) {
GPU_DPRINTF("Function: rocshmem_ctx_n_pes (ctx=%zd)\n",
ctx.ctx_opaque);
TeamInfo *tinfo = reinterpret_cast<TeamInfo *>(ctx.team_opaque);
return tinfo->size;
}
__device__ int rocshmem_n_pes() {
return get_internal_ctx(ROCSHMEM_CTX_DEFAULT)->num_pes;
}
__device__ int rocshmem_ctx_my_pe(rocshmem_ctx_t ctx) {
GPU_DPRINTF("Function: rocshmem_ctx_my_pe (ctx=%zd)\n",
ctx.ctx_opaque);
TeamInfo *tinfo = reinterpret_cast<TeamInfo *>(ctx.team_opaque);
int my_pe{get_internal_ctx(ctx)->my_pe};
int pe_start{tinfo->pe_start};
int stride{tinfo->stride};
int size{tinfo->size};
int translated_pe = (my_pe - pe_start) / stride;
if ((my_pe < pe_start) ||
((my_pe - pe_start) % stride) ||
(translated_pe >= size)) {
translated_pe = -1;
}
return translated_pe;
}
__device__ int rocshmem_my_pe() {
return get_internal_ctx(ROCSHMEM_CTX_DEFAULT)->my_pe;
}
template <typename T>
__device__ T rocshmem_atomic_fetch_add(rocshmem_ctx_t ctx, T *dest, T val,
int pe) {
GPU_DPRINTF("Function: rocshmem_atomic_fetch_add (ctx=%zd, dest=%p, val=%g, pe=%d w%d)\n",
ctx.ctx_opaque, dest, (double)val, pe, translate_pe(ctx, pe));
return get_internal_ctx(ctx)->amo_fetch_add<T>(dest, val, pe);
}
template <typename T>
__device__ T rocshmem_atomic_compare_swap(rocshmem_ctx_t ctx, T *dest, T cond,
T val, int pe) {
GPU_DPRINTF("Function: rocshmem_atomic_compare_swap (ctx=%zd, dest=%p, cond=%g, val=%g, pe=%d w%d)\n",
ctx.ctx_opaque, dest, cond, (double)val, pe, translate_pe(ctx, pe));
return get_internal_ctx(ctx)->amo_fetch_cas(dest, val, cond, pe);
}
template <typename T>
__device__ T rocshmem_atomic_fetch_inc(rocshmem_ctx_t ctx, T *dest, int pe) {
GPU_DPRINTF("Function: rocshmem_atomic_fetch_inc (ctx=%zd, dest=%p, pe=%d w%d)\n",
ctx.ctx_opaque, dest, pe, translate_pe(ctx, pe));
return get_internal_ctx(ctx)->amo_fetch_add<T>(dest, 1, pe);
}
template <typename T>
__device__ T rocshmem_atomic_fetch(rocshmem_ctx_t ctx, T *source, int pe) {
GPU_DPRINTF("Function: rocshmem_atomic_fetch (ctx=%zd, source=%p, pe=%d w%d)\n",
ctx.ctx_opaque, source, pe, translate_pe(ctx, pe));
return get_internal_ctx(ctx)->amo_fetch_add<T>(source, 0, pe);
}
template <typename T>
__device__ void rocshmem_atomic_add(rocshmem_ctx_t ctx, T *dest, T val,
int pe) {
GPU_DPRINTF("Function: rocshmem_atomic_add (ctx=%zd, dest=%p, val=%g, pe=%d w%d)\n",
ctx.ctx_opaque, dest, (double)val, pe, translate_pe(ctx, pe));
get_internal_ctx(ctx)->amo_add<T>(dest, val, pe);
}
template <typename T>
__device__ void rocshmem_atomic_inc(rocshmem_ctx_t ctx, T *dest, int pe) {
GPU_DPRINTF("Function: rocshmem_atomic_inc (ctx=%zd, dest=%p, pe=%d w%d)\n",
ctx.ctx_opaque, dest, pe, translate_pe(ctx, pe));
get_internal_ctx(ctx)->amo_add<T>(dest, 1, pe);
}
template <typename T>
__device__ void rocshmem_atomic_set(rocshmem_ctx_t ctx, T *dest, T val,
int pe) {
GPU_DPRINTF("Function: rocshmem_atomic_set (ctx=%zd, dest=%p, val=%g, pe=%d w%d)\n",
ctx.ctx_opaque, dest, (double)val, pe, translate_pe(ctx, pe));
get_internal_ctx(ctx)->amo_set(dest, val, pe);
}
template <typename T>
__device__ T rocshmem_atomic_swap(rocshmem_ctx_t ctx, T *dest, T val,
int pe) {
GPU_DPRINTF("Function: rocshmem_atomic_swap (ctx=%zd, dest=%p, val=%g, pe=%d w%d)\n",
ctx.ctx_opaque, dest, (double)val, pe, translate_pe(ctx, pe));
return get_internal_ctx(ctx)->amo_swap(dest, val, pe);
}
template <typename T>
__device__ T rocshmem_atomic_fetch_and(rocshmem_ctx_t ctx, T *dest, T val,
int pe) {
GPU_DPRINTF("Function: rocshmem_atomic_fetch_and (ctx=%zd, dest=%p, val=%g, pe=%d w%d)\n",
ctx.ctx_opaque, dest, (double)val, pe, translate_pe(ctx, pe));
return get_internal_ctx(ctx)->amo_fetch_and(dest, val, pe);
}
template <typename T>
__device__ void rocshmem_atomic_and(rocshmem_ctx_t ctx, T *dest, T val,
int pe) {
GPU_DPRINTF("Function: rocshmem_atomic_and (ctx=%zd, dest=%p, val=%g, pe=%d w%d)\n",
ctx.ctx_opaque, dest, (double)val, pe, translate_pe(ctx, pe));
get_internal_ctx(ctx)->amo_and(dest, val, pe);
}
template <typename T>
__device__ T rocshmem_atomic_fetch_or(rocshmem_ctx_t ctx, T *dest, T val,
int pe) {
GPU_DPRINTF("Function: rocshmem_atomic_fetch_or (ctx=%zd, dest=%p, val=%g, pe=%d w%d)\n",
ctx.ctx_opaque, dest, (double)val, pe, translate_pe(ctx, pe));
return get_internal_ctx(ctx)->amo_fetch_or(dest, val, pe);
}
template <typename T>
__device__ void rocshmem_atomic_or(rocshmem_ctx_t ctx, T *dest, T val,
int pe) {
GPU_DPRINTF("Function: rocshmem_atomic_or (ctx=%zd, dest=%p, val=%g, pe=%d w%d)\n",
ctx.ctx_opaque, dest, (double)val, pe, translate_pe(ctx, pe));
get_internal_ctx(ctx)->amo_or(dest, val, pe);
}
template <typename T>
__device__ T rocshmem_atomic_fetch_xor(rocshmem_ctx_t ctx, T *dest, T val,
int pe) {
GPU_DPRINTF("Function: rocshmem_atomic_fetch_xor (ctx=%zd, dest=%p, val=%g, pe=%d w%d)\n",
ctx.ctx_opaque, dest, (double)val, pe, translate_pe(ctx, pe));
return get_internal_ctx(ctx)->amo_fetch_xor(dest, val, pe);
}
template <typename T>
__device__ void rocshmem_atomic_xor(rocshmem_ctx_t ctx, T *dest, T val,
int pe) {
GPU_DPRINTF("Function: rocshmem_atomic_xor (ctx=%zd, dest=%p, val=%g, pe=%d w%d)\n",
ctx.ctx_opaque, dest, (double)val, pe, translate_pe(ctx, pe));
get_internal_ctx(ctx)->amo_xor(dest, val, pe);
}
/**
* SHMEM X RMA API for WG and Wave level
*/
__device__ void rocshmem_ctx_putmem_wave(rocshmem_ctx_t ctx, void *dest,
const void *source, size_t nelems,
int pe) {
GPU_DPRINTF("Function: rocshmem_ctx_putmem_wave (ctx=%zd, dest=%p, source=%p, nelems=%d, pe=%d w%d)\n",
ctx.ctx_opaque, dest, source, nelems, pe, translate_pe(ctx, pe));
get_internal_ctx(ctx)->putmem_wave(dest, source, nelems, pe);
}
__device__ void rocshmem_ctx_putmem_wg(rocshmem_ctx_t ctx, void *dest,
const void *source, size_t nelems,
int pe) {
GPU_DPRINTF("Function: rocshmem_ctx_putmem_wg (ctx=%zd, dest=%p, source=%p, nelems=%d, pe=%d w%d)\n",
ctx.ctx_opaque, dest, source, nelems, pe, translate_pe(ctx, pe));
get_internal_ctx(ctx)->putmem_wg(dest, source, nelems, pe);
}
__device__ void rocshmem_ctx_putmem_nbi_wave(rocshmem_ctx_t ctx, void *dest,
const void *source,
size_t nelems, int pe) {
GPU_DPRINTF("Function: rocshmem_ctx_putmem_nbi_wave (ctx=%zd, dest=%p, source=%p, nelems=%d, pe=%d w%d)\n",
ctx.ctx_opaque, dest, source, nelems, pe, translate_pe(ctx, pe));
get_internal_ctx(ctx)->putmem_nbi_wave(dest, source, nelems, pe);
}
__device__ void rocshmem_ctx_putmem_nbi_wg(rocshmem_ctx_t ctx, void *dest,
const void *source, size_t nelems,
int pe) {
GPU_DPRINTF("Function: rocshmem_ctx_putmem_nbi_wg (ctx=%zd, dest=%p, source=%p, nelems=%d, pe=%d w%d)\n",
ctx.ctx_opaque, dest, source, nelems, pe, translate_pe(ctx, pe));
get_internal_ctx(ctx)->putmem_nbi_wg(dest, source, nelems, pe);
}
template <typename T>
__device__ void rocshmem_put_wave(rocshmem_ctx_t ctx, T *dest,
const T *source, size_t nelems, int pe) {
GPU_DPRINTF("Function: rocshmem_put_wave (ctx=%zd, dest=%p, source=%p, nelems=%d, pe=%d w%d)\n",
ctx.ctx_opaque, dest, source, nelems, pe, translate_pe(ctx, pe));
get_internal_ctx(ctx)->put_wave(dest, source, nelems, pe);
}
template <typename T>
__device__ void rocshmem_put_wg(rocshmem_ctx_t ctx, T *dest, const T *source,
size_t nelems, int pe) {
GPU_DPRINTF("Function: rocshmem_put_wg (ctx=%zd, dest=%p, source=%p, nelems=%d, pe=%d w%d)\n",
ctx.ctx_opaque, dest, source, nelems, pe, translate_pe(ctx, pe));
get_internal_ctx(ctx)->put_wg(dest, source, nelems, pe);
}
template <typename T>
__device__ void rocshmem_put_nbi_wave(rocshmem_ctx_t ctx, T *dest,
const T *source, size_t nelems,
int pe) {
GPU_DPRINTF("Function: rocshmem_put_nbi_wave (ctx=%zd, dest=%p, source=%p, nelems=%d, pe=%d w%d)\n",
ctx.ctx_opaque, dest, source, nelems, pe, translate_pe(ctx, pe));
get_internal_ctx(ctx)->put_nbi_wave(dest, source, nelems, pe);
}
template <typename T>
__device__ void rocshmem_put_nbi_wg(rocshmem_ctx_t ctx, T *dest,
const T *source, size_t nelems, int pe) {
GPU_DPRINTF("Function: rocshmem_put_nbi_wg (ctx=%zd, dest=%p, source=%p, nelems=%d, pe=%d w%d)\n",
ctx.ctx_opaque, dest, source, nelems, pe, translate_pe(ctx, pe));
get_internal_ctx(ctx)->put_nbi_wg(dest, source, nelems, pe);
}
__device__ void rocshmem_ctx_getmem_wg(rocshmem_ctx_t ctx, void *dest,
const void *source, size_t nelems,
int pe) {
GPU_DPRINTF("Function: rocshmem_ctx_getmem_wg (ctx=%zd, dest=%p, source=%p, nelems=%d, pe=%d w%d)\n",
ctx.ctx_opaque, dest, source, nelems, pe, translate_pe(ctx, pe));
get_internal_ctx(ctx)->getmem_wg(dest, source, nelems, pe);
}
__device__ void rocshmem_ctx_getmem_wave(rocshmem_ctx_t ctx, void *dest,
const void *source, size_t nelems,
int pe) {
GPU_DPRINTF("Function: rocshmem_ctx_getmem_wave (ctx=%zd, dest=%p, source=%p, nelems=%d, pe=%d w%d)\n",
ctx.ctx_opaque, dest, source, nelems, pe, translate_pe(ctx, pe));
get_internal_ctx(ctx)->getmem_wave(dest, source, nelems, pe);
}
template <typename T>
__device__ void rocshmem_get_wg(rocshmem_ctx_t ctx, T *dest, const T *source,
size_t nelems, int pe) {
GPU_DPRINTF("Function: rocshmem_get_wg (ctx=%zd, dest=%p, source=%p, nelems=%d, pe=%d w%d)\n",
ctx.ctx_opaque, dest, source, nelems, pe, translate_pe(ctx, pe));
get_internal_ctx(ctx)->get_wg(dest, source, nelems, pe);
}
template <typename T>
__device__ void rocshmem_get_wave(rocshmem_ctx_t ctx, T *dest,
const T *source, size_t nelems, int pe) {
GPU_DPRINTF("Function: rocshmem_get_wave (ctx=%zd, dest=%p, source=%p, nelems=%d, pe=%d w%d)\n",
ctx.ctx_opaque, dest, source, nelems, pe, translate_pe(ctx, pe));
get_internal_ctx(ctx)->get_wave(dest, source, nelems, pe);
}
__device__ void rocshmem_ctx_getmem_nbi_wg(rocshmem_ctx_t ctx, void *dest,
const void *source, size_t nelems,
int pe) {
GPU_DPRINTF("Function: rocshmem_ctx_getmem_nbi_wg (ctx=%zd, dest=%p, source=%p, nelems=%d, pe=%d w%d)\n",
ctx.ctx_opaque, dest, source, nelems, pe, translate_pe(ctx, pe));
get_internal_ctx(ctx)->getmem_nbi_wg(dest, source, nelems, pe);
}
template <typename T>
__device__ void rocshmem_get_nbi_wg(rocshmem_ctx_t ctx, T *dest,
const T *source, size_t nelems, int pe) {
GPU_DPRINTF("Function: rocshmem_get_nbi_wg (ctx=%zd, dest=%p, source=%p, nelems=%d, pe=%d w%d)\n",
ctx.ctx_opaque, dest, source, nelems, pe, translate_pe(ctx, pe));
get_internal_ctx(ctx)->get_nbi_wg(dest, source, nelems, pe);
}
__device__ void rocshmem_ctx_getmem_nbi_wave(rocshmem_ctx_t ctx, void *dest,
const void *source,
size_t nelems, int pe) {
GPU_DPRINTF("Function: rocshmem_ctx_getmem_nbi_wave (ctx=%zd, dest=%p, source=%p, nelems=%d, pe=%d w%d)\n",
ctx.ctx_opaque, dest, source, nelems, pe, translate_pe(ctx, pe));
get_internal_ctx(ctx)->getmem_nbi_wave(dest, source, nelems, pe);
}
template <typename T>
__device__ void rocshmem_get_nbi_wave(rocshmem_ctx_t ctx, T *dest,
const T *source, size_t nelems,
int pe) {
GPU_DPRINTF("Function: rocshmem_get_nbi_wave (ctx=%zd, dest=%p, source=%p, nelems=%d, pe=%d w%d)\n",
ctx.ctx_opaque, dest, source, nelems, pe, translate_pe(ctx, pe));
get_internal_ctx(ctx)->get_nbi_wave(dest, source, nelems, pe);
}
#define ROCSHMEM_CTX_PUTMEM_SIGNAL_DEF(SUFFIX) \
__device__ void rocshmem_ctx_putmem_signal##SUFFIX(rocshmem_ctx_t ctx, \
void *dest, const void *source, \
size_t nelems, \
uint64_t *sig_addr, uint64_t signal, \
int sig_op, \
int pe) { \
GPU_DPRINTF("Function: rocshmem_ctx_putmem_signal##SUFFIX (ctx=%zd, dest=%p, " \
"source=%p, nelems=%d, sig_addr=%p, signal=%ld, sig_op=%d, pe=%d w%d)\n", \
ctx.ctx_opaque, dest, source, nelems, \
sig_addr, signal, sig_op, pe, translate_pe(ctx, pe)); \
\
get_internal_ctx(ctx)->putmem_signal##SUFFIX(dest, source, nelems, \
sig_addr, signal, sig_op, pe); \
} \
\
template <typename T> \
__device__ void rocshmem_ctx_put_signal##SUFFIX(rocshmem_ctx_t ctx, \
T *dest, const T *source, \
size_t nelems, \
uint64_t *sig_addr, uint64_t signal, \
int sig_op, int pe) { \
GPU_DPRINTF("Function: rocshmem_ctx_put_signal##SUFFIX (ctx=%zd, dest=%p, " \
"source=%p, nelems=%d, sig_addr=%p, signal=%ld, sig_op=%d, pe=%d w%d)\n", \
ctx.ctx_opaque, dest, source, nelems, \
sig_addr, signal, sig_op, pe, translate_pe(ctx, pe)); \
\
get_internal_ctx(ctx)->put_signal##SUFFIX(dest, source, nelems, \
sig_addr, signal, sig_op, pe); \
}
ROCSHMEM_CTX_PUTMEM_SIGNAL_DEF()
ROCSHMEM_CTX_PUTMEM_SIGNAL_DEF(_wg)
ROCSHMEM_CTX_PUTMEM_SIGNAL_DEF(_wave)
ROCSHMEM_CTX_PUTMEM_SIGNAL_DEF(_nbi)
ROCSHMEM_CTX_PUTMEM_SIGNAL_DEF(_nbi_wg)
ROCSHMEM_CTX_PUTMEM_SIGNAL_DEF(_nbi_wave)
#define ROCSHMEM_SIGNAL_FETCH_DEF(SUFFIX) \
__device__ uint64_t rocshmem_signal_fetch##SUFFIX(const uint64_t *sig_addr) { \
return get_internal_ctx(ROCSHMEM_CTX_DEFAULT)->signal_fetch##SUFFIX(sig_addr); \
}
ROCSHMEM_SIGNAL_FETCH_DEF()
ROCSHMEM_SIGNAL_FETCH_DEF(_wg)
ROCSHMEM_SIGNAL_FETCH_DEF(_wave)
/******************************************************************************
****************************** Teams Interface *******************************
*****************************************************************************/
__device__ int rocshmem_team_translate_pe(rocshmem_team_t src_team,
int src_pe,
rocshmem_team_t dst_team) {
return team_translate_pe(src_team, src_pe, dst_team);
}
/******************************************************************************
************************* Template Generation Macros *************************
*****************************************************************************/
/**
* Template generator for reductions
*/
#define REDUCTION_GEN(T, Op) \
template __device__ int rocshmem_reduce_wg<T, Op>( \
rocshmem_ctx_t ctx, rocshmem_team_t team, T * dest, const T *source, \
int nreduce);
/**
* Declare templates for the required datatypes (for the compiler)
*/
#define RMA_GEN(T) \
template __device__ void rocshmem_put<T>( \
rocshmem_ctx_t ctx, T * dest, const T *source, size_t nelems, int pe); \
template __device__ void rocshmem_put_nbi<T>( \
rocshmem_ctx_t ctx, T * dest, const T *source, size_t nelems, int pe); \
template __device__ void rocshmem_p<T>(rocshmem_ctx_t ctx, T * dest, \
T value, int pe); \
template __device__ void rocshmem_get<T>( \
rocshmem_ctx_t ctx, T * dest, const T *source, size_t nelems, int pe); \
template __device__ void rocshmem_get_nbi<T>( \
rocshmem_ctx_t ctx, T * dest, const T *source, size_t nelems, int pe); \
template __device__ T rocshmem_g<T>(rocshmem_ctx_t ctx, const T *source, \
int pe); \
template __device__ void rocshmem_put<T>(T * dest, const T *source, \
size_t nelems, int pe); \
template __device__ void rocshmem_put_nbi<T>(T * dest, const T *source, \
size_t nelems, int pe); \
template __device__ void rocshmem_p<T>(T * dest, T value, int pe); \
template __device__ void rocshmem_get<T>(T * dest, const T *source, \
size_t nelems, int pe); \
template __device__ void rocshmem_get_nbi<T>(T * dest, const T *source, \
size_t nelems, int pe); \
template __device__ T rocshmem_g<T>(const T *source, int pe); \
template __device__ void rocshmem_broadcast_wg<T>( \
rocshmem_ctx_t ctx, rocshmem_team_t team, T * dest, const T *source, \
int nelem, int pe_root); \
template __device__ void rocshmem_alltoall_wg<T>( \
rocshmem_ctx_t ctx, rocshmem_team_t team, T * dest, const T *source, \
int nelem); \
template __device__ void rocshmem_fcollect_wg<T>( \
rocshmem_ctx_t ctx, rocshmem_team_t team, T * dest, const T *source, \
int nelem); \
template __device__ void rocshmem_put_wave<T>( \
rocshmem_ctx_t ctx, T * dest, const T *source, size_t nelems, int pe); \
template __device__ void rocshmem_put_wg<T>( \
rocshmem_ctx_t ctx, T * dest, const T *source, size_t nelems, int pe); \
template __device__ void rocshmem_put_wave<T>(T * dest, const T *source, \
size_t nelems, int pe); \
template __device__ void rocshmem_put_wg<T>(T * dest, const T *source, \
size_t nelems, int pe); \
template __device__ void rocshmem_put_nbi_wave<T>( \
rocshmem_ctx_t ctx, T * dest, const T *source, size_t nelems, int pe); \
template __device__ void rocshmem_put_nbi_wg<T>( \
rocshmem_ctx_t ctx, T * dest, const T *source, size_t nelems, int pe); \
template __device__ void rocshmem_put_nbi_wave<T>( \
T * dest, const T *source, size_t nelems, int pe); \
template __device__ void rocshmem_put_nbi_wg<T>(T * dest, const T *source, \
size_t nelems, int pe); \
template __device__ void rocshmem_get_wave<T>( \
rocshmem_ctx_t ctx, T * dest, const T *source, size_t nelems, int pe); \
template __device__ void rocshmem_get_wg<T>( \
rocshmem_ctx_t ctx, T * dest, const T *source, size_t nelems, int pe); \
template __device__ void rocshmem_get_wave<T>(T * dest, const T *source, \
size_t nelems, int pe); \
template __device__ void rocshmem_get_wg<T>(T * dest, const T *source, \
size_t nelems, int pe); \
template __device__ void rocshmem_get_nbi_wave<T>( \
rocshmem_ctx_t ctx, T * dest, const T *source, size_t nelems, int pe); \
template __device__ void rocshmem_get_nbi_wg<T>( \
rocshmem_ctx_t ctx, T * dest, const T *source, size_t nelems, int pe); \
template __device__ void rocshmem_get_nbi_wave<T>( \
T * dest, const T *source, size_t nelems, int pe); \
template __device__ void rocshmem_get_nbi_wg<T>(T * dest, const T *source, \
size_t nelems, int pe);
/**
* Declare templates for the standard amo types
*/
#define AMO_STANDARD_GEN(T) \
template __device__ T rocshmem_atomic_compare_swap<T>( \
rocshmem_ctx_t ctx, T * dest, T cond, T value, int pe); \
template __device__ T rocshmem_atomic_compare_swap<T>(T * dest, T cond, \
T value, int pe); \
template __device__ T rocshmem_atomic_fetch_inc<T>(rocshmem_ctx_t ctx, \
T * dest, int pe); \
template __device__ T rocshmem_atomic_fetch_inc<T>(T * dest, int pe); \
template __device__ void rocshmem_atomic_inc<T>(rocshmem_ctx_t ctx, \
T * dest, int pe); \
template __device__ void rocshmem_atomic_inc<T>(T * dest, int pe); \
template __device__ T rocshmem_atomic_fetch_add<T>( \
rocshmem_ctx_t ctx, T * dest, T value, int pe); \
template __device__ T rocshmem_atomic_fetch_add<T>(T * dest, T value, \
int pe); \
template __device__ void rocshmem_atomic_add<T>(rocshmem_ctx_t ctx, \
T * dest, T value, int pe); \
template __device__ void rocshmem_atomic_add<T>(T * dest, T value, int pe);
/**
* Declare templates for the extended amo types
*/
#define AMO_EXTENDED_GEN(T) \
template __device__ T rocshmem_atomic_fetch<T>(rocshmem_ctx_t ctx, \
T * dest, int pe); \
template __device__ T rocshmem_atomic_fetch<T>(T * dest, int pe); \
template __device__ void rocshmem_atomic_set<T>(rocshmem_ctx_t ctx, \
T * dest, T value, int pe); \
template __device__ void rocshmem_atomic_set<T>(T * dest, T value, int pe); \
template __device__ T rocshmem_atomic_swap<T>(rocshmem_ctx_t ctx, \
T * dest, T value, int pe); \
template __device__ T rocshmem_atomic_swap<T>(T * dest, T value, int pe);
/**
* Declare templates for the bitwise amo types
*/
#define AMO_BITWISE_GEN(T) \
template __device__ T rocshmem_atomic_fetch_and<T>( \
rocshmem_ctx_t ctx, T * dest, T value, int pe); \
template __device__ T rocshmem_atomic_fetch_and<T>(T * dest, T value, \
int pe); \
template __device__ void rocshmem_atomic_and<T>(rocshmem_ctx_t ctx, \
T * dest, T value, int pe); \
template __device__ void rocshmem_atomic_and<T>(T * dest, T value, int pe); \
template __device__ T rocshmem_atomic_fetch_or<T>( \
rocshmem_ctx_t ctx, T * dest, T value, int pe); \
template __device__ T rocshmem_atomic_fetch_or<T>(T * dest, T value, \
int pe); \
template __device__ void rocshmem_atomic_or<T>(rocshmem_ctx_t ctx, \
T * dest, T value, int pe); \
template __device__ void rocshmem_atomic_or<T>(T * dest, T value, int pe); \
template __device__ T rocshmem_atomic_fetch_xor<T>( \
rocshmem_ctx_t ctx, T * dest, T value, int pe); \
template __device__ T rocshmem_atomic_fetch_xor<T>(T * dest, T value, \
int pe); \
template __device__ void rocshmem_atomic_xor<T>(rocshmem_ctx_t ctx, \
T * dest, T value, int pe); \
template __device__ void rocshmem_atomic_xor<T>(T * dest, T value, int pe);
/**
* Declare templates for the wait types
*/
#define WAIT_GEN(T) \
template __device__ void rocshmem_wait_until<T>(T *ivars, \
int cmp, T val); \
template __device__ size_t rocshmem_wait_until_any<T>(T *ivars, \
size_t nelems, const int* status, \
int cmp, T val); \
template __device__ void rocshmem_wait_until_all<T>(T *ivars, \
size_t nelems, const int* status, \
int cmp, T val); \
template __device__ size_t rocshmem_wait_until_some<T>(T *ivars, \
size_t nelems, size_t* indices, \
const int* status, \
int cmp, T val); \
template __device__ size_t rocshmem_wait_until_any_vector<T>(T *ivars, \
size_t nelems, const int* status, \
int cmp, T* vals); \
template __device__ void rocshmem_wait_until_all_vector<T>(T *ivars, \
size_t nelems, const int* status, \
int cmp, T* vals); \
template __device__ size_t rocshmem_wait_until_some_vector<T>(T *ivars, \
size_t nelems, size_t* indices, \
const int* status, int cmp, \
T* vals); \
template __device__ int rocshmem_test<T>(T *ivars, int cmp, \
T val); \
template __device__ void Context::wait_until<T>(T *ivars, int cmp, \
T val); \
template __device__ size_t Context::wait_until_any<T>(T *ivars, \
size_t nelems, const int* status, \
int cmp, T val); \
template __device__ void Context::wait_until_all<T>(T *ivars, \
size_t nelems, const int* status, \
int cmp, T val); \
template __device__ size_t Context::wait_until_some<T>(T *ivars, \
size_t nelems, \
size_t* indices, const int* status, \
int cmp, T val); \
template __device__ size_t Context::wait_until_any_vector<T>(T *ivars, \
size_t nelems, const int* status, \
int cmp, T* vals); \
template __device__ void Context::wait_until_all_vector<T>(T *ivars, \
size_t nelems, const int* status, \
int cmp, T* vals); \
template __device__ size_t Context::wait_until_some_vector<T>(T *ivars, \
size_t nelems, size_t* indices, \
const int* status, int cmp, \
T* vals); \
template __device__ int Context::test<T>(T *ivars, int cmp, T val);
#define ARITH_REDUCTION_GEN(T) \
REDUCTION_GEN(T, ROCSHMEM_SUM) \
REDUCTION_GEN(T, ROCSHMEM_MIN) \
REDUCTION_GEN(T, ROCSHMEM_MAX) \
REDUCTION_GEN(T, ROCSHMEM_PROD)
#define BITWISE_REDUCTION_GEN(T) \
REDUCTION_GEN(T, ROCSHMEM_OR) \
REDUCTION_GEN(T, ROCSHMEM_AND) \
REDUCTION_GEN(T, ROCSHMEM_XOR)
#define INT_REDUCTION_GEN(T) \
ARITH_REDUCTION_GEN(T) \
BITWISE_REDUCTION_GEN(T)
#define FLOAT_REDUCTION_GEN(T) ARITH_REDUCTION_GEN(T)
/**
* Define APIs to call the template functions
**/
#define REDUCTION_DEF_GEN(T, TNAME, Op_API, Op) \
__device__ int rocshmem_ctx_##TNAME##_##Op_API##_reduce_wg( \
rocshmem_ctx_t ctx, rocshmem_team_t team, T *dest, const T *source, \
int nreduce) { \
return rocshmem_reduce_wg<T, Op>(ctx, team, dest, source, nreduce); \
}
#define ARITH_REDUCTION_DEF_GEN(T, TNAME) \
REDUCTION_DEF_GEN(T, TNAME, sum, ROCSHMEM_SUM) \
REDUCTION_DEF_GEN(T, TNAME, min, ROCSHMEM_MIN) \
REDUCTION_DEF_GEN(T, TNAME, max, ROCSHMEM_MAX) \
REDUCTION_DEF_GEN(T, TNAME, prod, ROCSHMEM_PROD)
#define BITWISE_REDUCTION_DEF_GEN(T, TNAME) \
REDUCTION_DEF_GEN(T, TNAME, or, ROCSHMEM_OR) \
REDUCTION_DEF_GEN(T, TNAME, and, ROCSHMEM_AND) \
REDUCTION_DEF_GEN(T, TNAME, xor, ROCSHMEM_XOR)
#define INT_REDUCTION_DEF_GEN(T, TNAME) \
ARITH_REDUCTION_DEF_GEN(T, TNAME) \
BITWISE_REDUCTION_DEF_GEN(T, TNAME)
#define FLOAT_REDUCTION_DEF_GEN(T, TNAME) ARITH_REDUCTION_DEF_GEN(T, TNAME)
#define RMA_DEF_GEN(T, TNAME) \
__device__ void rocshmem_ctx_##TNAME##_put( \
rocshmem_ctx_t ctx, T *dest, const T *source, size_t nelems, int pe) { \
rocshmem_put<T>(ctx, dest, source, nelems, pe); \
} \
__device__ void rocshmem_ctx_##TNAME##_put_nbi( \
rocshmem_ctx_t ctx, T *dest, const T *source, size_t nelems, int pe) { \
rocshmem_put_nbi<T>(ctx, dest, source, nelems, pe); \
} \
__device__ void rocshmem_ctx_##TNAME##_p(rocshmem_ctx_t ctx, T *dest, \
T value, int pe) { \
rocshmem_p<T>(ctx, dest, value, pe); \
} \
__device__ void rocshmem_ctx_##TNAME##_get( \
rocshmem_ctx_t ctx, T *dest, const T *source, size_t nelems, int pe) { \
rocshmem_get<T>(ctx, dest, source, nelems, pe); \
} \
__device__ T rocshmem_ctx_##TNAME##_g(rocshmem_ctx_t ctx, const T *source, \
int pe) { \
return rocshmem_g<T>(ctx, source, pe); \
} \
__device__ void rocshmem_ctx_##TNAME##_get_nbi( \
rocshmem_ctx_t ctx, T *dest, const T *source, size_t nelems, int pe) { \
rocshmem_get_nbi<T>(ctx, dest, source, nelems, pe); \
} \
__device__ void rocshmem_##TNAME##_put(T *dest, const T *source, \
size_t nelems, int pe) { \
rocshmem_put<T>(dest, source, nelems, pe); \
} \
__device__ void rocshmem_##TNAME##_put_nbi(T *dest, const T *source, \
size_t nelems, int pe) { \
rocshmem_put_nbi<T>(dest, source, nelems, pe); \
} \
__device__ void rocshmem_##TNAME##_p(T *dest, T value, int pe) { \
rocshmem_p<T>(dest, value, pe); \
} \
__device__ void rocshmem_##TNAME##_get(T *dest, const T *source, \
size_t nelems, int pe) { \
rocshmem_get<T>(dest, source, nelems, pe); \
} \
__device__ void rocshmem_##TNAME##_get_nbi(T *dest, const T *source, \
size_t nelems, int pe) { \
rocshmem_get_nbi<T>(dest, source, nelems, pe); \
} \
__device__ T rocshmem_##TNAME##_g(const T *source, int pe) { \
return rocshmem_g<T>(source, pe); \
} \
__device__ void rocshmem_ctx_##TNAME##_put_wave( \
rocshmem_ctx_t ctx, T *dest, const T *source, size_t nelems, int pe) { \
rocshmem_put_wave<T>(ctx, dest, source, nelems, pe); \
} \
__device__ void rocshmem_ctx_##TNAME##_put_wg( \
rocshmem_ctx_t ctx, T *dest, const T *source, size_t nelems, int pe) { \
rocshmem_put_wg<T>(ctx, dest, source, nelems, pe); \
} \
__device__ void rocshmem_##TNAME##_put_wave(T *dest, const T *source, \
size_t nelems, int pe) { \
rocshmem_put_wave<T>(dest, source, nelems, pe); \
} \
__device__ void rocshmem_##TNAME##_put_wg(T *dest, const T *source, \
size_t nelems, int pe) { \
rocshmem_put_wg<T>(dest, source, nelems, pe); \
} \
__device__ void rocshmem_ctx_##TNAME##_put_nbi_wave( \
rocshmem_ctx_t ctx, T *dest, const T *source, size_t nelems, int pe) { \
rocshmem_put_nbi_wave<T>(ctx, dest, source, nelems, pe); \
} \
__device__ void rocshmem_ctx_##TNAME##_put_nbi_wg( \
rocshmem_ctx_t ctx, T *dest, const T *source, size_t nelems, int pe) { \
rocshmem_put_nbi_wg<T>(ctx, dest, source, nelems, pe); \
} \
__device__ void rocshmem_##TNAME##_put_nbi_wave(T *dest, const T *source, \
size_t nelems, int pe) { \
rocshmem_put_nbi_wave<T>(dest, source, nelems, pe); \
} \
__device__ void rocshmem_##TNAME##_put_nbi_wg(T *dest, const T *source, \
size_t nelems, int pe) { \
rocshmem_put_nbi_wg<T>(dest, source, nelems, pe); \
} \
__device__ void rocshmem_ctx_##TNAME##_get_wave( \
rocshmem_ctx_t ctx, T *dest, const T *source, size_t nelems, int pe) { \
rocshmem_get_wave<T>(ctx, dest, source, nelems, pe); \
} \
__device__ void rocshmem_ctx_##TNAME##_get_wg( \
rocshmem_ctx_t ctx, T *dest, const T *source, size_t nelems, int pe) { \
rocshmem_get_wg<T>(ctx, dest, source, nelems, pe); \
} \
__device__ void rocshmem_##TNAME##_get_wave(T *dest, const T *source, \
size_t nelems, int pe) { \
rocshmem_get_wave<T>(dest, source, nelems, pe); \
} \
__device__ void rocshmem_##TNAME##_get_wg(T *dest, const T *source, \
size_t nelems, int pe) { \
rocshmem_get_wg<T>(dest, source, nelems, pe); \
} \
__device__ void rocshmem_ctx_##TNAME##_get_nbi_wave( \
rocshmem_ctx_t ctx, T *dest, const T *source, size_t nelems, int pe) { \
rocshmem_get_nbi_wave<T>(ctx, dest, source, nelems, pe); \
} \
__device__ void rocshmem_ctx_##TNAME##_get_nbi_wg( \
rocshmem_ctx_t ctx, T *dest, const T *source, size_t nelems, int pe) { \
rocshmem_get_nbi_wg<T>(ctx, dest, source, nelems, pe); \
} \
__device__ void rocshmem_##TNAME##_get_nbi_wave(T *dest, const T *source, \
size_t nelems, int pe) { \
rocshmem_get_nbi_wave<T>(dest, source, nelems, pe); \
} \
__device__ void rocshmem_##TNAME##_get_nbi_wg(T *dest, const T *source, \
size_t nelems, int pe) { \
rocshmem_get_nbi_wg<T>(dest, source, nelems, pe); \
} \
__device__ void rocshmem_ctx_##TNAME##_broadcast_wg( \
rocshmem_ctx_t ctx, rocshmem_team_t team, T *dest, const T *source, \
int nelem, int pe_root) { \
rocshmem_broadcast_wg<T>(ctx, team, dest, source, nelem, pe_root); \
} \
__device__ void rocshmem_ctx_##TNAME##_alltoall_wg( \
rocshmem_ctx_t ctx, rocshmem_team_t team, T *dest, const T *source, \
int nelem) { \
rocshmem_alltoall_wg<T>(ctx, team, dest, source, nelem); \
} \
__device__ void rocshmem_ctx_##TNAME##_fcollect_wg( \
rocshmem_ctx_t ctx, rocshmem_team_t team, T *dest, const T *source, \
int nelem) { \
rocshmem_fcollect_wg<T>(ctx, team, dest, source, nelem); \
}
#define AMO_STANDARD_DEF_GEN(T, TNAME) \
__device__ T rocshmem_ctx_##TNAME##_atomic_compare_swap( \
rocshmem_ctx_t ctx, T *dest, T cond, T value, int pe) { \
return rocshmem_atomic_compare_swap<T>(ctx, dest, cond, value, pe); \
} \
__device__ T rocshmem_##TNAME##_atomic_compare_swap(T *dest, T cond, \
T value, int pe) { \
return rocshmem_atomic_compare_swap<T>(dest, cond, value, pe); \
} \
__device__ T rocshmem_ctx_##TNAME##_atomic_fetch_inc(rocshmem_ctx_t ctx, \
T *dest, int pe) { \
return rocshmem_atomic_fetch_inc<T>(ctx, dest, pe); \
} \
__device__ T rocshmem_##TNAME##_atomic_fetch_inc(T *dest, int pe) { \
return rocshmem_atomic_fetch_inc<T>(dest, pe); \
} \
__device__ void rocshmem_ctx_##TNAME##_atomic_inc(rocshmem_ctx_t ctx, \
T *dest, int pe) { \
rocshmem_atomic_inc<T>(ctx, dest, pe); \
} \
__device__ void rocshmem_##TNAME##_atomic_inc(T *dest, int pe) { \
rocshmem_atomic_inc<T>(dest, pe); \
} \
__device__ T rocshmem_ctx_##TNAME##_atomic_fetch_add( \
rocshmem_ctx_t ctx, T *dest, T value, int pe) { \
return rocshmem_atomic_fetch_add<T>(ctx, dest, value, pe); \
} \
__device__ T rocshmem_##TNAME##_atomic_fetch_add(T *dest, T value, \
int pe) { \
return rocshmem_atomic_fetch_add<T>(dest, value, pe); \
} \
__device__ void rocshmem_ctx_##TNAME##_atomic_add( \
rocshmem_ctx_t ctx, T *dest, T value, int pe) { \
rocshmem_atomic_add<T>(ctx, dest, value, pe); \
} \
__device__ void rocshmem_##TNAME##_atomic_add(T *dest, T value, int pe) { \
rocshmem_atomic_add<T>(dest, value, pe); \
}
#define AMO_EXTENDED_DEF_GEN(T, TNAME) \
__device__ T rocshmem_ctx_##TNAME##_atomic_fetch(rocshmem_ctx_t ctx, \
T *source, int pe) { \
return rocshmem_atomic_fetch<T>(ctx, source, pe); \
} \
__device__ T rocshmem_##TNAME##_atomic_fetch(T *source, int pe) { \
return rocshmem_atomic_fetch<T>(source, pe); \
} \
__device__ void rocshmem_ctx_##TNAME##_atomic_set( \
rocshmem_ctx_t ctx, T *dest, T value, int pe) { \
rocshmem_atomic_set<T>(ctx, dest, value, pe); \
} \
__device__ void rocshmem_##TNAME##_atomic_set(T *dest, T value, int pe) { \
rocshmem_atomic_set<T>(dest, value, pe); \
} \
__device__ T rocshmem_ctx_##TNAME##_atomic_swap(rocshmem_ctx_t ctx, \
T *dest, T value, int pe) {\
return rocshmem_atomic_swap<T>(ctx, dest, value, pe); \
} \
__device__ T rocshmem_##TNAME##_atomic_swap(T *dest, T value, int pe) { \
return rocshmem_atomic_swap<T>(dest, value, pe); \
}
#define AMO_BITWISE_DEF_GEN(T, TNAME) \
__device__ T rocshmem_ctx_##TNAME##_atomic_fetch_and( \
rocshmem_ctx_t ctx, T *dest, T value, int pe) { \
return rocshmem_atomic_fetch_and<T>(ctx, dest, value, pe); \
} \
__device__ T rocshmem_##TNAME##_atomic_fetch_and(T *dest, T value, \
int pe) { \
return rocshmem_atomic_fetch_and<T>(dest, value, pe); \
} \
__device__ void rocshmem_ctx_##TNAME##_atomic_and( \
rocshmem_ctx_t ctx, T *dest, T value, int pe) { \
rocshmem_atomic_and<T>(ctx, dest, value, pe); \
} \
__device__ void rocshmem_##TNAME##_atomic_and(T *dest, T value, int pe) { \
rocshmem_atomic_and<T>(dest, value, pe); \
} \
__device__ T rocshmem_ctx_##TNAME##_atomic_fetch_or( \
rocshmem_ctx_t ctx, T *dest, T value, int pe) { \
return rocshmem_atomic_fetch_or<T>(ctx, dest, value, pe); \
} \
__device__ T rocshmem_##TNAME##_atomic_fetch_or(T *dest, T value, int pe) { \
return rocshmem_atomic_fetch_or<T>(dest, value, pe); \
} \
__device__ void rocshmem_ctx_##TNAME##_atomic_or( \
rocshmem_ctx_t ctx, T *dest, T value, int pe) { \
rocshmem_atomic_or<T>(ctx, dest, value, pe); \
} \
__device__ void rocshmem_##TNAME##_atomic_or(T *dest, T value, int pe) { \
rocshmem_atomic_or<T>(dest, value, pe); \
} \
__device__ T rocshmem_ctx_##TNAME##_atomic_fetch_xor( \
rocshmem_ctx_t ctx, T *dest, T value, int pe) { \
return rocshmem_atomic_fetch_xor<T>(ctx, dest, value, pe); \
} \
__device__ T rocshmem_##TNAME##_atomic_fetch_xor(T *dest, T value, \
int pe) { \
return rocshmem_atomic_fetch_xor<T>(dest, value, pe); \
} \
__device__ void rocshmem_ctx_##TNAME##_atomic_xor( \
rocshmem_ctx_t ctx, T *dest, T value, int pe) { \
rocshmem_atomic_xor<T>(ctx, dest, value, pe); \
} \
__device__ void rocshmem_##TNAME##_atomic_xor(T *dest, T value, int pe) { \
rocshmem_atomic_xor<T>(dest, value, pe); \
}
#define WAIT_DEF_GEN(T, TNAME) \
__device__ void rocshmem_##TNAME##_wait_until(T *ivars, int cmp, \
T val) { \
rocshmem_wait_until<T>(ivars, cmp, val); \
} \
__device__ size_t rocshmem_##TNAME##_wait_until_any(T *ivars, size_t nelems,\
const int* status, \
int cmp, \
T val) { \
return rocshmem_wait_until_any<T>(ivars, nelems, status, cmp, val); \
} \
__device__ void rocshmem_##TNAME##_wait_until_all(T *ivars, size_t nelems, \
const int* status, \
int cmp, \
T val) { \
rocshmem_wait_until_all<T>(ivars, nelems, status, cmp, val); \
} \
__device__ size_t rocshmem_##TNAME##_wait_until_some(T *ivars, \
size_t nelems, \
size_t* indices, \
const int* status, \
int cmp, \
T val) { \
return rocshmem_wait_until_some<T>(ivars, nelems, indices, status, cmp, \
val); \
} \
__device__ size_t rocshmem_##TNAME##_wait_until_any_vector(T *ivars, \
size_t nelems, \
const int* status, \
int cmp, \
T* vals) { \
return rocshmem_wait_until_any_vector<T>(ivars, nelems, status, cmp, \
vals); \
} \
__device__ void rocshmem_##TNAME##_wait_until_all_vector(T *ivars, \
size_t nelems, \
const int* status, \
int cmp, \
T* vals) { \
rocshmem_wait_until_all_vector<T>(ivars, nelems, status, cmp, vals); \
} \
__device__ size_t rocshmem_##TNAME##_wait_until_some_vector(T *ivars, \
size_t nelems, \
size_t* indices, \
const int* status, \
int cmp, \
T* vals) { \
return rocshmem_wait_until_some_vector<T>(ivars, nelems, indices, \
status, cmp, vals); \
} \
__device__ int rocshmem_##TNAME##_test(T *ivars, int cmp, T val) { \
return rocshmem_test<T>(ivars, cmp, val); \
}
#define RMA_SIGNAL_SUFFIX_DEC(SUFFIX) \
template <typename T> \
__device__ void rocshmem_ctx_put_signal##SUFFIX(rocshmem_ctx_t ctx, \
T *dest, const T *source, \
size_t nelems, \
uint64_t *sig_addr, uint64_t signal, \
int sig_op, int pe); \
\
template <typename T> \
__device__ void rocshmem_put_signal##SUFFIX(T *dest, const T *source, size_t nelems, \
uint64_t *sig_addr, uint64_t signal, \
int sig_op, int pe); \
#define RMA_SIGNAL_SUFFIX_DEF(T, TNAME, SUFFIX) \
__device__ void rocshmem_ctx_##TNAME##_put_signal##SUFFIX(rocshmem_ctx_t ctx, \
T *dest, const T *source, \
size_t nelems, \
uint64_t *sig_addr, uint64_t signal, \
int sig_op, int pe) { \
rocshmem_ctx_put_signal##SUFFIX<T>(ctx, dest, source, nelems, sig_addr, signal, sig_op, pe); \
} \
\
__device__ void rocshmem_##TNAME##_put_signal##SUFFIX(T *dest, const T *source, size_t nelems, \
uint64_t *sig_addr, uint64_t signal, \
int sig_op, int pe) { \
rocshmem_put_signal##SUFFIX(dest, source, nelems, sig_addr, signal, sig_op, pe); \
}
#define RMA_SIGNAL_GEN(SUFFIX) \
RMA_SIGNAL_SUFFIX_DEC(SUFFIX) \
RMA_SIGNAL_SUFFIX_DEF(float, float, SUFFIX) \
RMA_SIGNAL_SUFFIX_DEF(double, double, SUFFIX) \
RMA_SIGNAL_SUFFIX_DEF(char, char, SUFFIX) \
RMA_SIGNAL_SUFFIX_DEF(signed char, schar, SUFFIX) \
RMA_SIGNAL_SUFFIX_DEF(short, short, SUFFIX) \
RMA_SIGNAL_SUFFIX_DEF(int, int, SUFFIX) \
RMA_SIGNAL_SUFFIX_DEF(long, long, SUFFIX) \
RMA_SIGNAL_SUFFIX_DEF(long long, longlong, SUFFIX) \
RMA_SIGNAL_SUFFIX_DEF(unsigned char, uchar, SUFFIX) \
RMA_SIGNAL_SUFFIX_DEF(unsigned short, ushort, SUFFIX) \
RMA_SIGNAL_SUFFIX_DEF(unsigned int, uint, SUFFIX) \
RMA_SIGNAL_SUFFIX_DEF(unsigned long, ulong, SUFFIX) \
RMA_SIGNAL_SUFFIX_DEF(unsigned long long, ulonglong, SUFFIX) \
RMA_SIGNAL_SUFFIX_DEF(int8_t, int8, SUFFIX) \
RMA_SIGNAL_SUFFIX_DEF(int16_t, int16, SUFFIX) \
RMA_SIGNAL_SUFFIX_DEF(int32_t, int32, SUFFIX) \
RMA_SIGNAL_SUFFIX_DEF(int64_t, int64, SUFFIX) \
RMA_SIGNAL_SUFFIX_DEF(uint8_t, uint8, SUFFIX) \
RMA_SIGNAL_SUFFIX_DEF(uint16_t, uint16, SUFFIX) \
RMA_SIGNAL_SUFFIX_DEF(uint32_t, uint32, SUFFIX) \
RMA_SIGNAL_SUFFIX_DEF(uint64_t, uint64, SUFFIX) \
RMA_SIGNAL_SUFFIX_DEF(size_t, size, SUFFIX) \
RMA_SIGNAL_SUFFIX_DEF(ptrdiff_t, ptrdiff, SUFFIX)
RMA_SIGNAL_GEN(_wg)
RMA_SIGNAL_GEN()
RMA_SIGNAL_GEN(_wave)
RMA_SIGNAL_GEN(_nbi)
RMA_SIGNAL_GEN(_nbi_wg)
RMA_SIGNAL_GEN(_nbi_wave)
/******************************************************************************
************************* Macro Invocation Per Type **************************
*****************************************************************************/
// clang-format off
INT_REDUCTION_GEN(int)
INT_REDUCTION_GEN(short)
INT_REDUCTION_GEN(long)
INT_REDUCTION_GEN(long long)
FLOAT_REDUCTION_GEN(float)
FLOAT_REDUCTION_GEN(double)
// long double reduction fails. hipcc/device may not support long double.
// so disable it for now.
// FLOAT_REDUCTION_GEN(long double)
RMA_GEN(float)
RMA_GEN(double)
// RMA_GEN(long double)
RMA_GEN(char)
RMA_GEN(signed char)
RMA_GEN(short)
RMA_GEN(int)
RMA_GEN(long)
RMA_GEN(long long)
RMA_GEN(unsigned char)
RMA_GEN(unsigned short)
RMA_GEN(unsigned int)
RMA_GEN(unsigned long)
RMA_GEN(unsigned long long)
AMO_STANDARD_GEN(int)
AMO_STANDARD_GEN(long)
AMO_STANDARD_GEN(long long)
AMO_STANDARD_GEN(unsigned int)
AMO_STANDARD_GEN(unsigned long)
AMO_STANDARD_GEN(unsigned long long)
AMO_EXTENDED_GEN(float)
AMO_EXTENDED_GEN(double)
AMO_EXTENDED_GEN(int)
AMO_EXTENDED_GEN(long)
AMO_EXTENDED_GEN(long long)
AMO_EXTENDED_GEN(unsigned int)
AMO_EXTENDED_GEN(unsigned long)
AMO_EXTENDED_GEN(unsigned long long)
AMO_BITWISE_GEN(unsigned int)
AMO_BITWISE_GEN(unsigned long)
AMO_BITWISE_GEN(unsigned long long)
/* Supported synchronization types */
WAIT_GEN(float)
WAIT_GEN(double)
// WAIT_GEN(long double)
WAIT_GEN(char)
WAIT_GEN(unsigned char)
WAIT_GEN(unsigned short)
WAIT_GEN(signed char)
WAIT_GEN(short)
WAIT_GEN(int)
WAIT_GEN(long)
WAIT_GEN(long long)
WAIT_GEN(unsigned int)
WAIT_GEN(unsigned long)
WAIT_GEN(unsigned long long)
INT_REDUCTION_DEF_GEN(int, int)
INT_REDUCTION_DEF_GEN(short, short)
INT_REDUCTION_DEF_GEN(long, long)
INT_REDUCTION_DEF_GEN(long long, longlong)
FLOAT_REDUCTION_DEF_GEN(float, float)
FLOAT_REDUCTION_DEF_GEN(double, double)
// long double reduction fails. hipcc/device may not support long double.
// so disable it for now.
// FLOAT_REDUCTION_DEF_GEN(long double, longdouble)
RMA_DEF_GEN(float, float)
RMA_DEF_GEN(double, double)
RMA_DEF_GEN(char, char)
// RMA_DEF_GEN(long double, longdouble)
RMA_DEF_GEN(signed char, schar)
RMA_DEF_GEN(short, short)
RMA_DEF_GEN(int, int)
RMA_DEF_GEN(long, long)
RMA_DEF_GEN(long long, longlong)
RMA_DEF_GEN(unsigned char, uchar)
RMA_DEF_GEN(unsigned short, ushort)
RMA_DEF_GEN(unsigned int, uint)
RMA_DEF_GEN(unsigned long, ulong)
RMA_DEF_GEN(unsigned long long, ulonglong)
RMA_DEF_GEN(int8_t, int8)
RMA_DEF_GEN(int16_t, int16)
RMA_DEF_GEN(int32_t, int32)
RMA_DEF_GEN(int64_t, int64)
RMA_DEF_GEN(uint8_t, uint8)
RMA_DEF_GEN(uint16_t, uint16)
RMA_DEF_GEN(uint32_t, uint32)
RMA_DEF_GEN(uint64_t, uint64)
RMA_DEF_GEN(size_t, size)
RMA_DEF_GEN(ptrdiff_t, ptrdiff)
AMO_STANDARD_DEF_GEN(int, int)
AMO_STANDARD_DEF_GEN(long, long)
AMO_STANDARD_DEF_GEN(long long, longlong)
AMO_STANDARD_DEF_GEN(unsigned int, uint)
AMO_STANDARD_DEF_GEN(unsigned long, ulong)
AMO_STANDARD_DEF_GEN(unsigned long long, ulonglong)
AMO_STANDARD_DEF_GEN(int32_t, int32)
AMO_STANDARD_DEF_GEN(int64_t, int64)
AMO_STANDARD_DEF_GEN(uint32_t, uint32)
AMO_STANDARD_DEF_GEN(uint64_t, uint64)
AMO_STANDARD_DEF_GEN(size_t, size)
AMO_STANDARD_DEF_GEN(ptrdiff_t, ptrdiff)
AMO_EXTENDED_DEF_GEN(float, float)
AMO_EXTENDED_DEF_GEN(double, double)
AMO_EXTENDED_DEF_GEN(int, int)
AMO_EXTENDED_DEF_GEN(long, long)
AMO_EXTENDED_DEF_GEN(long long, longlong)
AMO_EXTENDED_DEF_GEN(unsigned int, uint)
AMO_EXTENDED_DEF_GEN(unsigned long, ulong)
AMO_EXTENDED_DEF_GEN(unsigned long long, ulonglong)
AMO_EXTENDED_DEF_GEN(int32_t, int32)
AMO_EXTENDED_DEF_GEN(int64_t, int64)
AMO_EXTENDED_DEF_GEN(uint32_t, uint32)
AMO_EXTENDED_DEF_GEN(uint64_t, uint64)
AMO_EXTENDED_DEF_GEN(size_t, size)
AMO_EXTENDED_DEF_GEN(ptrdiff_t, ptrdiff)
AMO_BITWISE_DEF_GEN(unsigned int, uint)
AMO_BITWISE_DEF_GEN(unsigned long, ulong)
AMO_BITWISE_DEF_GEN(unsigned long long, ulonglong)
AMO_BITWISE_DEF_GEN(int32_t, int32)
AMO_BITWISE_DEF_GEN(int64_t, int64)
AMO_BITWISE_DEF_GEN(uint32_t, uint32)
AMO_BITWISE_DEF_GEN(uint64_t, uint64)
WAIT_DEF_GEN(float, float)
WAIT_DEF_GEN(double, double)
// WAIT_DEF_GEN(long double, longdouble)
WAIT_DEF_GEN(char, char)
WAIT_DEF_GEN(signed char, schar)
WAIT_DEF_GEN(short, short)
WAIT_DEF_GEN(int, int)
WAIT_DEF_GEN(long, long)
WAIT_DEF_GEN(long long, longlong)
WAIT_DEF_GEN(unsigned char, uchar)
WAIT_DEF_GEN(unsigned short, ushort)
WAIT_DEF_GEN(unsigned int, uint)
WAIT_DEF_GEN(unsigned long, ulong)
WAIT_DEF_GEN(unsigned long long, ulonglong)
// clang-format on
} // namespace rocshmem
Criar uma nova questão referindo esta Ver Git Blame Copiar ligação permanente