ems-typed
Version:
Persistent Shared Memory and Parallel Programming Model
284 lines (256 loc) • 11.9 kB
JavaScript
/*-----------------------------------------------------------------------------+
| Extended Memory Semantics (EMS) Version 1.0.0 |
| Synthetic Semantics http://www.synsem.com/ mogill@synsem.com |
+-----------------------------------------------------------------------------+
| Copyright (c) 2011-2014, Synthetic Semantics LLC. All rights reserved. |
| Copyright (c) 2015-2016, Jace A Mogill. All rights reserved. |
| |
| Redistribution and use in source and binary forms, with or without |
| modification, are permitted provided that the following conditions are met: |
| * Redistributions of source code must retain the above copyright |
| notice, this list of conditions and the following disclaimer. |
| * Redistributions in binary form must reproduce the above copyright |
| notice, this list of conditions and the following disclaimer in the |
| documentation and/or other materials provided with the distribution. |
| * Neither the name of the Synthetic Semantics nor the names of its |
| contributors may be used to endorse or promote products derived |
| from this software without specific prior written permission. |
| |
| THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS |
| "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT |
| LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR |
| A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL SYNTHETIC |
| SEMANTICS LLC BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, |
| EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, |
| PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR |
| PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF |
| LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING |
| NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS |
| SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
| |
+-----------------------------------------------------------------------------+
| Program Description: |
| |
| Sequentially enqueue transactions from thread 0 while all other |
| threads consume them. When all the operations have been queued |
| thread 0 also begins dequeing work and processing transactions. |
| |
| Unlike the workQ_and_TM.js example, randomInRange() is only called |
| from thread 0 so it is defined only in the main thread's context. |
| |
+-----------------------------------------------------------------------------*/
;
// Initialize EMS to use the fork-join execution model
var ems = require('ems')(parseInt(process.argv[2]), true, 'fj');
var assert = require('assert');
var workQ;
var totalNops;
var checkNops;
var arrLen;
var heapSize;
var nTransactions;
var nTables;
var maxNops;
var tables;
//---------------------------------------------------------------------------
// Generate a random integer within a range (inclusive) from 'low' to 'high'
//
function randomInRange(low, high) {
return (Math.floor((Math.random() * (high - low)) + low));
}
//-------------------------------------------------------------------
// Timer functions
function timerStart() {
return new Date().getTime();
}
function timerStop(timer, nOps, label, myID) {
function fmtNumber(n) {
var s = ' ' + n.toString().replace(/\B(?=(\d{3})+(?!\d))/g, ",");
if (n < 1) return n;
else {
return s.substr(s.length - 15, s.length);
}
}
var now = new Date().getTime();
var opsPerSec = (nOps * 1000000) / ((now - timer) * 1000);
if (typeof myID === undefined || myID === 0) {
console.log(fmtNumber(nOps) + label + fmtNumber(Math.floor(opsPerSec).toString()) + " ops/sec");
}
}
//---------------------------------------------------------------------------
// Initialize shared data: global scalars, EMS buffers for statistics
// and checksums, an EMS array to be used as a work queue, and many
// tables to perform transactions on.
//
function initializeSharedData() {
arrLen = 1000000; // Maximum number of elements the EMS array can contain
heapSize = 100000; // Amount of persistent memory to reserve for transactions
nTransactions = 1000000; // Number of transactions to perform in the experiment
nTables = 6; // Number of EMS arrays to perform transactions across
maxNops = 5; // Maximum number of EMS arrays to update during a transaction
tables = [];
totalNops = ems.new(2);
checkNops = ems.new(1);
//---------------------------------------------------------------------------
// The queue of transactions to perform
// [ table#, index, read-only ]
workQ = ems.new({
dimensions: [nTransactions + ems.nThreads],
heapSize: nTransactions * 200,
useExisting: false,
setFEtags: 'empty'
});
//---------------------------------------------------------------------------
// Create all the tables
for (var tableN = 0; tableN < nTables; tableN++) {
tables[tableN] = ems.new({
dimensions: [arrLen],
heapSize: 0,
useExisting: false,
filename: '/tmp/EMS_tm' + tableN,
doDataFill: true,
dataFill: 0,
setFEtags: 'full'
});
}
}
//---------------------------------------------------------------------------
// Create 'nTransactions' many transactions, each having a random number (up
// to maxNops) randomly chosen values, some of which are read-only.
// Because tables and elements are chosen at random, it is necessary to
// remove duplicate elements in a single transaction, to prevent deadlocks.
//
// Transactions are pushed onto a shared queue
function generateTransactions() {
//---------------------------------------------------------------------------
// Generate operations involving random elements in random EMS arrays
// and enqueue them on the work queue
for (var transN = 0; transN < nTransactions; transN++) {
var ops = [];
var nOps = randomInRange(1, maxNops);
// var indexes = [];
for (var opN = 0; opN < nOps; opN++) {
var tableN = randomInRange(0, nTables);
var idx = randomInRange(0, arrLen);
if (transN % 2 == 0 || opN % 3 > 0) {
ops.push([tableN, idx, true]);
} else {
ops.push([tableN, idx]);
}
}
// De-duplicate operands in a single transaction as that would deadlock
var indicies = [];
var uids = [];
for (var i = 0; i < ops.length; ++i) {
indicies[i] = i;
uids[i] = (ops[i][0] * 1000000000000) + ops[i][1];
}
var uniq = [];
for (opN = 0; opN < ops.length; opN++) {
var isDupl = false;
for (var checkN = 0; checkN < ops.length; checkN++) {
if (opN != checkN && uids[opN] == uids[checkN]) {
isDupl = true;
break;
}
}
if (!isDupl) {
uniq.push(ops[opN]);
}
}
workQ.enqueue(JSON.stringify(uniq));
}
}
//------------------------------------------------------------------
// Simultaneously generate and consume transactions
// Thread 0 enqueues them while the rest dequeue transactions.
// When thread 0 finishes creating new work, it also begins to
// dequeue transactions from the workQ.
//
// If there is nothing on the queue, keep checking until the "DONE"
// message is dequeued.
//
function performTransactions() {
//------------------------------------------------------------------
// Generate the transactions concurrently with their consumption
if (ems.myID == 0) {
var startTime = timerStart();
generateTransactions();
// After all the work has been enqueued, add DONE semaphores to the
// end of the queue so they are processed only after all the work
// has been issued. Each thread enqueues one event and can only
// consume one before exiting.
for (var taskN = 0; taskN < ems.nThreads; taskN++) {
workQ.enqueue("DONE");
}
timerStop(startTime, nTransactions, " transactions enqueued ", ems.myID);
}
//------------------------------------------------------------------
// Process the transactions
var rwNops = 0;
var readNops = 0;
while (true) {
var str = workQ.dequeue();
if (str !== undefined) {
if (str === "DONE") {
break
} else {
var ops = JSON.parse(str);
for (var opN = 0; opN < ops.length; opN++) {
ops[opN][0] = tables[ops[opN][0]];
}
var transaction = ems.tmStart(ops);
ops.forEach(function (op) {
var tmp = op[0].read(op[1]);
if (op[2] != true) {
rwNops++;
op[0].write(op[1], tmp + 1);
} else {
readNops++;
}
});
ems.tmEnd(transaction, true);
}
} else {
// Queue was empty, but have not yet seen DONE event, so
// keep trying to dequeue work
}
}
totalNops.faa(0, rwNops);
totalNops.faa(1, readNops);
}
//=======================================================================
// B E G I N P R O G R A M M A I N
//-------------------------------------------------------
// Initialize the shared data
var startTime = timerStart();
ems.parallel(initializeSharedData);
timerStop(startTime, nTables, " tables initialized ", ems.myID);
// Data to be initialized only once
totalNops.writeXF(0, 0);
totalNops.writeXF(1, 0);
checkNops.writeXF(0, 0);
// Perform all the transactions
startTime = timerStart();
ems.parallel(performTransactions);
timerStop(startTime, nTransactions, " transactions performed", ems.myID);
timerStop(startTime, totalNops.readFF(0), " table updates ", ems.myID);
timerStop(startTime, totalNops.readFF(0) + totalNops.readFF(1), " elements referenced ", ems.myID);
//------------------------------------------------------------------
// Sum all the values in the tables to account for all the transactions
startTime = timerStart();
ems.parallel(function () {
ems.parForEach(0, nTables, function (tableN) {
var localSum = 0;
for (var idx = 0; idx < arrLen; idx++) {
localSum += tables[tableN].read(idx);
}
checkNops.faa(0, localSum);
});
});
timerStop(startTime, nTables * arrLen, " elements checked ", ems.myID);
assert(checkNops.readFF(0) == totalNops.readFF(0),
"Error in final sum = " + checkNops.readFF(0) + " should be=" + totalNops.readFF(0));
console.log("Results are correct");
// One liner to get all processes to exit
ems.parallel(function () { process.exit(0) });