[clang] [CodeGen][OpenMP] Set the default schedule for loops to static,1 if TSan is enabled (PR #68891)
via cfe-commits
cfe-commits at lists.llvm.org
Thu Oct 12 06:44:25 PDT 2023
https://github.com/felilxtomski created https://github.com/llvm/llvm-project/pull/68891
To ease the race detection in OpenMP for loops where no schedule is specified as in the new archer test from DRB.
>From 6f6cba41a85598f28e70fbfdebcd3e5aef4e9df4 Mon Sep 17 00:00:00 2001
From: "felix.tomski" <tomski at itc.rwth-aachen.de>
Date: Thu, 12 Oct 2023 15:03:48 +0200
Subject: [PATCH 1/2] [openmp] Set the default schedule for loops to static,1
if TSan is enabled
To ease the race detection in for loops where no schedule is specified.
---
clang/lib/CodeGen/CGOpenMPRuntime.cpp | 7 ++++---
1 file changed, 4 insertions(+), 3 deletions(-)
diff --git a/clang/lib/CodeGen/CGOpenMPRuntime.cpp b/clang/lib/CodeGen/CGOpenMPRuntime.cpp
index aae1a0ea250eea2..c17d352fdbeb19e 100644
--- a/clang/lib/CodeGen/CGOpenMPRuntime.cpp
+++ b/clang/lib/CodeGen/CGOpenMPRuntime.cpp
@@ -2446,9 +2446,10 @@ unsigned CGOpenMPRuntime::getDefaultFlagsForBarriers(OpenMPDirectiveKind Kind) {
void CGOpenMPRuntime::getDefaultScheduleAndChunk(
CodeGenFunction &CGF, const OMPLoopDirective &S,
OpenMPScheduleClauseKind &ScheduleKind, const Expr *&ChunkExpr) const {
- // Check if the loop directive is actually a doacross loop directive. In this
- // case choose static, 1 schedule.
- if (llvm::any_of(
+ // Check if the loop directive is actually a doacross loop directive or
+ // ThreadSanitizer is enabled. In these cases choose static, 1 schedule.
+ if (CGF.SanOpts.has(SanitizerKind::Thread) ||
+ llvm::any_of(
S.getClausesOfKind<OMPOrderedClause>(),
[](const OMPOrderedClause *C) { return C->getNumForLoops(); })) {
ScheduleKind = OMPC_SCHEDULE_static;
>From 8b25e95008c4d794c4423d026c39382a0c7cfbf8 Mon Sep 17 00:00:00 2001
From: "felix.tomski" <tomski at itc.rwth-aachen.de>
Date: Wed, 11 Oct 2023 16:17:39 +0200
Subject: [PATCH 2/2] [archer] Add DRB test originally requiring 36 threads
with default schedule
---
.../races/DRB006-indirectaccess2-orig-yes.c | 161 ++++++++++++++++++
1 file changed, 161 insertions(+)
create mode 100644 openmp/tools/archer/tests/races/DRB006-indirectaccess2-orig-yes.c
diff --git a/openmp/tools/archer/tests/races/DRB006-indirectaccess2-orig-yes.c b/openmp/tools/archer/tests/races/DRB006-indirectaccess2-orig-yes.c
new file mode 100644
index 000000000000000..c9d1bf0ae0ed9bc
--- /dev/null
+++ b/openmp/tools/archer/tests/races/DRB006-indirectaccess2-orig-yes.c
@@ -0,0 +1,161 @@
+/*
+ * DRB006-indirectaccess2-orig-yes.c -- Archer testcase
+ */
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+//
+// See tools/archer/LICENSE.txt for details.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+/*
+Copyright (c) 2017, Lawrence Livermore National Security, LLC.
+Produced at the Lawrence Livermore National Laboratory
+Written by Chunhua Liao, Pei-Hung Lin, Joshua Asplund,
+Markus Schordan, and Ian Karlin
+(email: liao6 at llnl.gov, lin32 at llnl.gov, asplund1 at llnl.gov,
+schordan1 at llnl.gov, karlin1 at llnl.gov)
+LLNL-CODE-732144
+All rights reserved.
+
+This file is part of DataRaceBench. For details, see
+https://github.com/LLNL/dataracebench. Please also see the LICENSE file
+for our additional BSD notice.
+
+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 disclaimer below.
+
+* Redistributions in binary form must reproduce the above copyright
+ notice, this list of conditions and the disclaimer (as noted below)
+ in the documentation and/or other materials provided with the
+ distribution.
+
+* Neither the name of the LLNS/LLNL 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 LAWRENCE LIVERMORE NATIONAL
+SECURITY, LLC, THE U.S. DEPARTMENT OF ENERGY OR CONTRIBUTORS 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.
+*/
+/*
+Two pointers have a distance of 12 (p1 - p2 = 12).
+They are used as base addresses for indirect array accesses using an index set (another array).
+
+The index set has two indices with a distance of 12 :
+indexSet[5]- indexSet[0] = 533 - 521 = 12
+So there is loop carried dependence (e.g. between loops with index values of 0 and 5).
+
+We use the default loop scheduling (static even) in OpenMP.
+It is possible that two dependent iterations will be scheduled
+within a same chunk to a same thread. So there is no runtime data races.
+
+When N is 180, two iterations with N=0 and N= 5 have loop carried dependencies.
+For static even scheduling, we must have at least 36 threads (180/36=5 iterations)
+so iteration 0 and 5 will be scheduled to two different threads.
+Data race pair: xa1[idx]@128:5:W vs. xa2[idx]@129:5:W
+*/
+
+// RUN: %libarcher-compile-and-run-race | FileCheck %s
+// RUN: %libarcher-compile-and-run-race-noserial | FileCheck %s
+// REQUIRES: tsan
+#include <omp.h>
+#include <stdio.h>
+#include <unistd.h>
+#include <stdlib.h>
+
+#define NUM_THREADS 2
+
+#define N 180
+int indexSet[N] = {
+521, 523, 525, 527, 529, 533, // 521+12=533
+547, 549, 551, 553, 555, 557,
+573, 575, 577, 579, 581, 583,
+599, 601, 603, 605, 607, 609,
+625, 627, 629, 631, 633, 635,
+
+651, 653, 655, 657, 659, 661,
+859, 861, 863, 865, 867, 869,
+885, 887, 889, 891, 893, 895,
+911, 913, 915, 917, 919, 921,
+937, 939, 941, 943, 945, 947,
+
+963, 965, 967, 969, 971, 973,
+989, 991, 993, 995, 997, 999,
+1197, 1199, 1201, 1203, 1205, 1207,
+1223, 1225, 1227, 1229, 1231, 1233,
+1249, 1251, 1253, 1255, 1257, 1259,
+
+1275, 1277, 1279, 1281, 1283, 1285,
+1301, 1303, 1305, 1307, 1309, 1311,
+1327, 1329, 1331, 1333, 1335, 1337,
+1535, 1537, 1539, 1541, 1543, 1545,
+1561, 1563, 1565, 1567, 1569, 1571,
+
+1587, 1589, 1591, 1593, 1595, 1597,
+1613, 1615, 1617, 1619, 1621, 1623,
+1639, 1641, 1643, 1645, 1647, 1649,
+1665, 1667, 1669, 1671, 1673, 1675,
+1873, 1875, 1877, 1879, 1881, 1883,
+
+1899, 1901, 1903, 1905, 1907, 1909,
+1925, 1927, 1929, 1931, 1933, 1935,
+1951, 1953, 1955, 1957, 1959, 1961,
+1977, 1979, 1981, 1983, 1985, 1987,
+2003, 2005, 2007, 2009, 2011, 2013};
+
+int main (int argc, char* argv[])
+{
+ double * base = (double*) malloc(sizeof(double)* (2013+12+1));
+
+ if (base == 0) {
+ printf ("Error in malloc(). Aborting ...\n");
+ return 1;
+ }
+ double * xa1 = base;
+ double * xa2 = xa1 + 12;
+ int i;
+
+ // initialize segments touched by indexSet
+ for (i =521; i<= 2025; ++i) {
+ base[i]=0.5*i;
+ }
+
+#pragma omp parallel for num_threads(NUM_THREADS) //schedule(static, 1) // default static even scheduling may not trigger data race!
+ for (i =0; i< N; ++i) {
+ int idx = indexSet[i];
+ xa1[idx]+= 1.0;
+ xa2[idx]+= 3.0;
+ }
+ printf("x1[999]=%f xa2[1285]=%f\n", xa1[999], xa2[1285]);
+ free (base);
+ fprintf(stderr, "DONE\n");
+ return 0;
+}
+
+// CHECK: WARNING: ThreadSanitizer: data race
+// CHECK-NEXT: {{(Write|Read)}} of size 8
+// CHECK-NEXT: #0 {{.*}}DRB006-indirectaccess2-orig-yes.c:144
+// CHECK: Previous write of size 8
+// CHECK-NEXT: #0 {{.*}}DRB006-indirectaccess2-orig-yes.c:145
+// CHECK: DONE
+// CHECK: ThreadSanitizer: reported {{[0-9]+}} warnings
+
+
More information about the cfe-commits
mailing list