1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
|
Fix compilation with musl.
Bug: https://github.com/ROCm/ROCR-Runtime/issues/181
--- a/core/inc/checked.h
+++ b/core/inc/checked.h
@@ -58,7 +58,7 @@ template <uint64_t code, bool multiProcess = false> class Check final {
Check(const Check&) { object_ = uintptr_t(this) ^ uintptr_t(code); }
Check(Check&&) { object_ = uintptr_t(this) ^ uintptr_t(code); }
- ~Check() { object_ = NULL; }
+ ~Check() { object_ = uintptr_t(NULL); }
const Check& operator=(Check&& rhs) { return *this; }
const Check& operator=(const Check& rhs) { return *this; }
--- a/core/runtime/default_signal.cpp
+++ b/core/runtime/default_signal.cpp
@@ -57,7 +57,7 @@ int BusyWaitSignal::rtti_id_ = 0;
BusyWaitSignal::BusyWaitSignal(SharedSignal* abi_block, bool enableIPC)
: Signal(abi_block, enableIPC) {
signal_.kind = AMD_SIGNAL_KIND_USER;
- signal_.event_mailbox_ptr = NULL;
+ signal_.event_mailbox_ptr = uint64_t(NULL);
}
hsa_signal_value_t BusyWaitSignal::LoadRelaxed() {
--- a/core/runtime/hsa.cpp
+++ b/core/runtime/hsa.cpp
@@ -155,7 +155,7 @@ template <class T> struct ValidityError<const T*> {
template <class T>
static __forceinline bool IsValid(T* ptr) {
- return (ptr == NULL) ? NULL : ptr->IsValid();
+ return (ptr == NULL) ? false : ptr->IsValid();
}
namespace AMD {
--- a/core/util/lnx/os_linux.cpp
+++ b/core/util/lnx/os_linux.cpp
@@ -111,9 +111,12 @@ class os_thread {
}
}
+ int cores = 0;
+ cpu_set_t* cpuset = nullptr;
+
if (core::Runtime::runtime_singleton_->flag().override_cpu_affinity()) {
- int cores = get_nprocs_conf();
- cpu_set_t* cpuset = CPU_ALLOC(cores);
+ cores = get_nprocs_conf();
+ cpuset = CPU_ALLOC(cores);
if (cpuset == nullptr) {
fprintf(stderr, "CPU_ALLOC failed: %s\n", strerror(errno));
return;
@@ -122,12 +125,6 @@ class os_thread {
for (int i = 0; i < cores; i++) {
CPU_SET(i, cpuset);
}
- err = pthread_attr_setaffinity_np(&attrib, CPU_ALLOC_SIZE(cores), cpuset);
- CPU_FREE(cpuset);
- if (err != 0) {
- fprintf(stderr, "pthread_attr_setaffinity_np failed: %s\n", strerror(err));
- return;
- }
}
err = pthread_create(&thread, &attrib, ThreadTrampoline, args.get());
@@ -157,6 +154,14 @@ class os_thread {
if (err != 0) {
fprintf(stderr, "pthread_attr_destroy failed: %s\n", strerror(err));
}
+
+ if (thread && cores && cpuset) {
+ err = pthread_setaffinity_np(thread, CPU_ALLOC_SIZE(cores), cpuset);
+ CPU_FREE(cpuset);
+ if (err != 0) {
+ fprintf(stderr, "pthread_setaffinity_np failed: %s\n", strerror(err));
+ }
+ }
}
os_thread(os_thread&& rhs) {
@@ -617,11 +622,13 @@ SharedMutex CreateSharedMutex() {
fprintf(stderr, "rw lock attribute init failed: %s\n", strerror(err));
return nullptr;
}
+#if defined(__GLIBC__)
err = pthread_rwlockattr_setkind_np(&attrib, PTHREAD_RWLOCK_PREFER_WRITER_NONRECURSIVE_NP);
if (err != 0) {
fprintf(stderr, "Set rw lock attribute failure: %s\n", strerror(err));
return nullptr;
}
+#endif
pthread_rwlock_t* lock = new pthread_rwlock_t;
err = pthread_rwlock_init(lock, &attrib);
--- a/core/util/utils.h
+++ b/core/util/utils.h
@@ -74,7 +74,7 @@ static __forceinline void* _aligned_malloc(size_t size, size_t alignment) {
return aligned_alloc(alignment, size);
#else
void *mem = NULL;
- if (NULL != posix_memalign(&mem, alignment, size))
+ if (0 != posix_memalign(&mem, alignment, size))
return NULL;
return mem;
#endif
--- a/image/util.h
+++ b/image/util.h
@@ -95,7 +95,7 @@ static __forceinline void* _aligned_malloc(size_t size, size_t alignment) {
return aligned_alloc(alignment, size);
#else
void* mem = NULL;
- if (NULL != posix_memalign(&mem, alignment, size)) return NULL;
+ if (0 != posix_memalign(&mem, alignment, size)) return NULL;
return mem;
#endif
}
|
GitWeb