1
1
openmpi/opal/class/opal_fifo.h
Noah Evans ee3517427e Add threads framework
Add a framework to support different types of threading models including
user space thread packages such as Qthreads and argobot:

https://github.com/pmodels/argobots

https://github.com/Qthreads/qthreads

The default threading model is pthreads.  Alternate thread models are
specificed at configure time using the --with-threads=X option.

The framework is static.  The theading model to use is selected at
Open MPI configure/build time.

mca/threads: implement Argobots threading layer

config: fix thread configury

- Add double quotations
- Change Argobot to Argobots
config: implement Argobots check

If the poll time is too long, MPI hangs.

This quick fix just sets it to 0, but it is not good for the
Pthreads version. Need to find a good way to abstract it.

Note that even 1 (= 1 millisecond) causes disastrous performance
degradation.

rework threads MCA framework configury

It now works more like the ompi/mca/rte configury,
modulo some edge items that are special for threading package
linking, etc.

qthreads module
some argobots cleanup

Signed-off-by: Noah Evans <noah.evans@gmail.com>
Signed-off-by: Shintaro Iwasaki <siwasaki@anl.gov>
Signed-off-by: Howard Pritchard <howardp@lanl.gov>
2020-03-27 10:15:45 -06:00

350 строки
11 KiB
C

/* -*- Mode: C; c-basic-offset:4 ; indent-tabs-mode:nil -*- */
/*
* Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
* University Research and Technology
* Corporation. All rights reserved.
* Copyright (c) 2004-2007 The University of Tennessee and The University
* of Tennessee Research Foundation. All rights
* reserved.
* Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
* University of Stuttgart. All rights reserved.
* Copyright (c) 2004-2005 The Regents of the University of California.
* All rights reserved.
* Copyright (c) 2007 Voltaire All rights reserved.
* Copyright (c) 2010 IBM Corporation. All rights reserved.
* Copyright (c) 2014-2018 Los Alamos National Security, LLC. All rights
* reseved.
* $COPYRIGHT$
*
* Additional copyrights may follow
*
* $HEADER$
*/
#ifndef OPAL_FIFO_H_HAS_BEEN_INCLUDED
#define OPAL_FIFO_H_HAS_BEEN_INCLUDED
#include "opal_config.h"
#include "opal/class/opal_lifo.h"
#include "opal/sys/atomic.h"
#include "opal/mca/threads/mutex.h"
BEGIN_C_DECLS
/* Atomic First In First Out lists. If we are in a multi-threaded environment then the
* atomicity is insured via the compare-and-swap operation, if not we simply do a read
* and/or a write.
*
* There is a trick. The ghost element at the end of the list. This ghost element has
* the next pointer pointing to itself, therefore we cannot go past the end of the list.
* With this approach we will never have a NULL element in the list, so we never have
* to test for the NULL.
*/
struct opal_fifo_t {
opal_object_t super;
/** first element on the fifo */
volatile opal_counted_pointer_t opal_fifo_head;
/** last element on the fifo */
volatile opal_counted_pointer_t opal_fifo_tail;
/** list sentinel (always points to self) */
opal_list_item_t opal_fifo_ghost;
};
typedef struct opal_fifo_t opal_fifo_t;
OPAL_DECLSPEC OBJ_CLASS_DECLARATION(opal_fifo_t);
static inline opal_list_item_t *opal_fifo_head (opal_fifo_t* fifo)
{
return (opal_list_item_t *) fifo->opal_fifo_head.data.item;
}
static inline opal_list_item_t *opal_fifo_tail (opal_fifo_t* fifo)
{
return (opal_list_item_t *) fifo->opal_fifo_tail.data.item;
}
/* The ghost pointer will never change. The head will change via an atomic
* compare-and-swap. On most architectures the reading of a pointer is an
* atomic operation so we don't have to protect it.
*/
static inline bool opal_fifo_is_empty( opal_fifo_t* fifo )
{
return opal_fifo_head (fifo) == &fifo->opal_fifo_ghost;
}
#if OPAL_HAVE_ATOMIC_COMPARE_EXCHANGE_128
/* Add one element to the FIFO. We will return the last head of the list
* to allow the upper level to detect if this element is the first one in the
* list (if the list was empty before this operation).
*/
static inline opal_list_item_t *opal_fifo_push_atomic (opal_fifo_t *fifo,
opal_list_item_t *item)
{
opal_counted_pointer_t tail = {.value = fifo->opal_fifo_tail.value};
const opal_list_item_t * const ghost = &fifo->opal_fifo_ghost;
item->opal_list_next = (opal_list_item_t *) ghost;
opal_atomic_wmb ();
do {
if (opal_update_counted_pointer (&fifo->opal_fifo_tail, &tail, item)) {
break;
}
} while (1);
opal_atomic_wmb ();
if ((intptr_t) ghost == tail.data.item) {
/* update the head */
opal_counted_pointer_t head = {.value = fifo->opal_fifo_head.value};
opal_update_counted_pointer (&fifo->opal_fifo_head, &head, item);
} else {
/* update previous item */
((opal_list_item_t *) tail.data.item)->opal_list_next = item;
}
return (opal_list_item_t *) tail.data.item;
}
/* Retrieve one element from the FIFO. If we reach the ghost element then the FIFO
* is empty so we return NULL.
*/
static inline opal_list_item_t *opal_fifo_pop_atomic (opal_fifo_t *fifo)
{
opal_list_item_t *item, *next, *ghost = &fifo->opal_fifo_ghost;
opal_counted_pointer_t head, tail;
opal_read_counted_pointer (&fifo->opal_fifo_head, &head);
do {
tail.value = fifo->opal_fifo_tail.value;
opal_atomic_rmb ();
item = (opal_list_item_t *) head.data.item;
next = (opal_list_item_t *) item->opal_list_next;
if ((intptr_t) ghost == tail.data.item && ghost == item) {
return NULL;
}
/* the head or next pointer are in an inconsistent state. keep looping. */
if (tail.data.item != (intptr_t) item && (intptr_t) ghost != tail.data.item && ghost == next) {
opal_read_counted_pointer (&fifo->opal_fifo_head, &head);
continue;
}
/* try popping the head */
if (opal_update_counted_pointer (&fifo->opal_fifo_head, &head, next)) {
break;
}
} while (1);
opal_atomic_wmb ();
/* check for tail and head consistency */
if (ghost == next) {
/* the head was just set to &fifo->opal_fifo_ghost. try to update the tail as well */
if (!opal_update_counted_pointer (&fifo->opal_fifo_tail, &tail, ghost)) {
/* tail was changed by a push operation. wait for the item's next pointer to be se then
* update the head */
/* wait for next pointer to be updated by push */
do {
opal_atomic_rmb ();
} while (ghost == item->opal_list_next);
/* update the head with the real next value. note that no other thread
* will be attempting to update the head until after it has been updated
* with the next pointer. push will not see an empty list and other pop
* operations will loop until the head is consistent. */
fifo->opal_fifo_head.data.item = (intptr_t) item->opal_list_next;
opal_atomic_wmb ();
}
}
item->opal_list_next = NULL;
return item;
}
#else
/* When compare-and-set 128 is not available we avoid the ABA problem by
* using a spin-lock on the head (using the head counter). Otherwise
* the algorithm is identical to the compare-and-set 128 version. */
static inline opal_list_item_t *opal_fifo_push_atomic (opal_fifo_t *fifo,
opal_list_item_t *item)
{
const opal_list_item_t * const ghost = &fifo->opal_fifo_ghost;
opal_list_item_t *tail_item;
item->opal_list_next = (opal_list_item_t *) ghost;
opal_atomic_wmb ();
/* try to get the tail */
tail_item = (opal_list_item_t *) opal_atomic_swap_ptr (&fifo->opal_fifo_tail.data.item, (intptr_t) item);
opal_atomic_wmb ();
if (ghost == tail_item) {
/* update the head */
fifo->opal_fifo_head.data.item = (intptr_t) item;
} else {
/* update previous item */
tail_item->opal_list_next = item;
}
opal_atomic_wmb ();
return (opal_list_item_t *) tail_item;
}
/* Retrieve one element from the FIFO. If we reach the ghost element then the FIFO
* is empty so we return NULL.
*/
static inline opal_list_item_t *opal_fifo_pop_atomic (opal_fifo_t *fifo)
{
const opal_list_item_t * const ghost = &fifo->opal_fifo_ghost;
#if OPAL_HAVE_ATOMIC_LLSC_PTR
register opal_list_item_t *item, *next;
int attempt = 0, ret = 0;
/* use load-linked store-conditional to avoid ABA issues */
do {
if (++attempt == 5) {
/* deliberatly suspend this thread to allow other threads to run. this should
* only occur during periods of contention on the lifo. */
_opal_lifo_release_cpu ();
attempt = 0;
}
opal_atomic_ll_ptr(&fifo->opal_fifo_head.data.item, item);
if (ghost == item) {
if ((intptr_t) ghost == fifo->opal_fifo_tail.data.item) {
return NULL;
}
/* fifo does not appear empty. wait for the fifo to be made
* consistent by conflicting thread. */
continue;
}
next = (opal_list_item_t *) item->opal_list_next;
opal_atomic_sc_ptr(&fifo->opal_fifo_head.data.item, next, ret);
} while (!ret);
#else
opal_list_item_t *item, *next;
/* protect against ABA issues by "locking" the head */
do {
if (!opal_atomic_swap_32 ((opal_atomic_int32_t *) &fifo->opal_fifo_head.data.counter, 1)) {
break;
}
opal_atomic_wmb ();
} while (1);
opal_atomic_wmb();
item = opal_fifo_head (fifo);
if (ghost == item) {
fifo->opal_fifo_head.data.counter = 0;
return NULL;
}
next = (opal_list_item_t *) item->opal_list_next;
fifo->opal_fifo_head.data.item = (uintptr_t) next;
#endif
if (ghost == next) {
void *tmp = item;
if (!opal_atomic_compare_exchange_strong_ptr (&fifo->opal_fifo_tail.data.item, (intptr_t *) &tmp, (intptr_t) ghost)) {
do {
opal_atomic_rmb ();
} while (ghost == item->opal_list_next);
fifo->opal_fifo_head.data.item = (intptr_t) item->opal_list_next;
}
}
opal_atomic_wmb ();
/* unlock the head */
fifo->opal_fifo_head.data.counter = 0;
item->opal_list_next = NULL;
return item;
}
#endif
/* single threaded versions of push/pop */
static inline opal_list_item_t *opal_fifo_push_st (opal_fifo_t *fifo,
opal_list_item_t *item)
{
opal_list_item_t *prev = opal_fifo_tail (fifo);
item->opal_list_next = &fifo->opal_fifo_ghost;
fifo->opal_fifo_tail.data.item = (intptr_t) item;
if (&fifo->opal_fifo_ghost == opal_fifo_head (fifo)) {
fifo->opal_fifo_head.data.item = (intptr_t) item;
} else {
prev->opal_list_next = item;
}
return (opal_list_item_t *) item->opal_list_next;
}
static inline opal_list_item_t *opal_fifo_pop_st (opal_fifo_t *fifo)
{
opal_list_item_t *item = opal_fifo_head (fifo);
if (item == &fifo->opal_fifo_ghost) {
return NULL;
}
fifo->opal_fifo_head.data.item = (intptr_t) item->opal_list_next;
if (&fifo->opal_fifo_ghost == opal_fifo_head (fifo)) {
fifo->opal_fifo_tail.data.item = (intptr_t) &fifo->opal_fifo_ghost;
}
item->opal_list_next = NULL;
return item;
}
/* push/pop versions conditioned off opal_using_threads() */
static inline opal_list_item_t *opal_fifo_push (opal_fifo_t *fifo,
opal_list_item_t *item)
{
if (opal_using_threads ()) {
return opal_fifo_push_atomic (fifo, item);
}
return opal_fifo_push_st (fifo, item);
}
static inline opal_list_item_t *opal_fifo_pop (opal_fifo_t *fifo)
{
if (opal_using_threads ()) {
return opal_fifo_pop_atomic (fifo);
}
return opal_fifo_pop_st (fifo);
}
END_C_DECLS
#endif /* OPAL_FIFO_H_HAS_BEEN_INCLUDED */