[OpenMP][libomp] Introduce oneAPI compiler support
Introduce KMP_COMPILER_ICX macro to represent compilation with oneAPI
compiler.
Fixup flag detection and compiler ID detection in CMake. Older CMake's
detect IntelLLVM as Clang.
Fix compiler warnings.
Fixup many of the tests to have non-empty parallel regions as they are
elided by oneAPI compiler.
diff --git a/openmp/cmake/DetectTestCompiler/CMakeLists.txt b/openmp/cmake/DetectTestCompiler/CMakeLists.txt
index 7fa32a9..c8afd47 100644
--- a/openmp/cmake/DetectTestCompiler/CMakeLists.txt
+++ b/openmp/cmake/DetectTestCompiler/CMakeLists.txt
@@ -10,6 +10,7 @@
set(information "${information}\\;${CMAKE_${lang}_COMPILER_VERSION}")
set(information "${information}\\;${${lang}_FLAGS}")
set(information "${information}\\;${${lang}_HAS_TSAN_FLAG}")
+ set(information "${information}\\;${${lang}_HAS_OMIT_FRAME_POINTER}")
file(WRITE ${CMAKE_CURRENT_BINARY_DIR}/${lang}CompilerInformation.txt ${information})
endfunction(write_compiler_information)
@@ -40,6 +41,9 @@
add_experimental_isel_flag(CXX)
endif()
+check_c_compiler_flag("-fno-omit-frame-pointer" C_HAS_OMIT_FRAME_POINTER)
+check_cxx_compiler_flag("-fno-omit-frame-pointer" CXX_HAS_OMIT_FRAME_POINTER)
+
SET(CMAKE_REQUIRED_FLAGS "-fsanitize=thread")
check_c_compiler_flag("" C_HAS_TSAN_FLAG)
check_cxx_compiler_flag("" CXX_HAS_TSAN_FLAG)
diff --git a/openmp/cmake/OpenMPTesting.cmake b/openmp/cmake/OpenMPTesting.cmake
index 81732ce..0370b5f 100644
--- a/openmp/cmake/OpenMPTesting.cmake
+++ b/openmp/cmake/OpenMPTesting.cmake
@@ -76,12 +76,14 @@
list(GET information 2 version)
list(GET information 3 openmp_flags)
list(GET information 4 has_tsan_flags)
+ list(GET information 5 has_omit_frame_pointer_flags)
set(OPENMP_TEST_${lang}_COMPILER_PATH ${path})
set(OPENMP_TEST_${lang}_COMPILER_ID ${id})
set(OPENMP_TEST_${lang}_COMPILER_VERSION ${version})
set(OPENMP_TEST_${lang}_COMPILER_OPENMP_FLAGS ${openmp_flags})
set(OPENMP_TEST_${lang}_COMPILER_HAS_TSAN_FLAGS ${has_tsan_flags})
+ set(OPENMP_TEST_${lang}_COMPILER_HAS_OMIT_FRAME_POINTER_FLAGS ${has_omit_frame_pointer_flags})
endmacro()
# Function to set variables with information about the test compiler.
@@ -98,6 +100,7 @@
set(OPENMP_TEST_COMPILER_VERSION "${OPENMP_TEST_C_COMPILER_VERSION}" PARENT_SCOPE)
set(OPENMP_TEST_COMPILER_OPENMP_FLAGS "${OPENMP_TEST_C_COMPILER_OPENMP_FLAGS}" PARENT_SCOPE)
set(OPENMP_TEST_COMPILER_HAS_TSAN_FLAGS "${OPENMP_TEST_C_COMPILER_HAS_TSAN_FLAGS}" PARENT_SCOPE)
+ set(OPENMP_TEST_COMPILER_HAS_OMIT_FRAME_POINTER_FLAGS "${OPENMP_TEST_C_COMPILER_HAS_OMIT_FRAME_POINTER_FLAGS}" PARENT_SCOPE)
# Determine major version.
string(REGEX MATCH "[0-9]+" major "${OPENMP_TEST_C_COMPILER_VERSION}")
@@ -149,6 +152,7 @@
endif()
# TODO: Implement blockaddress in GlobalISel and remove this flag!
set(OPENMP_TEST_COMPILER_OPENMP_FLAGS "-fopenmp ${OPENMP_TEST_COMPILER_THREAD_FLAGS} -fno-experimental-isel")
+ set(OPENMP_TEST_COMPILER_HAS_OMIT_FRAME_POINTER_FLAGS 1)
endif()
# Function to set compiler features for use in lit.
diff --git a/openmp/cmake/config-ix.cmake b/openmp/cmake/config-ix.cmake
index b7c254a..d651523 100644
--- a/openmp/cmake/config-ix.cmake
+++ b/openmp/cmake/config-ix.cmake
@@ -1,4 +1,20 @@
include(CheckCXXCompilerFlag)
+include(CheckCXXSourceCompiles)
+
+# Check for oneAPI compiler (some older CMake versions detect as Clang)
+if (CMAKE_C_COMPILER_ID STREQUAL "Clang")
+ check_cxx_source_compiles("#if (defined(__INTEL_CLANG_COMPILER) || defined(__INTEL_LLVM_COMPILER))
+ int main() { return 0; }
+ #else
+ not oneAPI
+ #endif" OPENMP_HAVE_ONEAPI_COMPILER)
+ if (OPENMP_HAVE_ONEAPI_COMPILER)
+ # According to CMake documentation, the compiler id should
+ # be IntelLLVM when detected oneAPI
+ set(CMAKE_C_COMPILER_ID "IntelLLVM")
+ set(CMAKE_CXX_COMPILER_ID "IntelLLVM")
+ endif()
+endif()
check_cxx_compiler_flag(-Wall OPENMP_HAVE_WALL_FLAG)
check_cxx_compiler_flag(-Werror OPENMP_HAVE_WERROR_FLAG)
diff --git a/openmp/runtime/cmake/LibompCheckLinkerFlag.cmake b/openmp/runtime/cmake/LibompCheckLinkerFlag.cmake
index 4c30514..bcc0ca80 100644
--- a/openmp/runtime/cmake/LibompCheckLinkerFlag.cmake
+++ b/openmp/runtime/cmake/LibompCheckLinkerFlag.cmake
@@ -25,7 +25,7 @@
add_library(foo SHARED src_to_link.c)")
# Compiling as a part of runtimes introduces ARCH-unknown-linux-gnu as a part
# of a working directory. So adding a guard for unknown.
- set(failed_regexes "[Ee]rror;[Uu]nknown[^-];[Ss]kipping;LINK : warning")
+ set(failed_regexes "[Ee]rror;[Uu]nknown[^-];[Ss]kipping;LINK : warning;Unsupported command line")
set(base_dir ${CMAKE_BINARY_DIR}${CMAKE_FILES_DIRECTORY}/CMakeTmp/link_flag_check_${boolean})
file(MAKE_DIRECTORY ${base_dir})
file(MAKE_DIRECTORY ${base_dir}/build)
diff --git a/openmp/runtime/cmake/config-ix.cmake b/openmp/runtime/cmake/config-ix.cmake
index 0fe1266..d91e852 100644
--- a/openmp/runtime/cmake/config-ix.cmake
+++ b/openmp/runtime/cmake/config-ix.cmake
@@ -139,7 +139,7 @@
endif()
# Check Intel(R) C Compiler specific flags
-if(CMAKE_C_COMPILER_ID STREQUAL "Intel")
+if(CMAKE_C_COMPILER_ID STREQUAL "Intel" OR CMAKE_C_COMPILER_ID STREQUAL "IntelLLVM")
check_cxx_compiler_flag(/Qlong_double LIBOMP_HAVE_LONG_DOUBLE_FLAG)
check_cxx_compiler_flag(/Qdiag-disable:177 LIBOMP_HAVE_DIAG_DISABLE_177_FLAG)
check_cxx_compiler_flag(/Qinline-min-size=1 LIBOMP_HAVE_INLINE_MIN_SIZE_FLAG)
@@ -247,7 +247,7 @@
# Check if quad precision types are available
if(CMAKE_C_COMPILER_ID STREQUAL "GNU")
set(LIBOMP_HAVE_QUAD_PRECISION TRUE)
-elseif(CMAKE_C_COMPILER_ID STREQUAL "Intel")
+elseif(CMAKE_C_COMPILER_ID STREQUAL "Intel" OR CMAKE_C_COMPILER_ID STREQUAL "IntelLLVM")
if(LIBOMP_HAVE_EXTENDED_FLOAT_TYPES_FLAG)
set(LIBOMP_HAVE_QUAD_PRECISION TRUE)
else()
diff --git a/openmp/runtime/src/kmp.h b/openmp/runtime/src/kmp.h
index a41265a..6071fb1 100644
--- a/openmp/runtime/src/kmp.h
+++ b/openmp/runtime/src/kmp.h
@@ -1124,7 +1124,7 @@
#if KMP_OS_UNIX && (KMP_ARCH_X86 || KMP_ARCH_X86_64)
// HW TSC is used to reduce overhead (clock tick instead of nanosecond).
extern kmp_uint64 __kmp_ticks_per_msec;
-#if KMP_COMPILER_ICC
+#if KMP_COMPILER_ICC || KMP_COMPILER_ICX
#define KMP_NOW() ((kmp_uint64)_rdtsc())
#else
#define KMP_NOW() __kmp_hardware_timestamp()
diff --git a/openmp/runtime/src/kmp_affinity.cpp b/openmp/runtime/src/kmp_affinity.cpp
index d1f1b67..8a205a9 100644
--- a/openmp/runtime/src/kmp_affinity.cpp
+++ b/openmp/runtime/src/kmp_affinity.cpp
@@ -1765,7 +1765,7 @@
hw_thread_index = 0;
pu = NULL;
- while (pu = hwloc_get_next_obj_by_type(tp, HWLOC_OBJ_PU, pu)) {
+ while ((pu = hwloc_get_next_obj_by_type(tp, HWLOC_OBJ_PU, pu))) {
int index = depth - 1;
bool included = KMP_CPU_ISSET(pu->os_index, __kmp_affin_fullMask);
kmp_hw_thread_t &hw_thread = __kmp_topology->at(hw_thread_index);
diff --git a/openmp/runtime/src/kmp_atomic.cpp b/openmp/runtime/src/kmp_atomic.cpp
index 0bd7b1a..21c2c60 100644
--- a/openmp/runtime/src/kmp_atomic.cpp
+++ b/openmp/runtime/src/kmp_atomic.cpp
@@ -2452,6 +2452,7 @@
RTYPE, LCK_ID, MASK, GOMP_FLAG) \
ATOMIC_BEGIN_CPT_MIX(TYPE_ID, OP_ID, TYPE, RTYPE_ID, RTYPE) \
TYPE new_value; \
+ (void)new_value; \
OP_GOMP_CRITICAL_CPT(TYPE, OP, GOMP_FLAG) \
OP_CMPXCHG_CPT(TYPE, BITS, OP) \
}
@@ -2461,6 +2462,7 @@
LCK_ID, GOMP_FLAG) \
ATOMIC_BEGIN_CPT_MIX(TYPE_ID, OP_ID, TYPE, RTYPE_ID, RTYPE) \
TYPE new_value; \
+ (void)new_value; \
OP_GOMP_CRITICAL_CPT(TYPE, OP, GOMP_FLAG) /* send assignment */ \
OP_UPDATE_CRITICAL_CPT(TYPE, OP, LCK_ID) /* send assignment */ \
}
@@ -3162,6 +3164,7 @@
RTYPE, LCK_ID, MASK, GOMP_FLAG) \
ATOMIC_BEGIN_CPT_MIX(TYPE_ID, OP_ID, TYPE, RTYPE_ID, RTYPE) \
TYPE new_value; \
+ (void)new_value; \
OP_GOMP_CRITICAL_CPT_REV(TYPE, OP, GOMP_FLAG) \
OP_CMPXCHG_CPT_REV(TYPE, BITS, OP) \
}
@@ -3171,6 +3174,7 @@
LCK_ID, GOMP_FLAG) \
ATOMIC_BEGIN_CPT_MIX(TYPE_ID, OP_ID, TYPE, RTYPE_ID, RTYPE) \
TYPE new_value; \
+ (void)new_value; \
OP_GOMP_CRITICAL_CPT_REV(TYPE, OP, GOMP_FLAG) /* send assignment */ \
OP_CRITICAL_CPT_REV(TYPE, OP, LCK_ID) /* send assignment */ \
}
diff --git a/openmp/runtime/src/kmp_atomic.h b/openmp/runtime/src/kmp_atomic.h
index 079b917..19c02e9 100644
--- a/openmp/runtime/src/kmp_atomic.h
+++ b/openmp/runtime/src/kmp_atomic.h
@@ -251,6 +251,9 @@
kmp_cmplx128_a4_t() : q() {}
+#if defined(__cplusplus) && (KMP_OS_WINDOWS)
+ kmp_cmplx128_a4_t(const std::complex<_Quad> &c128) : q(c128) {}
+#endif
kmp_cmplx128_a4_t(const kmp_cmplx128 &c128) : q(c128) {}
kmp_cmplx128_a4_t operator+(const kmp_cmplx128_a4_t &b) {
@@ -314,6 +317,9 @@
kmp_cmplx128_a16_t() : q() {}
+#if defined(__cplusplus) && (KMP_OS_WINDOWS)
+ kmp_cmplx128_a16_t(const std::complex<_Quad> &c128) : q(c128) {}
+#endif
kmp_cmplx128_a16_t(const kmp_cmplx128 &c128) : q(c128) {}
kmp_cmplx128_a16_t operator+(const kmp_cmplx128_a16_t &b) {
diff --git a/openmp/runtime/src/kmp_barrier.cpp b/openmp/runtime/src/kmp_barrier.cpp
index ee05bb3..1a718b4 100644
--- a/openmp/runtime/src/kmp_barrier.cpp
+++ b/openmp/runtime/src/kmp_barrier.cpp
@@ -2163,7 +2163,6 @@
kmp_info_t *this_thr = __kmp_threads[gtid];
kmp_team_t *team;
- kmp_uint nproc;
int tid;
#ifdef KMP_DEBUG
int team_id;
@@ -2176,12 +2175,14 @@
itt_sync_obj = __kmp_itt_barrier_object(gtid, bs_forkjoin_barrier);
#endif
#endif /* USE_ITT_BUILD */
+#if ((USE_ITT_BUILD && USE_ITT_NOTIFY) || defined KMP_DEBUG)
+ int nproc = this_thr->th.th_team_nproc;
+#endif
KMP_MB();
// Get current info
team = this_thr->th.th_team;
- nproc = this_thr->th.th_team_nproc;
- KMP_DEBUG_ASSERT((int)nproc == team->t.t_nproc);
+ KMP_DEBUG_ASSERT(nproc == team->t.t_nproc);
tid = __kmp_tid_from_gtid(gtid);
#ifdef KMP_DEBUG
team_id = team->t.t_id;
@@ -2354,7 +2355,7 @@
// Set arrive time to zero to be able to check it in
// __kmp_invoke_task(); the same is done inside the loop below
this_thr->th.th_bar_arrive_time = 0;
- for (kmp_uint i = 1; i < nproc; ++i) {
+ for (int i = 1; i < nproc; ++i) {
delta += (cur_time - other_threads[i]->th.th_bar_arrive_time);
other_threads[i]->th.th_bar_arrive_time = 0;
}
diff --git a/openmp/runtime/src/kmp_csupport.cpp b/openmp/runtime/src/kmp_csupport.cpp
index e263558..d978044 100644
--- a/openmp/runtime/src/kmp_csupport.cpp
+++ b/openmp/runtime/src/kmp_csupport.cpp
@@ -685,13 +685,13 @@
if (!__kmp_cpuinfo.flags.sse2) {
// CPU cannot execute SSE2 instructions.
} else {
-#if KMP_COMPILER_ICC
+#if KMP_COMPILER_ICC || KMP_COMPILER_ICX
_mm_mfence();
#elif KMP_COMPILER_MSVC
MemoryBarrier();
#else
__sync_synchronize();
-#endif // KMP_COMPILER_ICC
+#endif // KMP_COMPILER_ICC || KMP_COMPILER_ICX
}
#endif // KMP_MIC
#elif (KMP_ARCH_ARM || KMP_ARCH_AARCH64 || KMP_ARCH_MIPS || KMP_ARCH_MIPS64 || \
diff --git a/openmp/runtime/src/kmp_debugger.cpp b/openmp/runtime/src/kmp_debugger.cpp
index 2a1f633..0abdb0f 100644
--- a/openmp/runtime/src/kmp_debugger.cpp
+++ b/openmp/runtime/src/kmp_debugger.cpp
@@ -226,16 +226,16 @@
when 64-bit value is assigned to 32-bit pointer. Use this function
to suppress the warning. */
static inline void *__kmp_convert_to_ptr(kmp_uint64 addr) {
-#if KMP_COMPILER_ICC
+#if KMP_COMPILER_ICC || KMP_COMPILER_ICX
#pragma warning(push)
#pragma warning(disable : 810) // conversion from "unsigned long long" to "char
// *" may lose significant bits
#pragma warning(disable : 1195) // conversion from integer to smaller pointer
-#endif // KMP_COMPILER_ICC
+#endif // KMP_COMPILER_ICC || KMP_COMPILER_ICX
return (void *)addr;
-#if KMP_COMPILER_ICC
+#if KMP_COMPILER_ICC || KMP_COMPILER_ICX
#pragma warning(pop)
-#endif // KMP_COMPILER_ICC
+#endif // KMP_COMPILER_ICC || KMP_COMPILER_ICX
} // __kmp_convert_to_ptr
static int kmp_location_match(kmp_str_loc_t *loc, kmp_omp_nthr_item_t *item) {
diff --git a/openmp/runtime/src/kmp_lock.cpp b/openmp/runtime/src/kmp_lock.cpp
index fff7305..8fcddc7 100644
--- a/openmp/runtime/src/kmp_lock.cpp
+++ b/openmp/runtime/src/kmp_lock.cpp
@@ -1954,7 +1954,7 @@
// We need a fence here, since we must ensure that no memory operations
// from later in this thread float above that read.
-#if KMP_COMPILER_ICC
+#if KMP_COMPILER_ICC || KMP_COMPILER_ICX
_mm_mfence();
#else
__sync_synchronize();
diff --git a/openmp/runtime/src/kmp_os.h b/openmp/runtime/src/kmp_os.h
index d71e9ae..0bc40c7 100644
--- a/openmp/runtime/src/kmp_os.h
+++ b/openmp/runtime/src/kmp_os.h
@@ -53,8 +53,12 @@
#define KMP_COMPILER_GCC 0
#define KMP_COMPILER_CLANG 0
#define KMP_COMPILER_MSVC 0
+#define KMP_COMPILER_ICX 0
-#if defined(__INTEL_COMPILER)
+#if __INTEL_CLANG_COMPILER
+#undef KMP_COMPILER_ICX
+#define KMP_COMPILER_ICX 1
+#elif defined(__INTEL_COMPILER)
#undef KMP_COMPILER_ICC
#define KMP_COMPILER_ICC 1
#elif defined(__clang__)
@@ -85,7 +89,7 @@
/* Check for quad-precision extension. */
#define KMP_HAVE_QUAD 0
#if KMP_ARCH_X86 || KMP_ARCH_X86_64
-#if KMP_COMPILER_ICC
+#if KMP_COMPILER_ICC || KMP_COMPILER_ICX
/* _Quad is already defined for icc */
#undef KMP_HAVE_QUAD
#define KMP_HAVE_QUAD 1
@@ -448,8 +452,10 @@
#pragma intrinsic(InterlockedExchangeAdd)
#pragma intrinsic(InterlockedCompareExchange)
#pragma intrinsic(InterlockedExchange)
+#if !(KMP_COMPILER_ICX && KMP_32_BIT_ARCH)
#pragma intrinsic(InterlockedExchange64)
#endif
+#endif
// Using InterlockedIncrement / InterlockedDecrement causes a library loading
// ordering problem, so we use InterlockedExchangeAdd instead.
@@ -842,8 +848,14 @@
(kmp_uint64)(sv))
#endif
+#if KMP_OS_DARWIN && defined(__INTEL_COMPILER) && __INTEL_COMPILER >= 1800
+#define KMP_XCHG_FIXED8(p, v) \
+ __atomic_exchange_1((volatile kmp_uint8 *)(p), (kmp_uint8)(v), \
+ __ATOMIC_SEQ_CST)
+#else
#define KMP_XCHG_FIXED8(p, v) \
__sync_lock_test_and_set((volatile kmp_uint8 *)(p), (kmp_uint8)(v))
+#endif
#define KMP_XCHG_FIXED16(p, v) \
__sync_lock_test_and_set((volatile kmp_uint16 *)(p), (kmp_uint16)(v))
#define KMP_XCHG_FIXED32(p, v) \
@@ -1026,7 +1038,7 @@
#endif
#if KMP_ARCH_X86 || KMP_ARCH_X86_64
-#if KMP_COMPILER_ICC
+#if KMP_COMPILER_ICC || KMP_COMPILER_ICX
#define KMP_MFENCE_() _mm_mfence()
#define KMP_SFENCE_() _mm_sfence()
#elif KMP_COMPILER_MSVC
diff --git a/openmp/runtime/src/kmp_runtime.cpp b/openmp/runtime/src/kmp_runtime.cpp
index e1af2f4..abb1d85 100644
--- a/openmp/runtime/src/kmp_runtime.cpp
+++ b/openmp/runtime/src/kmp_runtime.cpp
@@ -8953,19 +8953,16 @@
KMP_DEBUG_ASSERT(team->t.t_threads[f]->th.th_used_in_team.load() == 2);
}
// Release all the workers
- kmp_uint64 new_value; // new value for go
- new_value = team->t.b->go_release();
+ team->t.b->go_release();
KMP_MFENCE();
// Workers should see transition status 2 and move to 0; but may need to be
// woken up first
- size_t my_go_index;
int count = old_nthreads - 1;
while (count > 0) {
count = old_nthreads - 1;
for (int f = 1; f < old_nthreads; ++f) {
- my_go_index = f / team->t.b->threads_per_go;
if (other_threads[f]->th.th_used_in_team.load() != 0) {
if (__kmp_dflt_blocktime != KMP_MAX_BLOCKTIME) { // Wake up the workers
kmp_atomic_flag_64<> *flag = (kmp_atomic_flag_64<> *)CCAST(
diff --git a/openmp/runtime/src/kmp_stub.cpp b/openmp/runtime/src/kmp_stub.cpp
index 6aaec68..da7340f 100644
--- a/openmp/runtime/src/kmp_stub.cpp
+++ b/openmp/runtime/src/kmp_stub.cpp
@@ -159,11 +159,11 @@
}
void *kmp_aligned_malloc(size_t sz, size_t a) {
i;
- int err;
void *res;
#if KMP_OS_WINDOWS
res = _aligned_malloc(sz, a);
#else
+ int err;
if ((err = posix_memalign(&res, a, sz))) {
errno = err; // can be EINVAL or ENOMEM
res = NULL;
@@ -393,12 +393,12 @@
void *omp_aligned_alloc(size_t a, size_t size, omp_allocator_handle_t al) {
i;
- int err;
void *res;
#if KMP_OS_WINDOWS
res = _aligned_malloc(size, a);
#else
- if (err = posix_memalign(&res, a, size)) {
+ int err;
+ if ((err = posix_memalign(&res, a, size))) {
errno = err; // can be EINVAL or ENOMEM
res = NULL;
}
@@ -420,12 +420,12 @@
void *omp_aligned_calloc(size_t a, size_t nmemb, size_t size,
omp_allocator_handle_t al) {
i;
- int err;
void *res;
#if KMP_OS_WINDOWS
res = _aligned_recalloc(NULL, nmemb, size, a);
#else
- if (err = posix_memalign(&res, a, nmemb * size)) {
+ int err;
+ if ((err = posix_memalign(&res, a, nmemb * size))) {
errno = err; // can be EINVAL or ENOMEM
res = NULL;
}
diff --git a/openmp/runtime/src/kmp_version.cpp b/openmp/runtime/src/kmp_version.cpp
index db2454c..bb600c1 100644
--- a/openmp/runtime/src/kmp_version.cpp
+++ b/openmp/runtime/src/kmp_version.cpp
@@ -22,7 +22,9 @@
#define stringer(x) _stringer(x)
// Detect compiler.
-#if KMP_COMPILER_ICC
+#if KMP_COMPILER_ICX
+#define KMP_COMPILER __VERSION__
+#elif KMP_COMPILER_ICC
#if __INTEL_COMPILER == 1010
#define KMP_COMPILER "Intel(R) C++ Compiler 10.1"
#elif __INTEL_COMPILER == 1100
@@ -53,8 +55,10 @@
#define KMP_COMPILER "Intel(R) C++ Compiler 19.0"
#elif __INTEL_COMPILER == 1910
#define KMP_COMPILER "Intel(R) C++ Compiler 19.1"
-#elif __INTEL_COMPILER >= 9900
-#define KMP_COMPILER "Intel(R) C++ Compiler mainline"
+#elif __INTEL_COMPILER > 1910
+#define KMP_COMPILER \
+ "Intel(R) C++ Compiler Classic " stringer(__INTEL_COMPILER) "." stringer( \
+ __INTEL_COMPILER_UPDATE)
#endif
#elif KMP_COMPILER_CLANG
#define KMP_COMPILER \
diff --git a/openmp/runtime/src/ompt-general.cpp b/openmp/runtime/src/ompt-general.cpp
index c1468c0..1bc3bea 100644
--- a/openmp/runtime/src/ompt-general.cpp
+++ b/openmp/runtime/src/ompt-general.cpp
@@ -310,7 +310,8 @@
OMPT_VERBOSE_INIT_PRINT("Opening %s... ", fname);
HMODULE h = LoadLibrary(fname);
if (!h) {
- OMPT_VERBOSE_INIT_CONTINUED_PRINT("Failed: Error %u\n", GetLastError());
+ OMPT_VERBOSE_INIT_CONTINUED_PRINT("Failed: Error %u\n",
+ (unsigned)GetLastError());
} else {
OMPT_VERBOSE_INIT_CONTINUED_PRINT("Success. \n");
OMPT_VERBOSE_INIT_PRINT("Searching for ompt_start_tool in %s... ",
@@ -318,7 +319,7 @@
start_tool = (ompt_start_tool_t)GetProcAddress(h, "ompt_start_tool");
if (!start_tool) {
OMPT_VERBOSE_INIT_CONTINUED_PRINT("Failed: Error %u\n",
- GetLastError());
+ (unsigned)GetLastError());
} else
#else
#error Activation of OMPT is not supported on this platform.
diff --git a/openmp/runtime/src/z_Windows_NT_util.cpp b/openmp/runtime/src/z_Windows_NT_util.cpp
index 8fa1985..9c77e28 100644
--- a/openmp/runtime/src/z_Windows_NT_util.cpp
+++ b/openmp/runtime/src/z_Windows_NT_util.cpp
@@ -568,7 +568,8 @@
if (__kmp_init_gtid) {
KA_TRACE(50, ("__kmp_gtid_set_specific: T#%d key:%d\n", gtid,
__kmp_gtid_threadprivate_key));
- if (!TlsSetValue(__kmp_gtid_threadprivate_key, (LPVOID)(gtid + 1)))
+ kmp_intptr_t g = (kmp_intptr_t)gtid;
+ if (!TlsSetValue(__kmp_gtid_threadprivate_key, (LPVOID)(g + 1)))
KMP_FATAL(TLSSetValueFailed);
} else {
KA_TRACE(50, ("__kmp_gtid_set_specific: runtime shutdown, returning\n"));
@@ -934,9 +935,8 @@
}
void __kmp_clear_system_time(void) {
- BOOL status;
LARGE_INTEGER time;
- status = QueryPerformanceCounter(&time);
+ QueryPerformanceCounter(&time);
__kmp_win32_time = (kmp_int64)time.QuadPart;
}
@@ -960,9 +960,8 @@
/* Calculate the elapsed wall clock time for the user */
void __kmp_elapsed(double *t) {
- BOOL status;
LARGE_INTEGER now;
- status = QueryPerformanceCounter(&now);
+ QueryPerformanceCounter(&now);
*t = ((double)now.QuadPart) * __kmp_win32_tick;
}
@@ -972,11 +971,8 @@
void __kmp_read_system_time(double *delta) {
if (delta != NULL) {
- BOOL status;
LARGE_INTEGER now;
-
- status = QueryPerformanceCounter(&now);
-
+ QueryPerformanceCounter(&now);
*delta = ((double)(((kmp_int64)now.QuadPart) - __kmp_win32_time)) *
__kmp_win32_tick;
}
@@ -1020,6 +1016,7 @@
if (__kmp_stkoffset > 0 && gtid > 0) {
padding = KMP_ALLOCA(gtid * __kmp_stkoffset);
+ (void)padding;
}
KMP_FSYNC_RELEASING(&this_thr->th.th_info.ds.ds_alive);
@@ -1354,9 +1351,10 @@
/* NOTE: The ExitProcess(code) system call causes all threads to Terminate
with a exit_val = code. Because of this we can not rely on exit_val having
any particular value. */
+ kmp_intptr_t e = (kmp_intptr_t)exit_val;
if (exit_val == STILL_ACTIVE) {
KA_TRACE(1, ("__kmp_reap_common: thread still active.\n"));
- } else if ((void *)exit_val != (void *)th) {
+ } else if ((void *)e != (void *)th) {
KA_TRACE(1, ("__kmp_reap_common: ExitProcess / TerminateThread used?\n"));
}
@@ -1519,13 +1517,12 @@
// Determine whether the given address is mapped into the current address space.
int __kmp_is_address_mapped(void *addr) {
- DWORD status;
MEMORY_BASIC_INFORMATION lpBuffer;
SIZE_T dwLength;
dwLength = sizeof(MEMORY_BASIC_INFORMATION);
- status = VirtualQuery(addr, &lpBuffer, dwLength);
+ VirtualQuery(addr, &lpBuffer, dwLength);
return !(((lpBuffer.State == MEM_RESERVE) || (lpBuffer.State == MEM_FREE)) ||
((lpBuffer.Protect == PAGE_NOACCESS) ||
diff --git a/openmp/runtime/test/CMakeLists.txt b/openmp/runtime/test/CMakeLists.txt
index e31ca52..aa9a373 100644
--- a/openmp/runtime/test/CMakeLists.txt
+++ b/openmp/runtime/test/CMakeLists.txt
@@ -30,6 +30,7 @@
pythonize_bool(LIBOMP_HAVE_LIBM)
pythonize_bool(LIBOMP_HAVE_LIBATOMIC)
pythonize_bool(OPENMP_STANDALONE_BUILD)
+pythonize_bool(OPENMP_TEST_COMPILER_HAS_OMIT_FRAME_POINTER_FLAGS)
add_library(ompt-print-callback INTERFACE)
target_include_directories(ompt-print-callback INTERFACE ${CMAKE_CURRENT_SOURCE_DIR}/ompt)
diff --git a/openmp/runtime/test/affinity/format/increase.c b/openmp/runtime/test/affinity/format/increase.c
index b3942db..875d2f0 100644
--- a/openmp/runtime/test/affinity/format/increase.c
+++ b/openmp/runtime/test/affinity/format/increase.c
@@ -3,30 +3,22 @@
#include <stdio.h>
#include <stdlib.h>
#include <omp.h>
+#include "omp_testsuite.h"
int main(int argc, char** argv) {
omp_set_affinity_format("TESTER: tl:%L tn:%n nt:%N");
// should print all for first parallel
- omp_set_num_threads(4);
- #pragma omp parallel
- { }
+ go_parallel_nthreads(4);
// should print all because of new threads
- omp_set_num_threads(8);
- #pragma omp parallel
- { }
+ go_parallel_nthreads(8);
// should not print anything here
- omp_set_num_threads(6);
- #pragma omp parallel
- { }
+ go_parallel_nthreads(6);
// should print all because of new thread
- omp_set_num_threads(9);
- #pragma omp parallel
- { }
+ go_parallel_nthreads(9);
// should not print anything here
- omp_set_num_threads(2);
- #pragma omp parallel
- { }
- return 0;
+ go_parallel_nthreads(2);
+
+ return get_exit_value();
}
// CHECK: num_threads=4 TESTER: tl:1 tn:[0-3] nt:4
diff --git a/openmp/runtime/test/affinity/format/nested.c b/openmp/runtime/test/affinity/format/nested.c
index 2ecc918..db2e607 100644
--- a/openmp/runtime/test/affinity/format/nested.c
+++ b/openmp/runtime/test/affinity/format/nested.c
@@ -4,16 +4,16 @@
#include <stdio.h>
#include <stdlib.h>
#include <omp.h>
+#include "omp_testsuite.h"
int main(int argc, char** argv) {
omp_set_affinity_format("TESTER: tl:%L at:%a tn:%n nt:%N");
omp_set_nested(1);
#pragma omp parallel num_threads(4)
{
- #pragma omp parallel num_threads(3)
- { }
+ go_parallel_nthreads(3);
}
- return 0;
+ return get_exit_value();
}
// CHECK: num_threads=4 TESTER: tl:1 at:0 tn:[0-3] nt:4
diff --git a/openmp/runtime/test/affinity/format/nested2.c b/openmp/runtime/test/affinity/format/nested2.c
index 4b54912..f259aea 100644
--- a/openmp/runtime/test/affinity/format/nested2.c
+++ b/openmp/runtime/test/affinity/format/nested2.c
@@ -3,6 +3,7 @@
#include <stdio.h>
#include <stdlib.h>
#include <omp.h>
+#include "omp_testsuite.h"
// Currently, KMP_HOT_TEAMS_MAX_LEVEL has to be equal to the
// nest depth for intuitive behavior
@@ -11,14 +12,11 @@
omp_set_nested(1);
#pragma omp parallel num_threads(4)
{
- #pragma omp parallel num_threads(3)
- { }
- #pragma omp parallel num_threads(3)
- { }
+ go_parallel_nthreads(3);
+ go_parallel_nthreads(3);
}
- #pragma omp parallel num_threads(4)
- { }
- return 0;
+ go_parallel_nthreads(4);
+ return get_exit_value();
}
// CHECK: num_threads=4 TESTER: tl:1 tn:[0-3] nt:4
diff --git a/openmp/runtime/test/affinity/format/nested_mixed.c b/openmp/runtime/test/affinity/format/nested_mixed.c
index 1e4c753..288e1c2 100644
--- a/openmp/runtime/test/affinity/format/nested_mixed.c
+++ b/openmp/runtime/test/affinity/format/nested_mixed.c
@@ -3,30 +3,26 @@
#include <stdio.h>
#include <stdlib.h>
#include <omp.h>
+#include "omp_testsuite.h"
int main(int argc, char** argv) {
omp_set_affinity_format("TESTER: tl:%L at:%a tn:%n nt:%N");
omp_set_nested(1);
#pragma omp parallel num_threads(1)
{
- #pragma omp parallel num_threads(2)
- { }
+ go_parallel_nthreads(2);
#pragma omp parallel num_threads(2)
{
#pragma omp parallel num_threads(1)
{
- #pragma omp parallel num_threads(2)
- { }
+ go_parallel_nthreads(2);
}
}
- #pragma omp parallel num_threads(1)
- { }
+ go_parallel_nthreads(1);
}
- #pragma omp parallel num_threads(2)
- { }
- #pragma omp parallel num_threads(1)
- { }
- return 0;
+ go_parallel_nthreads(2);
+ go_parallel_nthreads(1);
+ return get_exit_value();
}
// CHECK: num_threads=1 TESTER: tl:1 at:0 tn:0 nt:1
diff --git a/openmp/runtime/test/affinity/format/nested_serial.c b/openmp/runtime/test/affinity/format/nested_serial.c
index 8b84ba6..70ccf5b 100644
--- a/openmp/runtime/test/affinity/format/nested_serial.c
+++ b/openmp/runtime/test/affinity/format/nested_serial.c
@@ -3,29 +3,24 @@
#include <stdio.h>
#include <stdlib.h>
#include <omp.h>
+#include "omp_testsuite.h"
int main(int argc, char** argv) {
omp_set_affinity_format("TESTER: tl:%L at:%a tn:%n nt:%N");
omp_set_nested(1);
#pragma omp parallel num_threads(1)
{
- #pragma omp parallel num_threads(1)
- { }
- #pragma omp parallel num_threads(1)
- { }
+ go_parallel_nthreads(1);
+ go_parallel_nthreads(1);
#pragma omp parallel num_threads(1)
{
- #pragma omp parallel num_threads(1)
- { }
+ go_parallel_nthreads(1);
}
- #pragma omp parallel num_threads(1)
- { }
+ go_parallel_nthreads(1);
}
- #pragma omp parallel num_threads(1)
- { }
- #pragma omp parallel num_threads(1)
- { }
- return 0;
+ go_parallel_nthreads(1);
+ go_parallel_nthreads(1);
+ return get_exit_value();
}
// CHECK: num_threads=1 TESTER: tl:1 at:0 tn:0 nt:1
diff --git a/openmp/runtime/test/affinity/format/proc_bind.c b/openmp/runtime/test/affinity/format/proc_bind.c
index 765c3ce..815f20f 100644
--- a/openmp/runtime/test/affinity/format/proc_bind.c
+++ b/openmp/runtime/test/affinity/format/proc_bind.c
@@ -4,26 +4,21 @@
#include <stdio.h>
#include <stdlib.h>
#include <omp.h>
+#include "omp_testsuite.h"
int main(int argc, char** argv) {
omp_set_affinity_format("TESTER: tl:%L tn:%n nt:%N aff:{%A}");
omp_set_num_threads(8);
// Initial parallel
- #pragma omp parallel proc_bind(spread)
- { }
- #pragma omp parallel proc_bind(spread)
- { }
+ go_parallel_spread();
+ go_parallel_spread();
// Affinity changes here
- #pragma omp parallel proc_bind(close)
- { }
- #pragma omp parallel proc_bind(close)
- { }
+ go_parallel_close();
+ go_parallel_close();
// Affinity changes here
- #pragma omp parallel proc_bind(master)
- { }
- #pragma omp parallel proc_bind(master)
- { }
- return 0;
+ go_parallel_master();
+ go_parallel_master();
+ return get_exit_value();
}
// CHECK: num_threads=8 TESTER: tl:1 tn:[0-7] nt:8 aff:
diff --git a/openmp/runtime/test/affinity/format/simple.c b/openmp/runtime/test/affinity/format/simple.c
index 701c207..c0a48553 100644
--- a/openmp/runtime/test/affinity/format/simple.c
+++ b/openmp/runtime/test/affinity/format/simple.c
@@ -9,14 +9,13 @@
#include <stdio.h>
#include <stdlib.h>
#include <omp.h>
+#include "omp_testsuite.h"
int main(int argc, char** argv) {
omp_set_affinity_format("TESTER: tl:%L tn:%n nt:%N");
- #pragma omp parallel
- { }
- #pragma omp parallel
- { }
- return 0;
+ go_parallel();
+ go_parallel();
+ return get_exit_value();
}
// NOTHING: NO_OUTPUT
diff --git a/openmp/runtime/test/affinity/format/simple_env.c b/openmp/runtime/test/affinity/format/simple_env.c
index f369e9f..850066d 100644
--- a/openmp/runtime/test/affinity/format/simple_env.c
+++ b/openmp/runtime/test/affinity/format/simple_env.c
@@ -4,13 +4,12 @@
#include <stdio.h>
#include <stdlib.h>
#include <omp.h>
+#include "omp_testsuite.h"
int main(int argc, char** argv) {
- #pragma omp parallel
- { }
- #pragma omp parallel
- { }
- return 0;
+ go_parallel();
+ go_parallel();
+ return get_exit_value();
}
// CHECK-8: num_threads=8 TESTER-ENV: tl:1 tn:[0-7] nt:8$
diff --git a/openmp/runtime/test/affinity/omp-places-invalid-syntax.c b/openmp/runtime/test/affinity/omp-places-invalid-syntax.c
index 31a3185..1157bb6 100644
--- a/openmp/runtime/test/affinity/omp-places-invalid-syntax.c
+++ b/openmp/runtime/test/affinity/omp-places-invalid-syntax.c
@@ -4,8 +4,9 @@
// CHECK-SAME: cores
// REQUIRES: affinity
+#include "omp_testsuite.h"
+
int main() {
- #pragma omp parallel
- {}
- return 0;
+ go_parallel();
+ return get_exit_value();
}
diff --git a/openmp/runtime/test/lit.cfg b/openmp/runtime/test/lit.cfg
index c1d240e..76727f3 100644
--- a/openmp/runtime/test/lit.cfg
+++ b/openmp/runtime/test/lit.cfg
@@ -45,9 +45,13 @@
flags = " -I " + config.test_source_root + \
" -L " + config.library_dir + \
" " + config.test_extra_flags
+if config.has_omit_frame_pointer_flag:
+ flags += " -fno-omit-frame-pointer"
+
config.test_flags = " -I " + config.omp_header_directory + flags
config.test_flags_use_compiler_omp_h = flags
+
# extra libraries
libs = ""
if config.has_libm:
diff --git a/openmp/runtime/test/lit.site.cfg.in b/openmp/runtime/test/lit.site.cfg.in
index dae5b90..4d8235d 100644
--- a/openmp/runtime/test/lit.site.cfg.in
+++ b/openmp/runtime/test/lit.site.cfg.in
@@ -17,6 +17,7 @@
config.has_libm = @LIBOMP_HAVE_LIBM@
config.has_libatomic = @LIBOMP_HAVE_LIBATOMIC@
config.is_standalone_build = @OPENMP_STANDALONE_BUILD@
+config.has_omit_frame_pointer_flag = @OPENMP_TEST_COMPILER_HAS_OMIT_FRAME_POINTER_FLAGS@
# Let the main config do the real work.
lit_config.load_config(config, "@LIBOMP_BASE_DIR@/test/lit.cfg")
diff --git a/openmp/runtime/test/omp_testsuite.h b/openmp/runtime/test/omp_testsuite.h
index eef5470..c3b5a90 100644
--- a/openmp/runtime/test/omp_testsuite.h
+++ b/openmp/runtime/test/omp_testsuite.h
@@ -20,6 +20,60 @@
#define NUM_TASKS 25
#define MAX_TASKS_PER_THREAD 5
+// Functions that call a parallel region that does very minimal work
+// Some compilers may optimize away an empty parallel region
+volatile int g_counter__;
+
+// If nthreads == 0, then do not use num_threads() clause
+static void go_parallel() {
+ g_counter__ = 0;
+ #pragma omp parallel
+ {
+ #pragma omp atomic
+ g_counter__++;
+ }
+}
+
+static void go_parallel_nthreads(int nthreads) {
+ g_counter__ = 0;
+ #pragma omp parallel num_threads(nthreads)
+ {
+ #pragma omp atomic
+ g_counter__++;
+ }
+}
+
+static void go_parallel_spread() {
+ g_counter__ = 0;
+ #pragma omp parallel proc_bind(spread)
+ {
+ #pragma omp atomic
+ g_counter__++;
+ }
+}
+
+static void go_parallel_close() {
+ g_counter__ = 0;
+ #pragma omp parallel proc_bind(close)
+ {
+ #pragma omp atomic
+ g_counter__++;
+ }
+}
+
+static void go_parallel_master() {
+ g_counter__ = 0;
+ #pragma omp parallel proc_bind(master)
+ {
+ #pragma omp atomic
+ g_counter__++;
+ }
+}
+
+static inline int get_exit_value() {
+ return ((g_counter__ == -1) ? EXIT_FAILURE : EXIT_SUCCESS);
+}
+
#ifdef _WIN32
// Windows versions of pthread_create() and pthread_join()
# include <windows.h>
diff --git a/openmp/runtime/test/ompt/misc/finalize_tool.c b/openmp/runtime/test/ompt/misc/finalize_tool.c
index 0a406c8..54709aa 100644
--- a/openmp/runtime/test/ompt/misc/finalize_tool.c
+++ b/openmp/runtime/test/ompt/misc/finalize_tool.c
@@ -1,16 +1,16 @@
// RUN: %libomp-compile-and-run | FileCheck %s
// REQUIRES: ompt
#include "callback.h"
+#include "omp_testsuite.h"
int main() {
-#pragma omp parallel num_threads(2)
- {}
+ go_parallel_nthreads(2);
printf("Before ompt_finalize_tool\n");
ompt_finalize_tool();
printf("After ompt_finalize_tool\n");
- return 0;
+ return get_exit_value();
}
// CHECK: 0: NULL_POINTER=[[NULL:.*$]]
diff --git a/openmp/runtime/test/ompt/misc/unset_callback.c b/openmp/runtime/test/ompt/misc/unset_callback.c
index 9074ad3..53669fc 100644
--- a/openmp/runtime/test/ompt/misc/unset_callback.c
+++ b/openmp/runtime/test/ompt/misc/unset_callback.c
@@ -1,19 +1,14 @@
// RUN: %libomp-compile-and-run | FileCheck %s
// REQUIRES: ompt
#include "callback.h"
+#include "omp_testsuite.h"
#include <omp.h>
int main()
{
- #pragma omp parallel num_threads(1)
- {
-
- }
+ go_parallel_nthreads(1);
ompt_set_callback(ompt_callback_parallel_begin, NULL);
- #pragma omp parallel num_threads(1)
- {
-
- }
+ go_parallel_nthreads(1);
// Check if libomp supports the callbacks for this test.
// CHECK-NOT: {{^}}0: Could not register callback 'ompt_callback_idle'
@@ -25,5 +20,5 @@
// CHECK-NOT: {{^}}[[THREAD_ID]]: ompt_event_parallel_begin:
// CHECK: {{^}}[[THREAD_ID]]: ompt_event_parallel_end:
- return 0;
+ return get_exit_value();
}
diff --git a/openmp/runtime/test/ompt/tasks/taskwait-depend.c b/openmp/runtime/test/ompt/tasks/taskwait-depend.c
index 752486c..74dad2a 100644
--- a/openmp/runtime/test/ompt/tasks/taskwait-depend.c
+++ b/openmp/runtime/test/ompt/tasks/taskwait-depend.c
@@ -8,8 +8,7 @@
// XFAIL: icc
// support for taskwait with depend clause introduced in clang-14
-// UNSUPPORTED: clang-5, clang-6, clang-6, clang-8, clang-9, clang-10, clang-11,
-// clang-12, clang-13
+// UNSUPPORTED: clang-5, clang-6, clang-6, clang-8, clang-9, clang-10, clang-11, clang-12, clang-13
#include "callback.h"
#include <omp.h>
