Star569
710
代码
代码
Issues52
Pull Requests156
Actions
项目讨论
Wiki
分析
Star569
710
已合并
【PR】: feat:remove all exclude #4195
duanpengliang创建于 18 天前
【PR】: feat:remove all exclude #4195
已合并
duanpengliang创建于 18 天前
从duanpengliang/runtime_newmaster_precommit合入到cann/runtimemaster
共 101 个文件变更+1693-1778
M.pre-commit-config.yaml+2-55
@@ -7,66 +7,13 @@ repos:
7 files: \.(c|h|cpp|hpp|cc|hh|cxx|hxx)$7 files: \.(c|h|cpp|hpp|cc|hh|cxx|hxx)$
8 exclude: |8 exclude: |
9 (?x)^(9 (?x)^(
10- docs10+ include/external/acl/(?:
11- | example
12- | include/external/acl/(?:
13 acl_prof11 acl_prof
14 | acl_rt12 | acl_rt
15 | acl_rt_allocator13 | acl_rt_allocator
16 | acl_tdt14 | acl_tdt
17 | acl_tdt_queue15 | acl_tdt_queue
18- )\.h16+ )\.h$
19- | include(?:
20- /(?!
21- (?:
22- external
23- | driver
24- | tsd_client
25- | dfx/base/err_msg.h
26- )(?:/|$)
27- )
28- | $
29- )
30- | pkg_inc(?:
31- /(?!
32- (?:
33- aicpu_sched
34- | base/err_mgr.h
35- | driver
36- | mmpa
37- | queue_schedule
38- | runtime
39- | tsd
40- )(?:/|$)
41- )
42- | $
43- )
44- | scripts
45- | src/(
46- dfx/adump(?!(?:/adcore)(?:/|$))
47- | dfx/log
48- | dfx/trace/(?:atrace|common|cpu_detect|inc)
49- )(/|$)
50- | stub
51- | tests(?:
52- /(?!
53- (?:
54- cmodel_test
55- | depends/mmpa
56- | depends/runtime
57- | depends/tdt
58- | ut/acl
59- | ut/aicpu_sched
60- | ut/error_manager
61- | ut/mmpa
62- | ut/queue_schedule
63- | ut/runtime
64- | ut/tsd
65- )(?:/|$)
66- | depends/acl_stub\.h$
67- )
68- | $
69- )
70 )17 )
71 args:18 args:
72 - "--style=file"19 - "--style=file"
Mexample/0_quickstart/1_error_handling/main.cpp+19-29
@@ -15,15 +15,9 @@
15#include "utils.h"15#include "utils.h"
16 16 
17namespace {17namespace {
18-const char *SafeString(const char *message)18+const char* SafeString(const char* message) { return message != nullptr ? message : "<null>"; }
19-{
20- return message != nullptr ? message : "<null>";
21-}
22 19 
23-const char *RunModeToString(aclrtRunMode runMode)20+const char* RunModeToString(aclrtRunMode runMode) { return runMode == ACL_HOST ? "ACL_HOST" : "ACL_DEVICE"; }
24-{
25- return runMode == ACL_HOST ? "ACL_HOST" : "ACL_DEVICE";
26-}
27 21 
28void PrintVerboseErrorInfo(int32_t deviceId)22void PrintVerboseErrorInfo(int32_t deviceId)
29{23{
@@ -34,20 +28,19 @@ void PrintVerboseErrorInfo(int32_t deviceId)
34 return;28 return;
35 }29 }
36 30 
37- INFO_LOG("Verbose error info: errorType=%d, tryRepair=%u, hasDetail=%u",31+ INFO_LOG(
38- static_cast<int32_t>(errorInfo.errorType),32+ "Verbose error info: errorType=%d, tryRepair=%u, hasDetail=%u", static_cast<int32_t>(errorInfo.errorType),
39- static_cast<uint32_t>(errorInfo.tryRepair),33+ static_cast<uint32_t>(errorInfo.tryRepair), static_cast<uint32_t>(errorInfo.hasDetail));
40- static_cast<uint32_t>(errorInfo.hasDetail));
41}34}
42} // namespace35} // namespace
43 36 
44-#define CHECK_ACL_RETURN(call) \37+#define CHECK_ACL_RETURN(call) \
45- do { \38+ do { \
46- aclError ret = (call); \39+ aclError ret = (call); \
47- if (ret != ACL_SUCCESS) { \40+ if (ret != ACL_SUCCESS) { \
48 ERROR_LOG("%s failed: ret=%d", #call, static_cast<int32_t>(ret)); \41 ERROR_LOG("%s failed: ret=%d", #call, static_cast<int32_t>(ret)); \
49- return -1; \42+ return -1; \
50- } \43+ } \
51 } while (0)44 } while (0)
52 45 
53int main()46int main()
@@ -72,22 +65,19 @@ int main()
72 } else {65 } else {
73 aclError peekError = aclrtPeekAtLastError(ACL_RT_THREAD_LEVEL);66 aclError peekError = aclrtPeekAtLastError(ACL_RT_THREAD_LEVEL);
74 aclError lastError = aclrtGetLastError(ACL_RT_THREAD_LEVEL);67 aclError lastError = aclrtGetLastError(ACL_RT_THREAD_LEVEL);
75- const char *recentErrMsg = aclGetRecentErrMsg();68+ const char* recentErrMsg = aclGetRecentErrMsg();
76- ERROR_LOG("Diagnostics: ret=%d, peekErr=%d, lastErr=%d, recentErrMsg=%s",69+ ERROR_LOG(
77- static_cast<int32_t>(expectedRet),70+ "Diagnostics: ret=%d, peekErr=%d, lastErr=%d, recentErrMsg=%s", static_cast<int32_t>(expectedRet),
78- static_cast<int32_t>(peekError),71+ static_cast<int32_t>(peekError), static_cast<int32_t>(lastError), SafeString(recentErrMsg));
79- static_cast<int32_t>(lastError),
80- SafeString(recentErrMsg));
81 PrintVerboseErrorInfo(deviceId);72 PrintVerboseErrorInfo(deviceId);
82 73 
83 // Read the diagnostics again to show the consumed state.74 // Read the diagnostics again to show the consumed state.
84 aclError clearedPeek = aclrtPeekAtLastError(ACL_RT_THREAD_LEVEL);75 aclError clearedPeek = aclrtPeekAtLastError(ACL_RT_THREAD_LEVEL);
85 aclError clearedLast = aclrtGetLastError(ACL_RT_THREAD_LEVEL);76 aclError clearedLast = aclrtGetLastError(ACL_RT_THREAD_LEVEL);
86- const char *clearedErrMsg = aclGetRecentErrMsg();77+ const char* clearedErrMsg = aclGetRecentErrMsg();
87- INFO_LOG("After diagnostics are consumed once: peekErr=%d, lastErr=%d, recentErrMsg=%s",78+ INFO_LOG(
88- static_cast<int32_t>(clearedPeek),79+ "After diagnostics are consumed once: peekErr=%d, lastErr=%d, recentErrMsg=%s",
89- static_cast<int32_t>(clearedLast),80+ static_cast<int32_t>(clearedPeek), static_cast<int32_t>(clearedLast), SafeString(clearedErrMsg));
90- SafeString(clearedErrMsg));
91 }81 }
92 82 
93 // Release ACL resources.83 // Release ACL resources.
Mexample/0_quickstart/2_system_info/main.cpp+6-9
@@ -14,10 +14,7 @@
14#include "utils.h"14#include "utils.h"
15 15 
16namespace {16namespace {
17-const char *RunModeToString(aclrtRunMode runMode)17+const char* RunModeToString(aclrtRunMode runMode) { return runMode == ACL_HOST ? "ACL_HOST" : "ACL_DEVICE"; }
18-{
19- return runMode == ACL_HOST ? "ACL_HOST" : "ACL_DEVICE";
20-}
21 18 
22aclError PrintCannVersionInfo()19aclError PrintCannVersionInfo()
23{20{
@@ -58,12 +55,12 @@ int main()
58 float originalValue = 1.625f;55 float originalValue = 1.625f;
59 aclFloat16 fp16Value = aclFloatToFloat16(originalValue);56 aclFloat16 fp16Value = aclFloatToFloat16(originalValue);
60 float restoredValue = aclFloat16ToFloat(fp16Value);57 float restoredValue = aclFloat16ToFloat(fp16Value);
61- INFO_LOG("Float conversion: %.6f -> 0x%04x -> %.6f", originalValue, static_cast<unsigned int>(fp16Value),58+ INFO_LOG(
62- restoredValue);59+ "Float conversion: %.6f -> 0x%04x -> %.6f", originalValue, static_cast<unsigned int>(fp16Value), restoredValue);
63 60 
64- INFO_LOG("Data type size: ACL_FLOAT=%zu, ACL_FLOAT16=%zu, ACL_INT64=%zu",61+ INFO_LOG(
65- aclDataTypeSize(aclDataType::ACL_FLOAT), aclDataTypeSize(aclDataType::ACL_FLOAT16),62+ "Data type size: ACL_FLOAT=%zu, ACL_FLOAT16=%zu, ACL_INT64=%zu", aclDataTypeSize(aclDataType::ACL_FLOAT),
66- aclDataTypeSize(aclDataType::ACL_INT64));63+ aclDataTypeSize(aclDataType::ACL_FLOAT16), aclDataTypeSize(aclDataType::ACL_INT64));
67 64 
68 // Finalize ACL after all queries finish.65 // Finalize ACL after all queries finish.
69 CHECK_ERROR(aclFinalize());66 CHECK_ERROR(aclFinalize());
Mexample/0_quickstart/3_cross_version/0_runtime_compatibility/main.cpp+6-5
@@ -16,7 +16,7 @@
16#include "utils.h"16#include "utils.h"
17 17 
18namespace {18namespace {
19-const char *CannAttrToString(aclCannAttr attr)19+const char* CannAttrToString(aclCannAttr attr)
20{20{
21 switch (attr) {21 switch (attr) {
22 case ACL_CANN_ATTR_INF_NAN:22 case ACL_CANN_ATTR_INF_NAN:
@@ -45,7 +45,7 @@ int QueryCannVersion()
45 45 
46int QueryCannAttributes()46int QueryCannAttributes()
47{47{
48- const aclCannAttr *attrList = nullptr;48+ const aclCannAttr* attrList = nullptr;
49 size_t attrCount = 0;49 size_t attrCount = 0;
50 CHECK_ERROR(aclGetCannAttributeList(&attrList, &attrCount));50 CHECK_ERROR(aclGetCannAttributeList(&attrList, &attrCount));
51 INFO_LOG("CANN attribute count: %zu", attrCount);51 INFO_LOG("CANN attribute count: %zu", attrCount);
@@ -61,7 +61,7 @@ int QueryCannAttributes()
61 61 
62int QueryCompatibilityAndCapability(int32_t deviceId)62int QueryCompatibilityAndCapability(int32_t deviceId)
63{63{
64- const char *socName = aclrtGetSocName();64+ const char* socName = aclrtGetSocName();
65 if (socName == nullptr || std::strlen(socName) == 0U) {65 if (socName == nullptr || std::strlen(socName) == 0U) {
66 WARN_LOG("aclrtGetSocName returned empty soc name. Skip architecture compatibility check.");66 WARN_LOG("aclrtGetSocName returned empty soc name. Skip architecture compatibility check.");
67 } else {67 } else {
@@ -75,7 +75,7 @@ int QueryCompatibilityAndCapability(int32_t deviceId)
75 INFO_LOG("Device capability ACL_FEATURE_TSCPU_TASK_UPDATE_SUPPORT_AIC_AIV: %d", capability);75 INFO_LOG("Device capability ACL_FEATURE_TSCPU_TASK_UPDATE_SUPPORT_AIC_AIV: %d", capability);
76 return 0;76 return 0;
77}77}
78-} // namespace78+} // namespace
79 79 
80int32_t main()80int32_t main()
81{81{
@@ -94,7 +94,8 @@ int32_t main()
94 94 
95 aclError cleanupRet = aclrtResetDeviceForce(deviceId);95 aclError cleanupRet = aclrtResetDeviceForce(deviceId);
96 if (cleanupRet != ACL_SUCCESS) {96 if (cleanupRet != ACL_SUCCESS) {
97- ERROR_LOG("Operation failed: aclrtResetDeviceForce(deviceId) returned error code %d",97+ ERROR_LOG(
98+ "Operation failed: aclrtResetDeviceForce(deviceId) returned error code %d",
98 static_cast<int32_t>(cleanupRet));99 static_cast<int32_t>(cleanupRet));
99 ret = ret == 0 ? -1 : ret;100 ret = ret == 0 ? -1 : ret;
100 }101 }
Mexample/0_quickstart/4_custom_kernel_launch/main.cpp+3-3
@@ -16,10 +16,10 @@
16#include "acl/acl.h"16#include "acl/acl.h"
17#include "vector_add_kernel.h"17#include "vector_add_kernel.h"
18 18 
19-#define CHECK_ERROR(ret) \19+#define CHECK_ERROR(ret) \
20- if ((ret) != ACL_SUCCESS) { \20+ if ((ret) != ACL_SUCCESS) { \
21 printf("Error at line %d, ret = %d\n", __LINE__, static_cast<int32_t>(ret)); \21 printf("Error at line %d, ret = %d\n", __LINE__, static_cast<int32_t>(ret)); \
22- return -1; \22+ return -1; \
23 }23 }
24 24 
25namespace {25namespace {
Mexample/0_quickstart/4_custom_kernel_launch/vector_add_kernel.cpp+1-3
@@ -23,9 +23,7 @@ extern "C" __global__ __aicore__ void VectorAddKernel(
23 // A5 编译时，生成器会在 kernel tail 中追加 dci()。关闭自动 DCCI 后，scalar store 产生的脏数据23 // A5 编译时，生成器会在 kernel tail 中追加 dci()。关闭自动 DCCI 后，scalar store 产生的脏数据
24 // 尚未回写到 GM 就被 dci() 失效，导致输出全为 0，因此需要显式调用 dcci() 将数据回写到 GM。24 // 尚未回写到 GM 就被 dci() 失效，导致输出全为 0，因此需要显式调用 dcci() 将数据回写到 GM。
25#if __NPU_ARCH__ == 351025#if __NPU_ARCH__ == 3510
26- dcci(reinterpret_cast<__gm__ int64_t*>(dst),26+ dcci(reinterpret_cast<__gm__ int64_t*>(dst), cache_line_t::ENTIRE_DATA_CACHE, dcci_dst_t::CACHELINE_OUT);
27- cache_line_t::ENTIRE_DATA_CACHE,
28- dcci_dst_t::CACHELINE_OUT);
29#endif27#endif
30}28}
31 29 
Mexample/1_basic_features/context/0_context_query/main.cpp+3-3
@@ -27,8 +27,8 @@ public:
27 }27 }
28 }28 }
29 29 
30- ContextGuard(const ContextGuard &) = delete;30+ ContextGuard(const ContextGuard&) = delete;
31- ContextGuard &operator=(const ContextGuard &) = delete;31+ ContextGuard& operator=(const ContextGuard&) = delete;
32 32 
33 aclError Destroy()33 aclError Destroy()
34 {34 {
@@ -79,7 +79,7 @@ int32_t RunContextQuerySample()
79 INFO_LOG("[SUCCESS] Context query sample completed successfully");79 INFO_LOG("[SUCCESS] Context query sample completed successfully");
80 return 0;80 return 0;
81}81}
82-} // namespace82+} // namespace
83 83 
84int32_t main()84int32_t main()
85{85{
Mexample/1_basic_features/device/0_device_normal/main.cpp+66-63
@@ -15,70 +15,73 @@
15#include "utils.h"15#include "utils.h"
16 16 
17namespace {17namespace {
18- int Init(int32_t deviceId, aclrtStream* stream) 18+int Init(int32_t deviceId, aclrtStream* stream)
19- {19+{
20- CHECK_ERROR(aclInit(nullptr));20+ CHECK_ERROR(aclInit(nullptr));
21- CHECK_ERROR(aclrtSetDevice(deviceId));21+ CHECK_ERROR(aclrtSetDevice(deviceId));
22- CHECK_ERROR(aclrtCreateStream(stream));22+ CHECK_ERROR(aclrtCreateStream(stream));
23- return 0;23+ return 0;
24+}
25+ 
26+int64_t GetShapeSize(const std::vector<int64_t>& shape)
27+{
28+ int64_t shape_size = 1;
29+ for (auto i : shape) {
30+ shape_size *= i;
31+ }
32+ return shape_size;
33+}
34+ 
35+template <typename T>
36+int CreateAclTensor(
37+ const std::vector<T>& hostData, const std::vector<int64_t>& shape, void** deviceAddr, aclDataType dataType,
38+ aclTensor** tensor)
39+{
40+ auto size = GetShapeSize(shape) * sizeof(T);
41+ 
42+ // Allocate memory on the device
43+ CHECK_ERROR(aclrtMalloc(deviceAddr, size, ACL_MEM_MALLOC_HUGE_FIRST));
44+ 
45+ // Copy memory from host to device synchronously
46+ CHECK_ERROR(aclrtMemcpy(*deviceAddr, size, hostData.data(), size, ACL_MEMCPY_HOST_TO_DEVICE));
47+ 
48+ // Calculate the strides
49+ std::vector<int64_t> strides(shape.size(), 1);
50+ for (int64_t i = shape.size() - 2; i >= 0; i--) {
51+ strides[i] = shape[i + 1] * strides[i + 1];
24 }52 }
25 53 
26- int64_t GetShapeSize(const std::vector<int64_t> &shape)54+ // Create the tensor
27- {55+ *tensor = aclCreateTensor(
28- int64_t shape_size = 1;56+ shape.data(), shape.size(), dataType, strides.data(), 0, aclFormat::ACL_FORMAT_ND, shape.data(), shape.size(),
29- for (auto i : shape) {57+ *deviceAddr);
30- shape_size *= i;58+ return 0;
31- }59+}
32- return shape_size;
33- }
34 60 
35- template <typename T>61+void DestroyTensorResources(aclTensor* self, aclTensor* other, aclScalar* alpha, aclTensor* out)
36- int CreateAclTensor(const std::vector<T> &hostData, const std::vector<int64_t> &shape, void **deviceAddr, 62+{
37- aclDataType dataType, aclTensor **tensor)63+ aclDestroyTensor(self);
38- {64+ aclDestroyTensor(other);
39- auto size = GetShapeSize(shape) * sizeof(T);65+ aclDestroyScalar(alpha);
66+ aclDestroyTensor(out);
67+}
40 68 
41- // Allocate memory on the device69+int GetAndPrintResult(void* outDeviceAddr, const std::vector<int64_t>& outShape)
42- CHECK_ERROR(aclrtMalloc(deviceAddr, size, ACL_MEM_MALLOC_HUGE_FIRST));70+{
71+ auto size = GetShapeSize(outShape);
72+ std::vector<float> resultData(size, 0);
73+ CHECK_ERROR(aclrtMemcpy(
74+ resultData.data(), resultData.size() * sizeof(resultData[0]), outDeviceAddr, size * sizeof(float),
75+ ACL_MEMCPY_DEVICE_TO_HOST));
43 76 
44- // Copy memory from host to device synchronously77+ for (int64_t i = 0; i < size; i++) {
45- CHECK_ERROR(aclrtMemcpy(*deviceAddr, size, hostData.data(), size, ACL_MEMCPY_HOST_TO_DEVICE));
46- 
47- // Calculate the strides
48- std::vector<int64_t> strides(shape.size(), 1);
49- for (int64_t i = shape.size() - 2; i >= 0; i--) {
50- strides[i] = shape[i + 1] * strides[i + 1];
51- }
52- 
53- // Create the tensor
54- *tensor = aclCreateTensor(shape.data(), shape.size(), dataType, strides.data(), 0, aclFormat::ACL_FORMAT_ND,
55- shape.data(), shape.size(), *deviceAddr);
56- return 0;
57- }
58- 
59- void DestroyTensorResources(aclTensor *self, aclTensor *other, aclScalar *alpha, aclTensor *out)
60- {
61- aclDestroyTensor(self);
62- aclDestroyTensor(other);
63- aclDestroyScalar(alpha);
64- aclDestroyTensor(out);
65- }
66- 
67- int GetAndPrintResult(void *outDeviceAddr, const std::vector<int64_t> &outShape)
68- {
69- auto size = GetShapeSize(outShape);
70- std::vector<float> resultData(size, 0);
71- CHECK_ERROR(aclrtMemcpy(resultData.data(), resultData.size() * sizeof(resultData[0]), outDeviceAddr,
72- size * sizeof(float), ACL_MEMCPY_DEVICE_TO_HOST));
73- 
74- for (int64_t i = 0; i < size; i++) {
75 INFO_LOG("result[%ld] is: %f", i, resultData[i]);78 INFO_LOG("result[%ld] is: %f", i, resultData[i]);
76- }
77- return 0;
78 }79 }
80+ return 0;
81+}
79} // namespace82} // namespace
80 83 
81-int32_t main(int argc, char const *argv[])84+int32_t main(int argc, char const* argv[])
82{85{
83 INFO_LOG("Start to run device_normal sample.");86 INFO_LOG("Start to run device_normal sample.");
84 int32_t deviceId = 0;87 int32_t deviceId = 0;
@@ -88,7 +91,7 @@ int32_t main(int argc, char const *argv[])
88 std::vector<int64_t> selfShape{4, 2}, otherShape{4, 2}, outShape{4, 2};91 std::vector<int64_t> selfShape{4, 2}, otherShape{4, 2}, outShape{4, 2};
89 void *selfDeviceAddr = nullptr, *otherDeviceAddr = nullptr, *outDeviceAddr = nullptr;92 void *selfDeviceAddr = nullptr, *otherDeviceAddr = nullptr, *outDeviceAddr = nullptr;
90 aclTensor *self = nullptr, *other = nullptr, *out = nullptr;93 aclTensor *self = nullptr, *other = nullptr, *out = nullptr;
91- aclScalar *alpha = nullptr;94+ aclScalar* alpha = nullptr;
92 std::vector<float> selfHostData = {0, 1, 2, 3, 4, 5, 6, 7};95 std::vector<float> selfHostData = {0, 1, 2, 3, 4, 5, 6, 7};
93 std::vector<float> otherHostData = {1, 1, 1, 2, 2, 2, 3, 3};96 std::vector<float> otherHostData = {1, 1, 1, 2, 2, 2, 3, 3};
94 std::vector<float> outHostData = {0, 0, 0, 0, 0, 0, 0, 0};97 std::vector<float> outHostData = {0, 0, 0, 0, 0, 0, 0, 0};
@@ -99,21 +102,21 @@ int32_t main(int argc, char const *argv[])
99 102 
100 alpha = aclCreateScalar(&alphaValue, aclDataType::ACL_FLOAT);103 alpha = aclCreateScalar(&alphaValue, aclDataType::ACL_FLOAT);
101 if (alpha == nullptr) {104 if (alpha == nullptr) {
102- ERROR_LOG("Create alpha Scalar failed.");105+ ERROR_LOG("Create alpha Scalar failed.");
103- return -1;106+ return -1;
104 }107 }
105 108 
106 CHECK_ERROR(CreateAclTensor(outHostData, outShape, &outDeviceAddr, aclDataType::ACL_FLOAT, &out));109 CHECK_ERROR(CreateAclTensor(outHostData, outShape, &outDeviceAddr, aclDataType::ACL_FLOAT, &out));
107 110 
108 // Call the CANN operator library API111 // Call the CANN operator library API
109 uint64_t workspaceSize = 0;112 uint64_t workspaceSize = 0;
110- aclOpExecutor *executor;113+ aclOpExecutor* executor;
111 CHECK_ERROR(aclnnAddGetWorkspaceSize(self, other, alpha, out, &workspaceSize, &executor));114 CHECK_ERROR(aclnnAddGetWorkspaceSize(self, other, alpha, out, &workspaceSize, &executor));
112 115 
113 // Allocate device memory based on the calculation results116 // Allocate device memory based on the calculation results
114- void *workspaceAddr = nullptr;117+ void* workspaceAddr = nullptr;
115 if (workspaceSize > 0lu) {118 if (workspaceSize > 0lu) {
116- CHECK_ERROR(aclrtMalloc(&workspaceAddr, workspaceSize, ACL_MEM_MALLOC_HUGE_FIRST));119+ CHECK_ERROR(aclrtMalloc(&workspaceAddr, workspaceSize, ACL_MEM_MALLOC_HUGE_FIRST));
117 }120 }
118 // Call the Add operator121 // Call the Add operator
119 CHECK_ERROR(aclnnAdd(workspaceAddr, workspaceSize, executor, stream));122 CHECK_ERROR(aclnnAdd(workspaceAddr, workspaceSize, executor, stream));
@@ -133,7 +136,7 @@ int32_t main(int argc, char const *argv[])
133 CHECK_ERROR(aclrtFree(otherDeviceAddr));136 CHECK_ERROR(aclrtFree(otherDeviceAddr));
134 CHECK_ERROR(aclrtFree(outDeviceAddr));137 CHECK_ERROR(aclrtFree(outDeviceAddr));
135 if (workspaceSize > 0lu) {138 if (workspaceSize > 0lu) {
136- CHECK_ERROR(aclrtFree(workspaceAddr));139+ CHECK_ERROR(aclrtFree(workspaceAddr));
137 }140 }
138 CHECK_ERROR(aclrtDestroyStream(stream));141 CHECK_ERROR(aclrtDestroyStream(stream));
139 CHECK_ERROR(aclrtResetDeviceForce(deviceId));142 CHECK_ERROR(aclrtResetDeviceForce(deviceId));
@@ -141,4 +144,4 @@ int32_t main(int argc, char const *argv[])
141 INFO_LOG("Run the device_normal sample successfully.");144 INFO_LOG("Run the device_normal sample successfully.");
142 145 
143 return 0;146 return 0;
144-}147+}
Mexample/1_basic_features/device/1_device_multi_thread/main.cpp+107-106
@@ -17,123 +17,124 @@
17using namespace std;17using namespace std;
18 18 
19namespace {19namespace {
20- int Init(aclrtStream* stream)20+int Init(aclrtStream* stream)
21- {21+{
22- CHECK_ERROR(aclrtSetDevice(0));22+ CHECK_ERROR(aclrtSetDevice(0));
23- CHECK_ERROR(aclrtCreateStream(stream));23+ CHECK_ERROR(aclrtCreateStream(stream));
24- return 0;24+ return 0;
25+}
26+ 
27+int DeviceInfoQuery()
28+{
29+ aclrtRunMode runMode;
30+ int64_t vectorCoreNum = 0;
31+ uint32_t deviceCount = 0;
32+ aclrtDeviceStatus deviceStatus;
33+ 
34+ string socName = aclrtGetSocName();
35+ INFO_LOG("Current Ascend chipset platform is: %s.", socName.c_str());
36+ 
37+ CHECK_ERROR(aclrtGetDeviceCount(&deviceCount));
38+ INFO_LOG("Get device count success. deviceCount: %d.", deviceCount);
39+ 
40+ CHECK_ERROR(aclrtQueryDeviceStatus(0, &deviceStatus));
41+ INFO_LOG("Query device status success. deviceStatus: %d.", deviceStatus);
42+ 
43+ CHECK_ERROR(aclrtGetRunMode(&runMode));
44+ if (runMode == 0) {
45+ INFO_LOG("RunMode is ACL_DEVICE.");
46+ } else if (runMode == 1) {
47+ INFO_LOG("RunMode is ACL_HOST.");
48+ } else {
49+ INFO_LOG("RunMode is invalid.");
25 }50 }
26 51 
27- int DeviceInfoQuery()52+ aclrtUtilizationInfo utilizationInfo;
28- {53+ utilizationInfo.utilizationExtend = nullptr;
29- aclrtRunMode runMode;54+ CHECK_ERROR(aclrtGetDeviceUtilizationRate(0, &utilizationInfo));
30- int64_t vectorCoreNum = 0;55+ INFO_LOG(
31- uint32_t deviceCount = 0;56+ "Get device utilizationRate success. cubeUtilization %d, vectorUtilization %d, aicpuUtilization %d,"
32- aclrtDeviceStatus deviceStatus;57+ "memoryUtilization %d.",
58+ utilizationInfo.cubeUtilization, utilizationInfo.vectorUtilization, utilizationInfo.aicpuUtilization,
59+ utilizationInfo.memoryUtilization);
33 60 
34- string socName = aclrtGetSocName();61+ int32_t leastPriority;
35- INFO_LOG("Current Ascend chipset platform is: %s.", socName.c_str());62+ int32_t greatestPriority;
63+ CHECK_ERROR(aclrtDeviceGetStreamPriorityRange(&leastPriority, &greatestPriority));
64+ INFO_LOG(
65+ "Get stream priorityRange success. leastPriority %d, greatestPriority %d.", leastPriority, greatestPriority);
36 66 
37- CHECK_ERROR(aclrtGetDeviceCount(&deviceCount));67+ CHECK_ERROR(aclrtGetDeviceInfo(0, ACL_DEV_ATTR_VECTOR_CORE_NUM, &vectorCoreNum));
38- INFO_LOG("Get device count success. deviceCount: %d.", deviceCount);68+ INFO_LOG("Get device info success. vectorCoreNum %ld.", vectorCoreNum);
69+ return 0;
70+}
39 71 
40- CHECK_ERROR(aclrtQueryDeviceStatus(0, &deviceStatus));72+void RunThread()
41- INFO_LOG("Query device status success. deviceStatus: %d.", deviceStatus);73+{
74+ aclrtStream stream = nullptr;
75+ uint32_t resLimitValue = 0;
42 76 
43- CHECK_ERROR(aclrtGetRunMode(&runMode));77+ auto ret = Init(&stream);
44- if (runMode == 0) {78+ if (ret != 0) {
45- INFO_LOG("RunMode is ACL_DEVICE.");79+ return;
46- } else if (runMode == 1) {
47- INFO_LOG("RunMode is ACL_HOST.");
48- } else {
49- INFO_LOG("RunMode is invalid.");
50- }
51- 
52- aclrtUtilizationInfo utilizationInfo;
53- utilizationInfo.utilizationExtend = nullptr;
54- CHECK_ERROR(aclrtGetDeviceUtilizationRate(0, &utilizationInfo));
55- INFO_LOG("Get device utilizationRate success. cubeUtilization %d, vectorUtilization %d, aicpuUtilization %d,"
56- "memoryUtilization %d.",
57- utilizationInfo.cubeUtilization, utilizationInfo.vectorUtilization, utilizationInfo.aicpuUtilization,
58- utilizationInfo.memoryUtilization);
59- 
60- int32_t leastPriority;
61- int32_t greatestPriority;
62- CHECK_ERROR(aclrtDeviceGetStreamPriorityRange(&leastPriority, &greatestPriority));
63- INFO_LOG("Get stream priorityRange success. leastPriority %d, greatestPriority %d.", leastPriority,
64- greatestPriority);
65-
66- CHECK_ERROR(aclrtGetDeviceInfo(0, ACL_DEV_ATTR_VECTOR_CORE_NUM, &vectorCoreNum));
67- INFO_LOG("Get device info success. vectorCoreNum %ld.", vectorCoreNum);
68- return 0;
69 }80 }
70 81 
71- void RunThread()82+ ret = DeviceInfoQuery();
72- {83+ if (ret != 0) {
73- aclrtStream stream = nullptr;84+ return;
74- uint32_t resLimitValue = 0;
75- 
76- auto ret = Init(&stream);
77- if (ret != 0) {
78- return;
79- }
80- 
81- ret = DeviceInfoQuery();
82- if (ret != 0) {
83- return;
84- }
85- 
86- // Define the parameters of the kernel function
87- const uint32_t TOTAL_SIZE = 1024;
88- const uint32_t DATA_SIZE = TOTAL_SIZE * static_cast<uint32_t>(sizeof(float));
89- 
90- // Allocate memory on the host
91- float *hostSrcA, *hostSrcB, *hostDst;
92- aclrtMallocHost(reinterpret_cast<void **>(&hostSrcA), DATA_SIZE);
93- aclrtMallocHost(reinterpret_cast<void **>(&hostSrcB), DATA_SIZE);
94- aclrtMallocHost(reinterpret_cast<void **>(&hostDst), DATA_SIZE);
95- 
96- // Allocate memory on the device
97- float *devSrcA, *devSrcB, *devDst;
98- aclrtMalloc(reinterpret_cast<void **>(&devSrcA), DATA_SIZE, ACL_MEM_MALLOC_HUGE_FIRST);
99- aclrtMalloc(reinterpret_cast<void **>(&devSrcB), DATA_SIZE, ACL_MEM_MALLOC_HUGE_FIRST);
100- aclrtMalloc(reinterpret_cast<void **>(&devDst), DATA_SIZE, ACL_MEM_MALLOC_HUGE_FIRST);
101- 
102- // Initialize the test data
103- for (uint32_t i = 0; i < TOTAL_SIZE; ++i) {
104- hostSrcA[i] = static_cast<float>(i);
105- hostSrcB[i] = static_cast<float>(i);
106- }
107- 
108- aclrtMemcpy(devSrcA, DATA_SIZE, hostSrcA, DATA_SIZE, ACL_MEMCPY_HOST_TO_DEVICE);
109- aclrtMemcpy(devSrcB, DATA_SIZE, hostSrcB, DATA_SIZE, ACL_MEMCPY_HOST_TO_DEVICE);
110- 
111- aclrtGetDeviceResLimit(0, ACL_RT_DEV_RES_VECTOR_CORE, &resLimitValue);
112- INFO_LOG("Get device resLimit success. VECTOR_CORE %d.", resLimitValue);
113- AddDo(resLimitValue, stream, devSrcA, devSrcB, devDst, TOTAL_SIZE);
114- aclrtSynchronizeStream(stream);
115- aclrtMemcpy(hostDst, DATA_SIZE, devDst, DATA_SIZE, ACL_MEMCPY_DEVICE_TO_HOST);
116- 
117- INFO_LOG("The results (first 10 elements) of the kernel function:");
118- for (int i = 0; i < 10; ++i) { // Display the first 10 elements
119- INFO_LOG("Result: hostDst[%d]: %f Expected value: %f", i, hostDst[i], hostSrcA[i] + hostSrcB[i]);
120- }
121- 
122- aclrtResetDeviceResLimit(0);
123- 
124- aclrtFree(devSrcA);
125- aclrtFree(devSrcB);
126- aclrtFree(devDst);
127- aclrtFreeHost(hostSrcA);
128- aclrtFreeHost(hostSrcB);
129- aclrtFreeHost(hostDst);
130- 
131- aclrtDestroyStream(stream);
132- aclrtResetDeviceForce(0);
133 }85 }
86+ 
87+ // Define the parameters of the kernel function
88+ const uint32_t TOTAL_SIZE = 1024;
89+ const uint32_t DATA_SIZE = TOTAL_SIZE * static_cast<uint32_t>(sizeof(float));
90+ 
91+ // Allocate memory on the host
92+ float *hostSrcA, *hostSrcB, *hostDst;
93+ aclrtMallocHost(reinterpret_cast<void**>(&hostSrcA), DATA_SIZE);
94+ aclrtMallocHost(reinterpret_cast<void**>(&hostSrcB), DATA_SIZE);
95+ aclrtMallocHost(reinterpret_cast<void**>(&hostDst), DATA_SIZE);
96+ 
97+ // Allocate memory on the device
98+ float *devSrcA, *devSrcB, *devDst;
99+ aclrtMalloc(reinterpret_cast<void**>(&devSrcA), DATA_SIZE, ACL_MEM_MALLOC_HUGE_FIRST);
100+ aclrtMalloc(reinterpret_cast<void**>(&devSrcB), DATA_SIZE, ACL_MEM_MALLOC_HUGE_FIRST);
101+ aclrtMalloc(reinterpret_cast<void**>(&devDst), DATA_SIZE, ACL_MEM_MALLOC_HUGE_FIRST);
102+ 
103+ // Initialize the test data
104+ for (uint32_t i = 0; i < TOTAL_SIZE; ++i) {
105+ hostSrcA[i] = static_cast<float>(i);
106+ hostSrcB[i] = static_cast<float>(i);
107+ }
108+ 
109+ aclrtMemcpy(devSrcA, DATA_SIZE, hostSrcA, DATA_SIZE, ACL_MEMCPY_HOST_TO_DEVICE);
110+ aclrtMemcpy(devSrcB, DATA_SIZE, hostSrcB, DATA_SIZE, ACL_MEMCPY_HOST_TO_DEVICE);
111+ 
112+ aclrtGetDeviceResLimit(0, ACL_RT_DEV_RES_VECTOR_CORE, &resLimitValue);
113+ INFO_LOG("Get device resLimit success. VECTOR_CORE %d.", resLimitValue);
114+ AddDo(resLimitValue, stream, devSrcA, devSrcB, devDst, TOTAL_SIZE);
115+ aclrtSynchronizeStream(stream);
116+ aclrtMemcpy(hostDst, DATA_SIZE, devDst, DATA_SIZE, ACL_MEMCPY_DEVICE_TO_HOST);
117+ 
118+ INFO_LOG("The results (first 10 elements) of the kernel function:");
119+ for (int i = 0; i < 10; ++i) { // Display the first 10 elements
120+ INFO_LOG("Result: hostDst[%d]: %f Expected value: %f", i, hostDst[i], hostSrcA[i] + hostSrcB[i]);
121+ }
122+ 
123+ aclrtResetDeviceResLimit(0);
124+ 
125+ aclrtFree(devSrcA);
126+ aclrtFree(devSrcB);
127+ aclrtFree(devDst);
128+ aclrtFreeHost(hostSrcA);
129+ aclrtFreeHost(hostSrcB);
130+ aclrtFreeHost(hostDst);
131+ 
132+ aclrtDestroyStream(stream);
133+ aclrtResetDeviceForce(0);
134+}
134} // namespace135} // namespace
135 136 
136-int32_t main(int argc, char const *argv[])137+int32_t main(int argc, char const* argv[])
137{138{
138 INFO_LOG("Start to run device_multi_thread sample.");139 INFO_LOG("Start to run device_multi_thread sample.");
139 CHECK_ERROR(aclInit(nullptr));140 CHECK_ERROR(aclInit(nullptr));
Mexample/1_basic_features/device/2_device_P2P/main.cpp+2-2
@@ -16,7 +16,7 @@
16 16 
17constexpr uint32_t UINT8_DATA_LEN = 256;17constexpr uint32_t UINT8_DATA_LEN = 256;
18 18 
19-int32_t main(int argc, const char *argv[])19+int32_t main(int argc, const char* argv[])
20{20{
21 INFO_LOG("Start to run device_P2P sample.");21 INFO_LOG("Start to run device_P2P sample.");
22 CHECK_ERROR(aclInit(NULL));22 CHECK_ERROR(aclInit(NULL));
@@ -62,7 +62,7 @@ int32_t main(int argc, const char *argv[])
62 CHECK_ERROR(aclrtMallocHost(&host, memSize));62 CHECK_ERROR(aclrtMallocHost(&host, memSize));
63 CHECK_ERROR(aclrtMemset(host, memSize, 0, memSize));63 CHECK_ERROR(aclrtMemset(host, memSize, 0, memSize));
64 64 
65- char *char_ptr = (char *)host;65+ char* char_ptr = (char*)host;
66 CHECK_ERROR(aclrtMemcpy(host, memSize, dev1, memSize, ACL_MEMCPY_DEVICE_TO_HOST));66 CHECK_ERROR(aclrtMemcpy(host, memSize, dev1, memSize, ACL_MEMCPY_DEVICE_TO_HOST));
67 INFO_LOG("Device 1 to host memcpy success.");67 INFO_LOG("Device 1 to host memcpy success.");
68 68 
Mexample/1_basic_features/event/0_event_status/main.cpp+6-6
@@ -21,7 +21,7 @@ int main()
21 int32_t deviceId = 0;21 int32_t deviceId = 0;
22 uint32_t blockDim = 1;22 uint32_t blockDim = 1;
23 uint32_t num = 0;23 uint32_t num = 0;
24- uint32_t *numDevice = nullptr;24+ uint32_t* numDevice = nullptr;
25 size_t size = sizeof(uint32_t);25 size_t size = sizeof(uint32_t);
26 aclrtContext context;26 aclrtContext context;
27 aclrtStream stream = nullptr;27 aclrtStream stream = nullptr;
@@ -33,17 +33,17 @@ int main()
33 CHECK_ERROR(aclrtSetDevice(deviceId));33 CHECK_ERROR(aclrtSetDevice(deviceId));
34 CHECK_ERROR(aclrtCreateContext(&context, deviceId));34 CHECK_ERROR(aclrtCreateContext(&context, deviceId));
35 CHECK_ERROR(aclrtCreateStream(&stream));35 CHECK_ERROR(aclrtCreateStream(&stream));
36- 36+ 
37 // 默认为遇错继续执行，调用该函数设置为遇错即停，当一个任务执行遇到错误，后续任务将不进行。37 // 默认为遇错继续执行，调用该函数设置为遇错即停，当一个任务执行遇到错误，后续任务将不进行。
38 CHECK_ERROR(aclrtSetStreamFailureMode(stream, ACL_STOP_ON_FAILURE));38 CHECK_ERROR(aclrtSetStreamFailureMode(stream, ACL_STOP_ON_FAILURE));
39 CHECK_ERROR(aclrtCreateEvent(&event));39 CHECK_ERROR(aclrtCreateEvent(&event));
40- 40+ 
41 // 查询创建后的event状态41 // 查询创建后的event状态
42 CHECK_ERROR(aclrtQueryEventStatus(event, &eventStatus));42 CHECK_ERROR(aclrtQueryEventStatus(event, &eventStatus));
43 INFO_LOG("After create event, current event status is %d.", eventStatus);43 INFO_LOG("After create event, current event status is %d.", eventStatus);
44- 44+ 
45 // 开始做长任务45 // 开始做长任务
46- CHECK_ERROR(aclrtMalloc((void **)&numDevice, size, ACL_MEM_MALLOC_HUGE_FIRST));46+ CHECK_ERROR(aclrtMalloc((void**)&numDevice, size, ACL_MEM_MALLOC_HUGE_FIRST));
47 CHECK_ERROR(aclrtMemcpy(numDevice, size, &num, size, ACL_MEMCPY_HOST_TO_DEVICE));47 CHECK_ERROR(aclrtMemcpy(numDevice, size, &num, size, ACL_MEMCPY_HOST_TO_DEVICE));
48 INFO_LOG("Applied resource successfully, begin assigning task.");48 INFO_LOG("Applied resource successfully, begin assigning task.");
49 INFO_LOG("Begin a long task.");49 INFO_LOG("Begin a long task.");
@@ -54,7 +54,7 @@ int main()
54 CHECK_ERROR(aclrtQueryEventStatus(event, &eventStatus));54 CHECK_ERROR(aclrtQueryEventStatus(event, &eventStatus));
55 INFO_LOG("0 is incompleted, 1 is completed.");55 INFO_LOG("0 is incompleted, 1 is completed.");
56 INFO_LOG("After record but before synchronize, current event status is %d.", eventStatus);56 INFO_LOG("After record but before synchronize, current event status is %d.", eventStatus);
57- 57+ 
58 // 查询同步后的event状态58 // 查询同步后的event状态
59 CHECK_ERROR(aclrtSynchronizeEvent(event));59 CHECK_ERROR(aclrtSynchronizeEvent(event));
60 CHECK_ERROR(aclrtQueryEventStatus(event, &eventStatus));60 CHECK_ERROR(aclrtQueryEventStatus(event, &eventStatus));
Mexample/1_basic_features/event/1_event_timestamp/main.cpp+4-4
@@ -23,7 +23,7 @@ int main()
23 int32_t deviceId = 0;23 int32_t deviceId = 0;
24 uint32_t blockDim = 1;24 uint32_t blockDim = 1;
25 uint32_t num = 0;25 uint32_t num = 0;
26- uint32_t *numDevice = nullptr;26+ uint32_t* numDevice = nullptr;
27 size_t size = sizeof(uint32_t);27 size_t size = sizeof(uint32_t);
28 uint64_t time = 0;28 uint64_t time = 0;
29 float useTime = 0;29 float useTime = 0;
@@ -43,7 +43,7 @@ int main()
43 CHECK_ERROR(aclrtCreateEvent(&endEvent));43 CHECK_ERROR(aclrtCreateEvent(&endEvent));
44 44 
45 // 先做一个短时任务查看耗时45 // 先做一个短时任务查看耗时
46- CHECK_ERROR(aclrtMalloc((void **)&numDevice, size, ACL_MEM_MALLOC_HUGE_FIRST));46+ CHECK_ERROR(aclrtMalloc((void**)&numDevice, size, ACL_MEM_MALLOC_HUGE_FIRST));
47 CHECK_ERROR(aclrtMemcpy(numDevice, size, &num, size, ACL_MEMCPY_HOST_TO_DEVICE));47 CHECK_ERROR(aclrtMemcpy(numDevice, size, &num, size, ACL_MEMCPY_HOST_TO_DEVICE));
48 INFO_LOG("Begin a short task.");48 INFO_LOG("Begin a short task.");
49 // 做任务前记录一次，做任务后再记录一次，相差时间则为任务耗时。49 // 做任务前记录一次，做任务后再记录一次，相差时间则为任务耗时。
@@ -62,10 +62,10 @@ int main()
62 INFO_LOG("End event timestamp: %ld.", time);62 INFO_LOG("End event timestamp: %ld.", time);
63 CHECK_ERROR(aclrtEventElapsedTime(&useTime, startEvent, endEvent));63 CHECK_ERROR(aclrtEventElapsedTime(&useTime, startEvent, endEvent));
64 INFO_LOG("The short task consume time %f ms.", useTime);64 INFO_LOG("The short task consume time %f ms.", useTime);
65- CHECK_ERROR(aclrtFree(numDevice));// 释放短任务分配的内存65+ CHECK_ERROR(aclrtFree(numDevice)); // 释放短任务分配的内存
66 66 
67 // 再做一个长耗时任务查看耗时67 // 再做一个长耗时任务查看耗时
68- CHECK_ERROR(aclrtMalloc((void **)&numDevice, size, ACL_MEM_MALLOC_HUGE_FIRST));68+ CHECK_ERROR(aclrtMalloc((void**)&numDevice, size, ACL_MEM_MALLOC_HUGE_FIRST));
69 CHECK_ERROR(aclrtMemcpy(numDevice, size, &num, size, ACL_MEMCPY_HOST_TO_DEVICE));69 CHECK_ERROR(aclrtMemcpy(numDevice, size, &num, size, ACL_MEMCPY_HOST_TO_DEVICE));
70 INFO_LOG("Begin a long task.");70 INFO_LOG("Begin a long task.");
71 CHECK_ERROR(aclrtRecordEvent(startEvent, stream));71 CHECK_ERROR(aclrtRecordEvent(startEvent, stream));
Mexample/1_basic_features/memory/0_h2h_memory_copy/main.cpp+2-2
@@ -26,8 +26,8 @@ int32_t main()
26 26 
27 // Allocate memory on the host27 // Allocate memory on the host
28 uint64_t size = 1 * 1024 * 1024;28 uint64_t size = 1 * 1024 * 1024;
29- int *hostPtrA;29+ int* hostPtrA;
30- int *hostPtrB;30+ int* hostPtrB;
31 CHECK_ERROR(aclrtMallocHost((void**)&hostPtrA, size));31 CHECK_ERROR(aclrtMallocHost((void**)&hostPtrA, size));
32 INFO_LOG("Allocate memory on the host memory %p successfully", hostPtrA);32 INFO_LOG("Allocate memory on the host memory %p successfully", hostPtrA);
33 33 
Mexample/1_basic_features/memory/10_ipc_memory_withpid/proc_a.cpp+1-1
@@ -31,7 +31,7 @@ int32_t main()
31 31 
32 // Allocate memory on the device32 // Allocate memory on the device
33 uint64_t size = 1 * 1024 * 1024;33 uint64_t size = 1 * 1024 * 1024;
34- int *devPtr;34+ int* devPtr;
35 CHECK_ERROR(aclrtMalloc((void**)&devPtr, size, ACL_MEM_MALLOC_HUGE_FIRST));35 CHECK_ERROR(aclrtMalloc((void**)&devPtr, size, ACL_MEM_MALLOC_HUGE_FIRST));
36 INFO_LOG("Allocate memory on the device %p successfully", devPtr);36 INFO_LOG("Allocate memory on the device %p successfully", devPtr);
37 37 
Mexample/1_basic_features/memory/10_ipc_memory_withpid/proc_b.cpp+1-1
@@ -43,7 +43,7 @@ int32_t main()
43 INFO_LOG("Process B: get the shareable memory identifier successfully, shareable identifier = %s", memName);43 INFO_LOG("Process B: get the shareable memory identifier successfully, shareable identifier = %s", memName);
44 44 
45 // Return a device memory pointer accessible to Process B45 // Return a device memory pointer accessible to Process B
46- void *devPtr = nullptr;46+ void* devPtr = nullptr;
47 uint64_t size = 1 * 1024 * 1024;47 uint64_t size = 1 * 1024 * 1024;
48 CHECK_ERROR(aclrtIpcMemImportByKey(&devPtr, memName, 0));48 CHECK_ERROR(aclrtIpcMemImportByKey(&devPtr, memName, 0));
49 49 
Mexample/1_basic_features/memory/11_ipc_memory_withoutpid/proc_a.cpp+1-1
@@ -31,7 +31,7 @@ int32_t main()
31 31 
32 // Allocate memory on the device32 // Allocate memory on the device
33 uint64_t size = 1 * 1024 * 1024;33 uint64_t size = 1 * 1024 * 1024;
34- int *devPtr;34+ int* devPtr;
35 CHECK_ERROR(aclrtMalloc((void**)&devPtr, size, ACL_MEM_MALLOC_HUGE_FIRST));35 CHECK_ERROR(aclrtMalloc((void**)&devPtr, size, ACL_MEM_MALLOC_HUGE_FIRST));
36 INFO_LOG("Allocate memory on the device %p successfully", devPtr);36 INFO_LOG("Allocate memory on the device %p successfully", devPtr);
37 37 
Mexample/1_basic_features/memory/11_ipc_memory_withoutpid/proc_b.cpp+1-1
@@ -33,7 +33,7 @@ int32_t main()
33 INFO_LOG("Process B: get the shareable memory identifier successfully, shareable identifier = %s", memName);33 INFO_LOG("Process B: get the shareable memory identifier successfully, shareable identifier = %s", memName);
34 34 
35 // Return a device memory pointer accessible to Process B35 // Return a device memory pointer accessible to Process B
36- void *devPtr = nullptr;36+ void* devPtr = nullptr;
37 uint64_t size = 1 * 1024 * 1024;37 uint64_t size = 1 * 1024 * 1024;
38 CHECK_ERROR(aclrtIpcMemImportByKey(&devPtr, memName, 0));38 CHECK_ERROR(aclrtIpcMemImportByKey(&devPtr, memName, 0));
39 39 
Mexample/1_basic_features/memory/12_cross_server_physical_memory_sharing_withoutpid/client.cpp+3-3
@@ -72,7 +72,7 @@ int32_t main(int argc, char* argv[])
72 aclrtMemFabricHandle shareableHandle = {};72 aclrtMemFabricHandle shareableHandle = {};
73 aclrtDrvMemHandle handle = nullptr;73 aclrtDrvMemHandle handle = nullptr;
74 socklen_t server_len = sizeof(server_addr);74 socklen_t server_len = sizeof(server_addr);
75- int recv_len = 75+ int recv_len =
76 recvfrom(sockfd, &shareableHandle, sizeof(shareableHandle), 0, (struct sockaddr*)&server_addr, &server_len);76 recvfrom(sockfd, &shareableHandle, sizeof(shareableHandle), 0, (struct sockaddr*)&server_addr, &server_len);
77 if (recv_len < 0) {77 if (recv_len < 0) {
78 close(sockfd);78 close(sockfd);
@@ -80,8 +80,8 @@ int32_t main(int argc, char* argv[])
80 }80 }
81 81 
82 // Import shareable handle from server82 // Import shareable handle from server
83- CHECK_ERROR(aclrtMemImportFromShareableHandleV2(&shareableHandle, ACL_MEM_SHARE_HANDLE_TYPE_FABRIC, 83+ CHECK_ERROR(aclrtMemImportFromShareableHandleV2(
84- ACL_RT_IPC_MEM_EXPORT_FLAG_DEFAULT, &handle));84+ &shareableHandle, ACL_MEM_SHARE_HANDLE_TYPE_FABRIC, ACL_RT_IPC_MEM_EXPORT_FLAG_DEFAULT, &handle));
85 INFO_LOG("Client: import shareable handle successfully");85 INFO_LOG("Client: import shareable handle successfully");
86 86 
87 const size_t data_size = 1024 * sizeof(float);87 const size_t data_size = 1024 * sizeof(float);
Mexample/1_basic_features/memory/12_cross_server_physical_memory_sharing_withoutpid/server.cpp+3-3
@@ -85,8 +85,8 @@ int32_t main(int argc, char* argv[])
85 // Export a shareable handle85 // Export a shareable handle
86 aclrtMemFabricHandle shareableHandle = {};86 aclrtMemFabricHandle shareableHandle = {};
87 aclrtMemSharedHandleType shareType = ACL_MEM_SHARE_HANDLE_TYPE_FABRIC;87 aclrtMemSharedHandleType shareType = ACL_MEM_SHARE_HANDLE_TYPE_FABRIC;
88- CHECK_ERROR(aclrtMemExportToShareableHandleV2(handle, ACL_RT_IPC_MEM_EXPORT_FLAG_DISABLE_PID_VALIDATION, 88+ CHECK_ERROR(aclrtMemExportToShareableHandleV2(
89- shareType, &shareableHandle));89+ handle, ACL_RT_IPC_MEM_EXPORT_FLAG_DISABLE_PID_VALIDATION, shareType, &shareableHandle));
90 INFO_LOG("Server: export shareable handle successfully");90 INFO_LOG("Server: export shareable handle successfully");
91 91 
92 // Transfer the shareable handle to client by UDP92 // Transfer the shareable handle to client by UDP
@@ -143,7 +143,7 @@ int32_t main(int argc, char* argv[])
143 close(sockfd);143 close(sockfd);
144 INFO_LOG("Server: ipc close successfully");144 INFO_LOG("Server: ipc close successfully");
145 145 
146- // Release memory resources146+ // Release memory resources
147 CHECK_ERROR(aclrtUnmapMem(virPtr));147 CHECK_ERROR(aclrtUnmapMem(virPtr));
148 CHECK_ERROR(aclrtReleaseMemAddress(virPtr));148 CHECK_ERROR(aclrtReleaseMemAddress(virPtr));
149 CHECK_ERROR(aclrtFreePhysical(handle));149 CHECK_ERROR(aclrtFreePhysical(handle));
Mexample/1_basic_features/memory/1_h2d_sync_memory_copy/main.cpp+2-2
@@ -27,8 +27,8 @@ int32_t main()
27 27 
28 // Allocate memory on the host and device28 // Allocate memory on the host and device
29 uint64_t size = 1 * 1024 * 1024;29 uint64_t size = 1 * 1024 * 1024;
30- int *hostPtrA;30+ int* hostPtrA;
31- int *devPtrB;31+ int* devPtrB;
32 CHECK_ERROR(aclrtMallocHost((void**)&hostPtrA, size));32 CHECK_ERROR(aclrtMallocHost((void**)&hostPtrA, size));
33 INFO_LOG("Allocate memory on the host memory %p successfully", hostPtrA);33 INFO_LOG("Allocate memory on the host memory %p successfully", hostPtrA);
34 34 
Mexample/1_basic_features/memory/2_h2d_async_memory_copy/main.cpp+2-2
@@ -27,8 +27,8 @@ int32_t main()
27 27 
28 // Allocate memory on the host and device28 // Allocate memory on the host and device
29 uint64_t size = 1 * 1024 * 1024;29 uint64_t size = 1 * 1024 * 1024;
30- int *hostPtrA;30+ int* hostPtrA;
31- int *devPtrB;31+ int* devPtrB;
32 CHECK_ERROR(aclrtMallocHost((void**)&hostPtrA, size));32 CHECK_ERROR(aclrtMallocHost((void**)&hostPtrA, size));
33 INFO_LOG("Allocate memory on the host memory %p successfully", hostPtrA);33 INFO_LOG("Allocate memory on the host memory %p successfully", hostPtrA);
34 34 
Mexample/1_basic_features/memory/3_d2h_sync_memory_copy/main.cpp+2-2
@@ -27,8 +27,8 @@ int32_t main()
27 27 
28 // Allocate memory on the host and device28 // Allocate memory on the host and device
29 uint64_t size = 1 * 1024 * 1024;29 uint64_t size = 1 * 1024 * 1024;
30- int *hostPtrA;30+ int* hostPtrA;
31- int *devPtrB;31+ int* devPtrB;
32 CHECK_ERROR(aclrtMallocHost((void**)&hostPtrA, size));32 CHECK_ERROR(aclrtMallocHost((void**)&hostPtrA, size));
33 INFO_LOG("Allocate memory on the host memory %p successfully", hostPtrA);33 INFO_LOG("Allocate memory on the host memory %p successfully", hostPtrA);
34 34 
Mexample/1_basic_features/memory/4_d2h_async_memory_copy/main.cpp+2-2
@@ -27,8 +27,8 @@ int32_t main()
27 27 
28 // Allocate memory on the host and device28 // Allocate memory on the host and device
29 uint64_t size = 1 * 1024 * 1024;29 uint64_t size = 1 * 1024 * 1024;
30- int *hostPtrA;30+ int* hostPtrA;
31- int *devPtrB;31+ int* devPtrB;
32 CHECK_ERROR(aclrtMallocHost((void**)&hostPtrA, size));32 CHECK_ERROR(aclrtMallocHost((void**)&hostPtrA, size));
33 INFO_LOG("Allocate memory on the host memory %p successfully", hostPtrA);33 INFO_LOG("Allocate memory on the host memory %p successfully", hostPtrA);
34 34 
Mexample/1_basic_features/memory/5_d2d_sync_memory_copy/main.cpp+2-2
@@ -27,8 +27,8 @@ int32_t main()
27 27 
28 // Allocate memory on the device28 // Allocate memory on the device
29 uint64_t size = 1 * 1024 * 1024;29 uint64_t size = 1 * 1024 * 1024;
30- int *devPtrA;30+ int* devPtrA;
31- int *devPtrB;31+ int* devPtrB;
32 CHECK_ERROR(aclrtMalloc((void**)&devPtrA, size, ACL_MEM_MALLOC_HUGE_FIRST));32 CHECK_ERROR(aclrtMalloc((void**)&devPtrA, size, ACL_MEM_MALLOC_HUGE_FIRST));
33 INFO_LOG("Allocate memory on the device memory %p successfully", devPtrA);33 INFO_LOG("Allocate memory on the device memory %p successfully", devPtrA);
34 34 
Mexample/1_basic_features/memory/6_d2d_async_memory_copy/main.cpp+2-2
@@ -27,8 +27,8 @@ int32_t main()
27 27 
28 // Allocate memory on the device28 // Allocate memory on the device
29 uint64_t size = 1 * 1024 * 1024;29 uint64_t size = 1 * 1024 * 1024;
30- int *devPtrA;30+ int* devPtrA;
31- int *devPtrB;31+ int* devPtrB;
32 CHECK_ERROR(aclrtMalloc((void**)&devPtrA, size, ACL_MEM_MALLOC_HUGE_FIRST));32 CHECK_ERROR(aclrtMalloc((void**)&devPtrA, size, ACL_MEM_MALLOC_HUGE_FIRST));
33 INFO_LOG("Allocate memory on the device memory %p successfully", devPtrA);33 INFO_LOG("Allocate memory on the device memory %p successfully", devPtrA);
34 34 
Mexample/1_basic_features/memory/7_physical_memory_sharing_withpid/proc_a.cpp+9-5
@@ -38,10 +38,12 @@ int32_t main()
38 38 
39 size_t granularity = 0UL;39 size_t granularity = 0UL;
40 CHECK_ERROR(aclrtMemGetAllocationGranularity(&prop, ACL_RT_MEM_ALLOC_GRANULARITY_MINIMUM, &granularity));40 CHECK_ERROR(aclrtMemGetAllocationGranularity(&prop, ACL_RT_MEM_ALLOC_GRANULARITY_MINIMUM, &granularity));
41- INFO_LOG("Process A: get memory allocation granularity successfully, granularity = %d", static_cast<int32_t>(granularity));41+ INFO_LOG(
42+ "Process A: get memory allocation granularity successfully, granularity = %d",
43+ static_cast<int32_t>(granularity));
42 44 
43 // Start allocating physical memory based on memory allocation granularity45 // Start allocating physical memory based on memory allocation granularity
44- size_t alignedSize = ((dataSize + granularity - 1U) / granularity) * granularity;46+ size_t alignedSize = ((dataSize + granularity - 1U) / granularity) * granularity;
45 INFO_LOG("Process A: aligned size = %zu", alignedSize);47 INFO_LOG("Process A: aligned size = %zu", alignedSize);
46 48 
47 aclrtDrvMemHandle handle = nullptr;49 aclrtDrvMemHandle handle = nullptr;
@@ -49,7 +51,7 @@ int32_t main()
49 INFO_LOG("Process A: allocate physical memory successfully");51 INFO_LOG("Process A: allocate physical memory successfully");
50 52 
51 // Reserve virtual memory53 // Reserve virtual memory
52- void *virPtr;54+ void* virPtr;
53 CHECK_ERROR(aclrtReserveMemAddress(&virPtr, granularity, 0, nullptr, 0));55 CHECK_ERROR(aclrtReserveMemAddress(&virPtr, granularity, 0, nullptr, 0));
54 INFO_LOG("Process A: reserve virtual memory successfully");56 INFO_LOG("Process A: reserve virtual memory successfully");
55 57 
@@ -68,13 +70,15 @@ int32_t main()
68 // Write the value to the virtual memory70 // Write the value to the virtual memory
69 constexpr uint32_t blockDim = 1;71 constexpr uint32_t blockDim = 1;
70 int writeValue = 123;72 int writeValue = 123;
71- WriteDo(blockDim, stream, (int *)virPtr, writeValue);73+ WriteDo(blockDim, stream, (int*)virPtr, writeValue);
72 INFO_LOG("Write data %d to the virtual address %p", writeValue, virPtr);74 INFO_LOG("Write data %d to the virtual address %p", writeValue, virPtr);
73 75 
74 // Export a shareable handle76 // Export a shareable handle
75 uint64_t shareableHandle = 0ULL;77 uint64_t shareableHandle = 0ULL;
76 CHECK_ERROR(aclrtMemExportToShareableHandle(handle, ACL_MEM_HANDLE_TYPE_NONE, 0, &shareableHandle));78 CHECK_ERROR(aclrtMemExportToShareableHandle(handle, ACL_MEM_HANDLE_TYPE_NONE, 0, &shareableHandle));
77- INFO_LOG("Process A: export a shareable handle successfully, shareable handle = %d", static_cast<int32_t>(shareableHandle));79+ INFO_LOG(
80+ "Process A: export a shareable handle successfully, shareable handle = %d",
81+ static_cast<int32_t>(shareableHandle));
78 82 
79 // Read Process B's pid from the file83 // Read Process B's pid from the file
80 int32_t pid = 0;84 int32_t pid = 0;
Mexample/1_basic_features/memory/7_physical_memory_sharing_withpid/proc_b.cpp+8-5
@@ -39,7 +39,8 @@ int32_t main()
39 uint64_t shareableHandle = 0ULL;39 uint64_t shareableHandle = 0ULL;
40 aclrtDrvMemHandle handle = nullptr;40 aclrtDrvMemHandle handle = nullptr;
41 memory::ReadFile("file/handle.bin", "file/handle.bin.done", &shareableHandle, sizeof(shareableHandle));41 memory::ReadFile("file/handle.bin", "file/handle.bin.done", &shareableHandle, sizeof(shareableHandle));
42- INFO_LOG("Process B: get a shareable handle successfully, shareable handle = %d", static_cast<int32_t>(shareableHandle));42+ INFO_LOG(
43+ "Process B: get a shareable handle successfully, shareable handle = %d", static_cast<int32_t>(shareableHandle));
43 44 
44 CHECK_ERROR(aclrtMemImportFromShareableHandle(shareableHandle, deviceId, &handle));45 CHECK_ERROR(aclrtMemImportFromShareableHandle(shareableHandle, deviceId, &handle));
45 46 
@@ -54,10 +55,12 @@ int32_t main()
54 55 
55 size_t granularity = 0UL;56 size_t granularity = 0UL;
56 CHECK_ERROR(aclrtMemGetAllocationGranularity(&prop, ACL_RT_MEM_ALLOC_GRANULARITY_MINIMUM, &granularity));57 CHECK_ERROR(aclrtMemGetAllocationGranularity(&prop, ACL_RT_MEM_ALLOC_GRANULARITY_MINIMUM, &granularity));
57- INFO_LOG("Process B: get memory allocation granularity successfully, granularity = %d", static_cast<int32_t>(granularity));58+ INFO_LOG(
59+ "Process B: get memory allocation granularity successfully, granularity = %d",
60+ static_cast<int32_t>(granularity));
58 61 
59 // Reserve virtual memory62 // Reserve virtual memory
60- void *virPtr = nullptr;63+ void* virPtr = nullptr;
61 CHECK_ERROR(aclrtReserveMemAddress(&virPtr, granularity, 0, nullptr, 0));64 CHECK_ERROR(aclrtReserveMemAddress(&virPtr, granularity, 0, nullptr, 0));
62 INFO_LOG("Process B: reserve virtual memory successfully");65 INFO_LOG("Process B: reserve virtual memory successfully");
63 66 
@@ -74,7 +77,7 @@ int32_t main()
74 INFO_LOG("Process B: set memory access permissions successfully");77 INFO_LOG("Process B: set memory access permissions successfully");
75 78 
76 // Copy memory from device to host79 // Copy memory from device to host
77- int *hostPtrA;80+ int* hostPtrA;
78 CHECK_ERROR(aclrtMallocHost(reinterpret_cast<void**>(&hostPtrA), granularity));81 CHECK_ERROR(aclrtMallocHost(reinterpret_cast<void**>(&hostPtrA), granularity));
79 CHECK_ERROR(aclrtMemcpy(hostPtrA, granularity, virPtr, granularity, ACL_MEMCPY_DEVICE_TO_HOST));82 CHECK_ERROR(aclrtMemcpy(hostPtrA, granularity, virPtr, granularity, ACL_MEMCPY_DEVICE_TO_HOST));
80 INFO_LOG("Process B: copy memory from device address %p to host address %p", virPtr, hostPtrA);83 INFO_LOG("Process B: copy memory from device address %p to host address %p", virPtr, hostPtrA);
@@ -102,4 +105,4 @@ int32_t main()
102 aclrtResetDeviceForce(deviceId);105 aclrtResetDeviceForce(deviceId);
103 aclFinalize();106 aclFinalize();
104 return 0;107 return 0;
105-}108+}
Mexample/1_basic_features/memory/8_physical_memory_sharing_withoutpid/proc_a.cpp+8-5
@@ -13,7 +13,6 @@
13 * using two independent processes (i.e., Process A and Process B) on a single device.13 * using two independent processes (i.e., Process A and Process B) on a single device.
14 */14 */
15 15 
16- 
17#include "acl/acl.h"16#include "acl/acl.h"
18#include "kernel_func/kernel_ops.h"17#include "kernel_func/kernel_ops.h"
19#include "utils.h"18#include "utils.h"
@@ -39,7 +38,9 @@ int32_t main()
39 38 
40 size_t granularity = 0UL;39 size_t granularity = 0UL;
41 CHECK_ERROR(aclrtMemGetAllocationGranularity(&prop, ACL_RT_MEM_ALLOC_GRANULARITY_MINIMUM, &granularity));40 CHECK_ERROR(aclrtMemGetAllocationGranularity(&prop, ACL_RT_MEM_ALLOC_GRANULARITY_MINIMUM, &granularity));
42- INFO_LOG("Process A: get memory allocation granularity successfully, granularity = %d", static_cast<int32_t>(granularity));41+ INFO_LOG(
42+ "Process A: get memory allocation granularity successfully, granularity = %d",
43+ static_cast<int32_t>(granularity));
43 44 
44 // Start allocating physical memory based on memory allocation granularity45 // Start allocating physical memory based on memory allocation granularity
45 size_t alignedSize = ((dataSize + granularity - 1U) / granularity) * granularity;46 size_t alignedSize = ((dataSize + granularity - 1U) / granularity) * granularity;
@@ -50,7 +51,7 @@ int32_t main()
50 INFO_LOG("Process A: allocate physical memory successfully");51 INFO_LOG("Process A: allocate physical memory successfully");
51 52 
52 // Reserve virtual memory53 // Reserve virtual memory
53- void *virPtr;54+ void* virPtr;
54 CHECK_ERROR(aclrtReserveMemAddress(&virPtr, granularity, 0, nullptr, 0));55 CHECK_ERROR(aclrtReserveMemAddress(&virPtr, granularity, 0, nullptr, 0));
55 INFO_LOG("Process A: reserve virtual memory successfully");56 INFO_LOG("Process A: reserve virtual memory successfully");
56 57 
@@ -69,13 +70,15 @@ int32_t main()
69 // Write the value to the virtual memory70 // Write the value to the virtual memory
70 constexpr uint32_t blockDim = 1;71 constexpr uint32_t blockDim = 1;
71 int writeValue = 123;72 int writeValue = 123;
72- WriteDo(blockDim, stream, (int *)virPtr, writeValue);73+ WriteDo(blockDim, stream, (int*)virPtr, writeValue);
73 INFO_LOG("Write the data %d to the virtual address %p", writeValue, virPtr);74 INFO_LOG("Write the data %d to the virtual address %p", writeValue, virPtr);
74 75 
75 // Export a shareable handle76 // Export a shareable handle
76 uint64_t shareableHandle = 0ULL;77 uint64_t shareableHandle = 0ULL;
77 CHECK_ERROR(aclrtMemExportToShareableHandle(handle, ACL_MEM_HANDLE_TYPE_NONE, 1, &shareableHandle));78 CHECK_ERROR(aclrtMemExportToShareableHandle(handle, ACL_MEM_HANDLE_TYPE_NONE, 1, &shareableHandle));
78- INFO_LOG("Process A: export a shareable handle successfully, shareable handle = %d", static_cast<int32_t>(shareableHandle));79+ INFO_LOG(
80+ "Process A: export a shareable handle successfully, shareable handle = %d",
81+ static_cast<int32_t>(shareableHandle));
79 82 
80 // Transfer the shareable handle to Process B by writing it to the file83 // Transfer the shareable handle to Process B by writing it to the file
81 memory::WriteFile("file/handle.bin", "file/handle.bin.done", &shareableHandle, sizeof(shareableHandle));84 memory::WriteFile("file/handle.bin", "file/handle.bin.done", &shareableHandle, sizeof(shareableHandle));
Mexample/1_basic_features/memory/8_physical_memory_sharing_withoutpid/proc_b.cpp+7-5
@@ -13,7 +13,6 @@
13 * using two independent processes (i.e., Process A and Process B) on a single device.13 * using two independent processes (i.e., Process A and Process B) on a single device.
14 */14 */
15 15 
16- 
17#include "acl/acl.h"16#include "acl/acl.h"
18#include "utils.h"17#include "utils.h"
19#include "mem_utils.h"18#include "mem_utils.h"
@@ -30,7 +29,8 @@ int32_t main()
30 uint64_t shareableHandle = 0ULL;29 uint64_t shareableHandle = 0ULL;
31 aclrtDrvMemHandle handle = nullptr;30 aclrtDrvMemHandle handle = nullptr;
32 memory::ReadFile("file/handle.bin", "file/handle.bin.done", &shareableHandle, sizeof(shareableHandle));31 memory::ReadFile("file/handle.bin", "file/handle.bin.done", &shareableHandle, sizeof(shareableHandle));
33- INFO_LOG("Process B: get a shareable handle successfully, shareable handle = %d", static_cast<int32_t>(shareableHandle));32+ INFO_LOG(
33+ "Process B: get a shareable handle successfully, shareable handle = %d", static_cast<int32_t>(shareableHandle));
34 34 
35 CHECK_ERROR(aclrtMemImportFromShareableHandle(shareableHandle, deviceId, &handle));35 CHECK_ERROR(aclrtMemImportFromShareableHandle(shareableHandle, deviceId, &handle));
36 36 
@@ -45,10 +45,12 @@ int32_t main()
45 45 
46 size_t granularity = 0UL;46 size_t granularity = 0UL;
47 CHECK_ERROR(aclrtMemGetAllocationGranularity(&prop, ACL_RT_MEM_ALLOC_GRANULARITY_MINIMUM, &granularity));47 CHECK_ERROR(aclrtMemGetAllocationGranularity(&prop, ACL_RT_MEM_ALLOC_GRANULARITY_MINIMUM, &granularity));
48- INFO_LOG("Process B: get memory allocation granularity successfully, granularity = %d", static_cast<int32_t>(granularity));48+ INFO_LOG(
49+ "Process B: get memory allocation granularity successfully, granularity = %d",
50+ static_cast<int32_t>(granularity));
49 51 
50 // Reserve virtual memory52 // Reserve virtual memory
51- void *virPtr = nullptr;53+ void* virPtr = nullptr;
52 CHECK_ERROR(aclrtReserveMemAddress(&virPtr, granularity, 0, nullptr, 0));54 CHECK_ERROR(aclrtReserveMemAddress(&virPtr, granularity, 0, nullptr, 0));
53 INFO_LOG("Process B: reserve virtual memory successfully");55 INFO_LOG("Process B: reserve virtual memory successfully");
54 56 
@@ -65,7 +67,7 @@ int32_t main()
65 INFO_LOG("Process B: set memory access permissions successfully");67 INFO_LOG("Process B: set memory access permissions successfully");
66 68 
67 // Copy memory from device to host69 // Copy memory from device to host
68- int *hostPtrA;70+ int* hostPtrA;
69 CHECK_ERROR(aclrtMallocHost(reinterpret_cast<void**>(&hostPtrA), granularity));71 CHECK_ERROR(aclrtMallocHost(reinterpret_cast<void**>(&hostPtrA), granularity));
70 CHECK_ERROR(aclrtMemcpy(hostPtrA, granularity, virPtr, granularity, ACL_MEMCPY_DEVICE_TO_HOST));72 CHECK_ERROR(aclrtMemcpy(hostPtrA, granularity, virPtr, granularity, ACL_MEMCPY_DEVICE_TO_HOST));
71 INFO_LOG("Process B: copy memory from device address %p to host address %p", virPtr, hostPtrA);73 INFO_LOG("Process B: copy memory from device address %p to host address %p", virPtr, hostPtrA);
Mexample/1_basic_features/memory/9_multistream_sync_memory/file_ops.cpp+6-4
@@ -16,7 +16,8 @@
16#include <unistd.h>16#include <unistd.h>
17 17 
18namespace memory {18namespace memory {
19-void WriteFileEx(const char* file, void* data, size_t size) {19+void WriteFileEx(const char* file, void* data, size_t size)
20+{
20 INFO_LOG("Start writing to %s", file);21 INFO_LOG("Start writing to %s", file);
21 22 
22 int fd = open(file, O_WRONLY | O_CREAT | O_TRUNC, S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH);23 int fd = open(file, O_WRONLY | O_CREAT | O_TRUNC, S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH);
@@ -39,19 +40,20 @@ void WriteFileEx(const char* file, void* data, size_t size) {
39} // namespace memory40} // namespace memory
40 41 
41namespace memory {42namespace memory {
42-void ReadFileEx(const char* file, void* data, size_t size) {43+void ReadFileEx(const char* file, void* data, size_t size)
44+{
43 INFO_LOG("Start reading from %s", file);45 INFO_LOG("Start reading from %s", file);
44 46 
45 // Read data from the target file47 // Read data from the target file
46 int fd = open(file, O_RDONLY);48 int fd = open(file, O_RDONLY);
47- if (fd == -1){49+ if (fd == -1) {
48 ERROR_LOG("Open file %s failed", file);50 ERROR_LOG("Open file %s failed", file);
49 (void)close(fd);51 (void)close(fd);
50 return;52 return;
51 }53 }
52 54 
53 ssize_t readSize = read(fd, data, size);55 ssize_t readSize = read(fd, data, size);
54- if (readSize != static_cast<ssize_t>(size)){56+ if (readSize != static_cast<ssize_t>(size)) {
55 ERROR_LOG("Partial read");57 ERROR_LOG("Partial read");
56 (void)close(fd);58 (void)close(fd);
57 return;59 return;
Mexample/1_basic_features/memory/9_multistream_sync_memory/main.cpp+5-4
@@ -13,7 +13,6 @@
13 * using two independent streams (i.e., Stream A and Stream B).13 * using two independent streams (i.e., Stream A and Stream B).
14 */14 */
15 15 
16- 
17#include "acl/acl.h"16#include "acl/acl.h"
18#include "utils.h"17#include "utils.h"
19#include "mem_utils.h"18#include "mem_utils.h"
@@ -23,7 +22,8 @@
23#include <fcntl.h>22#include <fcntl.h>
24#include <unistd.h>23#include <unistd.h>
25 24 
26-int ThreadWait(aclrtStream stream, int32_t deviceId, void* devPtr, uint64_t valueCompare, const char* filePath){25+int ThreadWait(aclrtStream stream, int32_t deviceId, void* devPtr, uint64_t valueCompare, const char* filePath)
26+{
27 // Wait for the data in the specified memory to meet the condition27 // Wait for the data in the specified memory to meet the condition
28 aclrtSetDevice(deviceId);28 aclrtSetDevice(deviceId);
29 29 
@@ -42,7 +42,8 @@ int ThreadWait(aclrtStream stream, int32_t deviceId, void* devPtr, uint64_t valu
42 return 0;42 return 0;
43}43}
44 44 
45-int ThreadWrite(aclrtStream stream, int32_t deviceId, void* devPtr, uint64_t valueWrite, const char* filePath){45+int ThreadWrite(aclrtStream stream, int32_t deviceId, void* devPtr, uint64_t valueWrite, const char* filePath)
46+{
46 // The writing thread writes a flag to check whether the waiting thread is blocked before the writing thread starts47 // The writing thread writes a flag to check whether the waiting thread is blocked before the writing thread starts
47 int32_t writeFlag = 123;48 int32_t writeFlag = 123;
48 memory::WriteFileEx(filePath, &writeFlag, sizeof(writeFlag));49 memory::WriteFileEx(filePath, &writeFlag, sizeof(writeFlag));
@@ -67,7 +68,7 @@ int32_t main()
67 68 
68 // Allocate memory on the device69 // Allocate memory on the device
69 uint64_t size = 1 * 1024 * 1024;70 uint64_t size = 1 * 1024 * 1024;
70- void *devPtrA;71+ void* devPtrA;
71 CHECK_ERROR(aclrtMalloc(&devPtrA, size, ACL_MEM_MALLOC_HUGE_FIRST));72 CHECK_ERROR(aclrtMalloc(&devPtrA, size, ACL_MEM_MALLOC_HUGE_FIRST));
72 INFO_LOG("Allocate memory on the device successfully");73 INFO_LOG("Allocate memory on the device successfully");
73 74 
Mexample/1_basic_features/memory/file_ops.cpp+12-10
@@ -7,7 +7,7 @@
7 * INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE.7 * INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE.
8 * See LICENSE in the root of the software repository for the full text of the License.8 * See LICENSE in the root of the software repository for the full text of the License.
9 */9 */
10- 10+ 
11#include "utils.h"11#include "utils.h"
12#include "mem_utils.h"12#include "mem_utils.h"
13#include <cstdio>13#include <cstdio>
@@ -17,7 +17,7 @@
17#include <sys/stat.h>17#include <sys/stat.h>
18 18 
19namespace {19namespace {
20-bool WriteBufferToFile(int fd, const void *data, size_t size)20+bool WriteBufferToFile(int fd, const void* data, size_t size)
21{21{
22 const ssize_t writeSize = write(fd, data, size);22 const ssize_t writeSize = write(fd, data, size);
23 if (writeSize == static_cast<ssize_t>(size)) {23 if (writeSize == static_cast<ssize_t>(size)) {
@@ -27,7 +27,7 @@ bool WriteBufferToFile(int fd, const void *data, size_t size)
27 return false;27 return false;
28}28}
29 29 
30-bool ReadBufferFromFile(int fd, void *data, size_t size)30+bool ReadBufferFromFile(int fd, void* data, size_t size)
31{31{
32 const ssize_t readSize = read(fd, data, size);32 const ssize_t readSize = read(fd, data, size);
33 if (readSize == static_cast<ssize_t>(size)) {33 if (readSize == static_cast<ssize_t>(size)) {
@@ -39,7 +39,8 @@ bool ReadBufferFromFile(int fd, void *data, size_t size)
39} // namespace39} // namespace
40 40 
41namespace memory {41namespace memory {
42-void WriteFile(const char* filePath, const char* doneFile, void* data, size_t size) {42+void WriteFile(const char* filePath, const char* doneFile, void* data, size_t size)
43+{
43 INFO_LOG("Start writing to %s", filePath);44 INFO_LOG("Start writing to %s", filePath);
44 45 
45 int fd = open(filePath, O_WRONLY | O_CREAT | O_TRUNC, S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH);46 int fd = open(filePath, O_WRONLY | O_CREAT | O_TRUNC, S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH);
@@ -66,7 +67,8 @@ void WriteFile(const char* filePath, const char* doneFile, void* data, size_t si
66} // namespace memory67} // namespace memory
67 68 
68namespace memory {69namespace memory {
69-void ReadFile(const char* filePath, const char* doneFile, void* data, size_t bufferSize) {70+void ReadFile(const char* filePath, const char* doneFile, void* data, size_t bufferSize)
71+{
70 INFO_LOG("Start reading from %s", filePath);72 INFO_LOG("Start reading from %s", filePath);
71 constexpr uint32_t waitTime = 1000000;73 constexpr uint32_t waitTime = 1000000;
72 int attempts = 0;74 int attempts = 0;
@@ -82,7 +84,7 @@ void ReadFile(const char* filePath, const char* doneFile, void* data, size_t buf
82 attempts++;84 attempts++;
83 WARN_LOG("Data writing not yet complete");85 WARN_LOG("Data writing not yet complete");
84 if (attempts >= attemptsMax) {86 if (attempts >= attemptsMax) {
85- ERROR_LOG("Open file %s failed", doneFile); 87+ ERROR_LOG("Open file %s failed", doneFile);
86 return;88 return;
87 }89 }
88 (void)usleep(waitTime);90 (void)usleep(waitTime);
@@ -91,7 +93,7 @@ void ReadFile(const char* filePath, const char* doneFile, void* data, size_t buf
91 93 
92 // Read data from the target file94 // Read data from the target file
93 int fd = open(filePath, O_RDONLY);95 int fd = open(filePath, O_RDONLY);
94- if (fd == -1){96+ if (fd == -1) {
95 ERROR_LOG("Open file %s failed", filePath);97 ERROR_LOG("Open file %s failed", filePath);
96 return;98 return;
97 }99 }
@@ -104,9 +106,9 @@ void ReadFile(const char* filePath, const char* doneFile, void* data, size_t buf
104 }106 }
105 size_t fileSize = static_cast<size_t>(fileStat.st_size);107 size_t fileSize = static_cast<size_t>(fileStat.st_size);
106 if (fileSize > bufferSize) {108 if (fileSize > bufferSize) {
107- ERROR_LOG("File size %zu exceeds buffer size %zu", fileSize, bufferSize);109+ ERROR_LOG("File size %zu exceeds buffer size %zu", fileSize, bufferSize);
108- (void)close(fd);110+ (void)close(fd);
109- return;111+ return;
110 }112 }
111 113 
112 if (!ReadBufferFromFile(fd, data, fileSize)) {114 if (!ReadBufferFromFile(fd, data, fileSize)) {
Mexample/1_basic_features/memory/mem_utils.h+6-6
@@ -14,9 +14,9 @@
14#include <cstddef>14#include <cstddef>
15 15 
16namespace memory {16namespace memory {
17- void WriteFile(const char* filePath, const char* doneFile, void* data, size_t size);17+void WriteFile(const char* filePath, const char* doneFile, void* data, size_t size);
18- void ReadFile(const char* filePath, const char* doneFile, void* data, size_t bufferSize);18+void ReadFile(const char* filePath, const char* doneFile, void* data, size_t bufferSize);
19- void WriteFileEx(const char* filePath, void* data, size_t size);19+void WriteFileEx(const char* filePath, void* data, size_t size);
20- void ReadFileEx(const char* filePath, void* data, size_t size);20+void ReadFileEx(const char* filePath, void* data, size_t size);
21-}21+} // namespace memory
22-#endif22+#endif
Mexample/1_basic_features/memory/memory_runtime_utils.h+1-1
@@ -14,7 +14,7 @@
14#include "runtime_init_utils.h"14#include "runtime_init_utils.h"
15 15 
16namespace memory {16namespace memory {
17-inline int InitDeviceAndCreateStream(int32_t deviceId, aclrtStream *stream)17+inline int InitDeviceAndCreateStream(int32_t deviceId, aclrtStream* stream)
18{18{
19 return static_cast<int>(runtime::InitRuntimeAndCreateStream(deviceId, stream));19 return static_cast<int>(runtime::InitRuntimeAndCreateStream(deviceId, stream));
20}20}
Mexample/1_basic_features/stream/0_simple_stream/main.cpp+2-2
@@ -24,7 +24,7 @@ int main()
24 int32_t deviceId = 0;24 int32_t deviceId = 0;
25 uint32_t blockDim = 1;25 uint32_t blockDim = 1;
26 uint32_t num = 0;26 uint32_t num = 0;
27- uint32_t *numDevice = nullptr;27+ uint32_t* numDevice = nullptr;
28 size_t size = sizeof(uint32_t);28 size_t size = sizeof(uint32_t);
29 aclrtStream stream = nullptr;29 aclrtStream stream = nullptr;
30 aclrtStreamStatus streamStatus;30 aclrtStreamStatus streamStatus;
@@ -54,7 +54,7 @@ int main()
54 CHECK_ERROR(aclrtSynchronizeStream(stream));54 CHECK_ERROR(aclrtSynchronizeStream(stream));
55 CHECK_ERROR(aclrtMemcpy(&num, size, numDevice, size, ACL_MEMCPY_DEVICE_TO_HOST));55 CHECK_ERROR(aclrtMemcpy(&num, size, numDevice, size, ACL_MEMCPY_DEVICE_TO_HOST));
56 INFO_LOG("After assigning the task through the created stream, the current result is: %d.", num);56 INFO_LOG("After assigning the task through the created stream, the current result is: %d.", num);
57- 57+ 
58 INFO_LOG("Begin 3000 task.");58 INFO_LOG("Begin 3000 task.");
59 // 循环下发任务并检查下发状态59 // 循环下发任务并检查下发状态
60 for (int i = 0; i < count; i++) {60 for (int i = 0; i < count; i++) {
Mexample/1_basic_features/stream/1_stream_with_failure_mode/main.cpp+14-13
@@ -14,13 +14,13 @@
14#include "kernel_func/kernel_ops.h"14#include "kernel_func/kernel_ops.h"
15 15 
16namespace {16namespace {
17-void StreamStateCallback(aclrtStream stream, aclrtStreamState state, void *args)17+void StreamStateCallback(aclrtStream stream, aclrtStreamState state, void* args)
18{18{
19- const char *sampleName = args == nullptr ? "stream_failure_mode" : static_cast<const char *>(args);19+ const char* sampleName = args == nullptr ? "stream_failure_mode" : static_cast<const char*>(args);
20 INFO_LOG("Stream state callback from %s: stream=%p state=%d", sampleName, stream, static_cast<int32_t>(state));20 INFO_LOG("Stream state callback from %s: stream=%p state=%d", sampleName, stream, static_cast<int32_t>(state));
21}21}
22 22 
23-void LogOptionalResult(const char *apiName, aclError ret)23+void LogOptionalResult(const char* apiName, aclError ret)
24{24{
25 if (ret == ACL_SUCCESS) {25 if (ret == ACL_SUCCESS) {
26 INFO_LOG("%s returned %d.", apiName, static_cast<int32_t>(ret));26 INFO_LOG("%s returned %d.", apiName, static_cast<int32_t>(ret));
@@ -34,28 +34,29 @@ void LogAbortSyncResult(aclError ret)
34 if (ret == ACL_SUCCESS) {34 if (ret == ACL_SUCCESS) {
35 INFO_LOG("aclrtSynchronizeStream(after abort) returned %d.", static_cast<int32_t>(ret));35 INFO_LOG("aclrtSynchronizeStream(after abort) returned %d.", static_cast<int32_t>(ret));
36 } else {36 } else {
37- INFO_LOG("aclrtSynchronizeStream(after abort) returned %d after stream abort, which is expected.",37+ INFO_LOG(
38- static_cast<int32_t>(ret));38+ "aclrtSynchronizeStream(after abort) returned %d after stream abort, which is expected.",
39+ static_cast<int32_t>(ret));
39 }40 }
40}41}
41 42 
42-void FreeOptionalBuffer(uint32_t *buffer, const char *name)43+void FreeOptionalBuffer(uint32_t* buffer, const char* name)
43{44{
44 if (buffer != nullptr) {45 if (buffer != nullptr) {
45 LogOptionalResult(name, aclrtFree(buffer));46 LogOptionalResult(name, aclrtFree(buffer));
46 }47 }
47}48}
48 49 
49-void DestroyOptionalStream(aclrtStream stream, const char *name)50+void DestroyOptionalStream(aclrtStream stream, const char* name)
50{51{
51 if (stream != nullptr) {52 if (stream != nullptr) {
52 LogOptionalResult(name, aclrtDestroyStreamForce(stream));53 LogOptionalResult(name, aclrtDestroyStreamForce(stream));
53 }54 }
54}55}
55 56 
56-bool PrepareScratchBuffer(uint32_t **scratch, size_t size, const char *demoName)57+bool PrepareScratchBuffer(uint32_t** scratch, size_t size, const char* demoName)
57{58{
58- aclError ret = aclrtMalloc(reinterpret_cast<void **>(scratch), size, ACL_MEM_MALLOC_HUGE_FIRST);59+ aclError ret = aclrtMalloc(reinterpret_cast<void**>(scratch), size, ACL_MEM_MALLOC_HUGE_FIRST);
59 if (ret != ACL_SUCCESS) {60 if (ret != ACL_SUCCESS) {
60 WARN_LOG("Skip %s demo: aclrtMalloc returned %d.", demoName, static_cast<int32_t>(ret));61 WARN_LOG("Skip %s demo: aclrtMalloc returned %d.", demoName, static_cast<int32_t>(ret));
61 return false;62 return false;
@@ -77,7 +78,7 @@ void DemonstrateStreamStop()
77 constexpr size_t kScratchSize = sizeof(uint32_t);78 constexpr size_t kScratchSize = sizeof(uint32_t);
78 constexpr uint32_t kBlockDim = 1;79 constexpr uint32_t kBlockDim = 1;
79 aclrtStream stopStream = nullptr;80 aclrtStream stopStream = nullptr;
80- uint32_t *stopScratch = nullptr;81+ uint32_t* stopScratch = nullptr;
81 82 
82 aclError ret = aclrtCreateStreamWithConfig(&stopStream, 0, ACL_STREAM_DEVICE_USE_ONLY);83 aclError ret = aclrtCreateStreamWithConfig(&stopStream, 0, ACL_STREAM_DEVICE_USE_ONLY);
83 if (ret != ACL_SUCCESS) {84 if (ret != ACL_SUCCESS) {
@@ -103,7 +104,7 @@ void DemonstrateStreamAbort()
103 constexpr size_t kScratchSize = sizeof(uint32_t);104 constexpr size_t kScratchSize = sizeof(uint32_t);
104 constexpr uint32_t kBlockDim = 1;105 constexpr uint32_t kBlockDim = 1;
105 aclrtStream abortStream = nullptr;106 aclrtStream abortStream = nullptr;
106- uint32_t *abortScratch = nullptr;107+ uint32_t* abortScratch = nullptr;
107 108 
108 aclError ret = aclrtCreateStream(&abortStream);109 aclError ret = aclrtCreateStream(&abortStream);
109 if (ret != ACL_SUCCESS) {110 if (ret != ACL_SUCCESS) {
@@ -138,7 +139,7 @@ int main()
138 const int32_t deviceId = 0;139 const int32_t deviceId = 0;
139 const uint32_t blockDim = 1;140 const uint32_t blockDim = 1;
140 uint32_t num = 0;141 uint32_t num = 0;
141- uint32_t *numDevice = nullptr;142+ uint32_t* numDevice = nullptr;
142 const size_t size = sizeof(uint32_t);143 const size_t size = sizeof(uint32_t);
143 aclrtStream stream = nullptr;144 aclrtStream stream = nullptr;
144 aclrtContext context = nullptr;145 aclrtContext context = nullptr;
@@ -150,7 +151,7 @@ int main()
150 CHECK_ERROR(aclrtSetDevice(deviceId));151 CHECK_ERROR(aclrtSetDevice(deviceId));
151 CHECK_ERROR(aclrtCreateContext(&context, deviceId));152 CHECK_ERROR(aclrtCreateContext(&context, deviceId));
152 CHECK_ERROR(aclrtCreateStream(&stream));153 CHECK_ERROR(aclrtCreateStream(&stream));
153- CHECK_ERROR(aclrtMalloc(reinterpret_cast<void **>(&numDevice), size, ACL_MEM_MALLOC_HUGE_FIRST));154+ CHECK_ERROR(aclrtMalloc(reinterpret_cast<void**>(&numDevice), size, ACL_MEM_MALLOC_HUGE_FIRST));
154 CHECK_ERROR(aclrtMemcpy(numDevice, size, &num, size, ACL_MEMCPY_HOST_TO_DEVICE));155 CHECK_ERROR(aclrtMemcpy(numDevice, size, &num, size, ACL_MEMCPY_HOST_TO_DEVICE));
155 156 
156 INFO_LOG("Assigning task without failure mode.");157 INFO_LOG("Assigning task without failure mode.");
Mexample/1_basic_features/stream/2_multi_stream/main.cpp+157-157
@@ -7,7 +7,7 @@
7 * INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE.7 * INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE.
8 * See LICENSE in the root of the software repository for the full text of the License.8 * See LICENSE in the root of the software repository for the full text of the License.
9 */9 */
10- 10+ 
11#include <iostream>11#include <iostream>
12#include "utils.h"12#include "utils.h"
13#include "acl/acl.h"13#include "acl/acl.h"
@@ -18,180 +18,180 @@ using namespace std;
18// 耗时较短的核函数，功能为x自乘218// 耗时较短的核函数，功能为x自乘2
19 19 
20namespace {20namespace {
21- constexpr int kWaitTimeMs = 1000;21+constexpr int kWaitTimeMs = 1000;
22- constexpr int32_t kDeviceId = 0;22+constexpr int32_t kDeviceId = 0;
23- constexpr uint32_t kBlockDim = 1;23+constexpr uint32_t kBlockDim = 1;
24 24 
25- // Event同步25+// Event同步
26- int EventSynchronization()26+int EventSynchronization()
27- {27+{
28- uint32_t num = 0;28+ uint32_t num = 0;
29- uint32_t *numDevice = nullptr;29+ uint32_t* numDevice = nullptr;
30- size_t size = sizeof(uint32_t);30+ size_t size = sizeof(uint32_t);
31- aclrtContext context = nullptr;31+ aclrtContext context = nullptr;
32- aclrtStream stream1 = nullptr;32+ aclrtStream stream1 = nullptr;
33- aclrtStream stream2 = nullptr;33+ aclrtStream stream2 = nullptr;
34- aclrtEvent event = nullptr;34+ aclrtEvent event = nullptr;
35- aclInit(nullptr);35+ aclInit(nullptr);
36 36 
37- // 申请设备37+ // 申请设备
38- CHECK_ERROR(aclrtSetDevice(kDeviceId));38+ CHECK_ERROR(aclrtSetDevice(kDeviceId));
39- CHECK_ERROR(aclrtCreateContext(&context, kDeviceId));39+ CHECK_ERROR(aclrtCreateContext(&context, kDeviceId));
40- CHECK_ERROR(aclrtCreateStream(&stream1));40+ CHECK_ERROR(aclrtCreateStream(&stream1));
41- CHECK_ERROR(aclrtCreateStream(&stream2));41+ CHECK_ERROR(aclrtCreateStream(&stream2));
42- // 默认为遇错继续执行，调用该函数设置为遇错即停，当一个任务执行遇到错误，后续任务将不进行。42+ // 默认为遇错继续执行，调用该函数设置为遇错即停，当一个任务执行遇到错误，后续任务将不进行。
43- CHECK_ERROR(aclrtSetStreamFailureMode(stream1, ACL_STOP_ON_FAILURE));43+ CHECK_ERROR(aclrtSetStreamFailureMode(stream1, ACL_STOP_ON_FAILURE));
44- CHECK_ERROR(aclrtSetStreamFailureMode(stream2, ACL_STOP_ON_FAILURE));44+ CHECK_ERROR(aclrtSetStreamFailureMode(stream2, ACL_STOP_ON_FAILURE));
45- CHECK_ERROR(aclrtCreateEvent(&event));45+ CHECK_ERROR(aclrtCreateEvent(&event));
46- INFO_LOG("Use event synchronize.");46+ INFO_LOG("Use event synchronize.");
47- // 开始做长任务47+ // 开始做长任务
48- CHECK_ERROR(aclrtMalloc((void**)&numDevice, size, ACL_MEM_MALLOC_HUGE_FIRST));48+ CHECK_ERROR(aclrtMalloc((void**)&numDevice, size, ACL_MEM_MALLOC_HUGE_FIRST));
49- CHECK_ERROR(aclrtMemcpy(numDevice, size, &num, size, ACL_MEMCPY_HOST_TO_DEVICE));49+ CHECK_ERROR(aclrtMemcpy(numDevice, size, &num, size, ACL_MEMCPY_HOST_TO_DEVICE));
50- INFO_LOG("Applied resource successfully, begin assigning task.");50+ INFO_LOG("Applied resource successfully, begin assigning task.");
51- INFO_LOG("Begin a long task, num += 1.");51+ INFO_LOG("Begin a long task, num += 1.");
52- LongOP(kBlockDim, stream1, numDevice);52+ LongOP(kBlockDim, stream1, numDevice);
53- CHECK_ERROR(aclrtRecordEvent(event, stream1));53+ CHECK_ERROR(aclrtRecordEvent(event, stream1));
54- INFO_LOG("Event synchronize.");54+ INFO_LOG("Event synchronize.");
55- CHECK_ERROR(aclrtStreamWaitEvent(stream2, event));55+ CHECK_ERROR(aclrtStreamWaitEvent(stream2, event));
56 56 
57- // 复位event，确保在同步后再复位，可以重复使用event，节省资源。57+ // 复位event，确保在同步后再复位，可以重复使用event，节省资源。
58- CHECK_ERROR(aclrtResetEvent(event, stream2));58+ CHECK_ERROR(aclrtResetEvent(event, stream2));
59- INFO_LOG("Begin a short task, num *= 2.");59+ INFO_LOG("Begin a short task, num *= 2.");
60- ShortOP(kBlockDim, stream2, numDevice);60+ ShortOP(kBlockDim, stream2, numDevice);
61- CHECK_ERROR(aclrtRecordEvent(event, stream2));61+ CHECK_ERROR(aclrtRecordEvent(event, stream2));
62- CHECK_ERROR(aclrtSynchronizeEvent(event));62+ CHECK_ERROR(aclrtSynchronizeEvent(event));
63- CHECK_ERROR(aclrtMemcpy(&num, size, numDevice, size, ACL_MEMCPY_DEVICE_TO_HOST));63+ CHECK_ERROR(aclrtMemcpy(&num, size, numDevice, size, ACL_MEMCPY_DEVICE_TO_HOST));
64- // 按照顺序执行结果应该为264+ // 按照顺序执行结果应该为2
65- INFO_LOG("The answer is %d.", num);65+ INFO_LOG("The answer is %d.", num);
66 66 
67- // event超时等待67+ // event超时等待
68- INFO_LOG("Simulate timeout, wait 1000ms, the task will fail.");68+ INFO_LOG("Simulate timeout, wait 1000ms, the task will fail.");
69- LongOP(kBlockDim, stream1, numDevice);69+ LongOP(kBlockDim, stream1, numDevice);
70- CHECK_ERROR(aclrtRecordEvent(event, stream1));70+ CHECK_ERROR(aclrtRecordEvent(event, stream1));
71- // 流的超时等待，如果超时直接报函数错误，将时间变长则不会报错。71+ // 流的超时等待，如果超时直接报函数错误，将时间变长则不会报错。
72- CHECK_ERROR_WITHOUT_RETURN(aclrtSynchronizeEventWithTimeout(event, kWaitTimeMs));72+ CHECK_ERROR_WITHOUT_RETURN(aclrtSynchronizeEventWithTimeout(event, kWaitTimeMs));
73 73 
74- // 释放资源74+ // 释放资源
75- CHECK_ERROR(aclrtFree(numDevice));75+ CHECK_ERROR(aclrtFree(numDevice));
76- CHECK_ERROR(aclrtDestroyEvent(event));76+ CHECK_ERROR(aclrtDestroyEvent(event));
77- CHECK_ERROR(aclrtDestroyStreamForce(stream1));77+ CHECK_ERROR(aclrtDestroyStreamForce(stream1));
78- CHECK_ERROR(aclrtDestroyStreamForce(stream2));78+ CHECK_ERROR(aclrtDestroyStreamForce(stream2));
79- CHECK_ERROR(aclrtDestroyContext(context));79+ CHECK_ERROR(aclrtDestroyContext(context));
80- CHECK_ERROR(aclrtResetDeviceForce(kDeviceId));80+ CHECK_ERROR(aclrtResetDeviceForce(kDeviceId));
81- aclFinalize();81+ aclFinalize();
82- INFO_LOG("Resource cleanup completed.");82+ INFO_LOG("Resource cleanup completed.");
83- return 0;83+ return 0;
84- }84+}
85 85 
86- // Stream同步86+// Stream同步
87- int StreamSynchronization()87+int StreamSynchronization()
88- {88+{
89- uint32_t num = 0;89+ uint32_t num = 0;
90- uint32_t *numDevice = nullptr;90+ uint32_t* numDevice = nullptr;
91- size_t size = sizeof(uint32_t);91+ size_t size = sizeof(uint32_t);
92- aclrtContext context = nullptr;92+ aclrtContext context = nullptr;
93- aclrtStream stream1 = nullptr;93+ aclrtStream stream1 = nullptr;
94- aclrtStream stream2 = nullptr;94+ aclrtStream stream2 = nullptr;
95- aclInit(nullptr);95+ aclInit(nullptr);
96 96 
97- // 申请设备97+ // 申请设备
98- CHECK_ERROR(aclrtSetDevice(kDeviceId));98+ CHECK_ERROR(aclrtSetDevice(kDeviceId));
99- CHECK_ERROR(aclrtCreateContext(&context, kDeviceId));99+ CHECK_ERROR(aclrtCreateContext(&context, kDeviceId));
100- CHECK_ERROR(aclrtCreateStream(&stream1));100+ CHECK_ERROR(aclrtCreateStream(&stream1));
101- CHECK_ERROR(aclrtCreateStream(&stream2));101+ CHECK_ERROR(aclrtCreateStream(&stream2));
102- // 默认为遇错继续执行，调用该函数设置为遇错即停，当一个任务执行遇到错误，后续任务将不进行。102+ // 默认为遇错继续执行，调用该函数设置为遇错即停，当一个任务执行遇到错误，后续任务将不进行。
103- CHECK_ERROR(aclrtSetStreamFailureMode(stream1, ACL_STOP_ON_FAILURE));103+ CHECK_ERROR(aclrtSetStreamFailureMode(stream1, ACL_STOP_ON_FAILURE));
104- CHECK_ERROR(aclrtSetStreamFailureMode(stream2, ACL_STOP_ON_FAILURE));104+ CHECK_ERROR(aclrtSetStreamFailureMode(stream2, ACL_STOP_ON_FAILURE));
105- INFO_LOG("Use stream synchronize.");105+ INFO_LOG("Use stream synchronize.");
106- // 开始做长任务106+ // 开始做长任务
107- CHECK_ERROR(aclrtMalloc((void**)&numDevice, size, ACL_MEM_MALLOC_HUGE_FIRST));107+ CHECK_ERROR(aclrtMalloc((void**)&numDevice, size, ACL_MEM_MALLOC_HUGE_FIRST));
108- CHECK_ERROR(aclrtMemcpy(numDevice, size, &num, size, ACL_MEMCPY_HOST_TO_DEVICE));108+ CHECK_ERROR(aclrtMemcpy(numDevice, size, &num, size, ACL_MEMCPY_HOST_TO_DEVICE));
109- INFO_LOG("Applied resource successfully, begin assigning task.");109+ INFO_LOG("Applied resource successfully, begin assigning task.");
110- INFO_LOG("Begin a long task, num += 1.");110+ INFO_LOG("Begin a long task, num += 1.");
111- LongOP(kBlockDim, stream1, numDevice);111+ LongOP(kBlockDim, stream1, numDevice);
112- INFO_LOG("Stream synchronize.");112+ INFO_LOG("Stream synchronize.");
113- CHECK_ERROR(aclrtSynchronizeStream(stream1));113+ CHECK_ERROR(aclrtSynchronizeStream(stream1));
114 114 
115- // 复位event，确保在同步后再复位，可以重复使用event，节省资源。115+ // 复位event，确保在同步后再复位，可以重复使用event，节省资源。
116- INFO_LOG("Begin a short task, num *= 2.");116+ INFO_LOG("Begin a short task, num *= 2.");
117- ShortOP(kBlockDim, stream2, numDevice);117+ ShortOP(kBlockDim, stream2, numDevice);
118- CHECK_ERROR(aclrtSynchronizeStream(stream2));118+ CHECK_ERROR(aclrtSynchronizeStream(stream2));
119- CHECK_ERROR(aclrtMemcpy(&num, size, numDevice, size, ACL_MEMCPY_DEVICE_TO_HOST));119+ CHECK_ERROR(aclrtMemcpy(&num, size, numDevice, size, ACL_MEMCPY_DEVICE_TO_HOST));
120- // 按照顺序执行结果应该为2120+ // 按照顺序执行结果应该为2
121- INFO_LOG("The answer is %d.", num);121+ INFO_LOG("The answer is %d.", num);
122 122 
123- // event超时等待123+ // event超时等待
124- INFO_LOG("Simulate timeout, wait 1000ms, the task will fail.");124+ INFO_LOG("Simulate timeout, wait 1000ms, the task will fail.");
125- LongOP(kBlockDim, stream1, numDevice);125+ LongOP(kBlockDim, stream1, numDevice);
126- // 流的超时等待，如果超时直接报函数错误，将时间变长则不会报错。126+ // 流的超时等待，如果超时直接报函数错误，将时间变长则不会报错。
127- CHECK_ERROR_WITHOUT_RETURN(aclrtSynchronizeStreamWithTimeout(stream1, kWaitTimeMs));127+ CHECK_ERROR_WITHOUT_RETURN(aclrtSynchronizeStreamWithTimeout(stream1, kWaitTimeMs));
128 128 
129- // 释放资源129+ // 释放资源
130- CHECK_ERROR(aclrtFree(numDevice));130+ CHECK_ERROR(aclrtFree(numDevice));
131- CHECK_ERROR(aclrtDestroyStreamForce(stream1));131+ CHECK_ERROR(aclrtDestroyStreamForce(stream1));
132- CHECK_ERROR(aclrtDestroyStreamForce(stream2));132+ CHECK_ERROR(aclrtDestroyStreamForce(stream2));
133- CHECK_ERROR(aclrtDestroyContext(context));133+ CHECK_ERROR(aclrtDestroyContext(context));
134- CHECK_ERROR(aclrtResetDeviceForce(kDeviceId));134+ CHECK_ERROR(aclrtResetDeviceForce(kDeviceId));
135- aclFinalize();135+ aclFinalize();
136- INFO_LOG("Resource cleanup completed.");136+ INFO_LOG("Resource cleanup completed.");
137- return 0;137+ return 0;
138- }138+}
139 139 
140- // Notify同步140+// Notify同步
141- int NotifySynchronization()141+int NotifySynchronization()
142- {142+{
143- uint32_t num = 0;143+ uint32_t num = 0;
144- uint32_t *numDevice = nullptr;144+ uint32_t* numDevice = nullptr;
145- size_t size = sizeof(uint32_t);145+ size_t size = sizeof(uint32_t);
146- aclrtContext context = nullptr;146+ aclrtContext context = nullptr;
147- aclrtNotify notify;147+ aclrtNotify notify;
148- aclrtStream stream1 = nullptr;148+ aclrtStream stream1 = nullptr;
149- aclrtStream stream2 = nullptr;149+ aclrtStream stream2 = nullptr;
150- aclInit(nullptr);150+ aclInit(nullptr);
151 151 
152- // 申请设备152+ // 申请设备
153- CHECK_ERROR(aclrtSetDevice(kDeviceId));153+ CHECK_ERROR(aclrtSetDevice(kDeviceId));
154- CHECK_ERROR(aclrtCreateContext(&context, kDeviceId));154+ CHECK_ERROR(aclrtCreateContext(&context, kDeviceId));
155- CHECK_ERROR(aclrtCreateStream(&stream1));155+ CHECK_ERROR(aclrtCreateStream(&stream1));
156- CHECK_ERROR(aclrtCreateStream(&stream2));156+ CHECK_ERROR(aclrtCreateStream(&stream2));
157- CHECK_ERROR(aclrtCreateNotify(&notify, 0x00U));157+ CHECK_ERROR(aclrtCreateNotify(&notify, 0x00U));
158- // 默认为遇错继续执行，调用该函数设置为遇错即停，当一个任务执行遇到错误，后续任务将不进行。158+ // 默认为遇错继续执行，调用该函数设置为遇错即停，当一个任务执行遇到错误，后续任务将不进行。
159- CHECK_ERROR(aclrtSetStreamFailureMode(stream1, ACL_STOP_ON_FAILURE));159+ CHECK_ERROR(aclrtSetStreamFailureMode(stream1, ACL_STOP_ON_FAILURE));
160- CHECK_ERROR(aclrtSetStreamFailureMode(stream2, ACL_STOP_ON_FAILURE));160+ CHECK_ERROR(aclrtSetStreamFailureMode(stream2, ACL_STOP_ON_FAILURE));
161- INFO_LOG("Use notify synchronize.");161+ INFO_LOG("Use notify synchronize.");
162- // 开始做长任务162+ // 开始做长任务
163- CHECK_ERROR(aclrtMalloc((void**)&numDevice, size, ACL_MEM_MALLOC_HUGE_FIRST));163+ CHECK_ERROR(aclrtMalloc((void**)&numDevice, size, ACL_MEM_MALLOC_HUGE_FIRST));
164- CHECK_ERROR(aclrtMemcpy(numDevice, size, &num, size, ACL_MEMCPY_HOST_TO_DEVICE));164+ CHECK_ERROR(aclrtMemcpy(numDevice, size, &num, size, ACL_MEMCPY_HOST_TO_DEVICE));
165- INFO_LOG("Applied resource successfully, begin assigning task.");165+ INFO_LOG("Applied resource successfully, begin assigning task.");
166- INFO_LOG("Begin a long task, num += 1.");166+ INFO_LOG("Begin a long task, num += 1.");
167- LongOP(kBlockDim, stream1, numDevice);167+ LongOP(kBlockDim, stream1, numDevice);
168- INFO_LOG("Stream synchronize.");168+ INFO_LOG("Stream synchronize.");
169- CHECK_ERROR(aclrtRecordNotify(notify, stream1));169+ CHECK_ERROR(aclrtRecordNotify(notify, stream1));
170 170 
171- CHECK_ERROR(aclrtWaitAndResetNotify(notify, stream2, 0));171+ CHECK_ERROR(aclrtWaitAndResetNotify(notify, stream2, 0));
172- INFO_LOG("Begin a short task, num *= 2.");172+ INFO_LOG("Begin a short task, num *= 2.");
173- ShortOP(kBlockDim, stream2, numDevice);173+ ShortOP(kBlockDim, stream2, numDevice);
174- CHECK_ERROR(aclrtSynchronizeStream(stream2));174+ CHECK_ERROR(aclrtSynchronizeStream(stream2));
175- CHECK_ERROR(aclrtMemcpy(&num, size, numDevice, size, ACL_MEMCPY_DEVICE_TO_HOST));175+ CHECK_ERROR(aclrtMemcpy(&num, size, numDevice, size, ACL_MEMCPY_DEVICE_TO_HOST));
176- // 按照顺序执行结果应该为2176+ // 按照顺序执行结果应该为2
177- INFO_LOG("The answer is %d.", num);177+ INFO_LOG("The answer is %d.", num);
178- // 释放资源178+ // 释放资源
179- CHECK_ERROR(aclrtFree(numDevice));179+ CHECK_ERROR(aclrtFree(numDevice));
180- CHECK_ERROR(aclrtDestroyNotify(notify));180+ CHECK_ERROR(aclrtDestroyNotify(notify));
181- CHECK_ERROR(aclrtDestroyStreamForce(stream1));181+ CHECK_ERROR(aclrtDestroyStreamForce(stream1));
182- CHECK_ERROR(aclrtDestroyStreamForce(stream2));182+ CHECK_ERROR(aclrtDestroyStreamForce(stream2));
183- CHECK_ERROR(aclrtDestroyContext(context));183+ CHECK_ERROR(aclrtDestroyContext(context));
184- CHECK_ERROR(aclrtResetDeviceForce(kDeviceId));184+ CHECK_ERROR(aclrtResetDeviceForce(kDeviceId));
185- aclFinalize();185+ aclFinalize();
186- INFO_LOG("Resource cleanup completed.");186+ INFO_LOG("Resource cleanup completed.");
187- return 0;187+ return 0;
188- }188+}
189} // namespace189} // namespace
190 190 
191int main()191int main()
192{192{
193 // 每个函数代表一种同步的方法193 // 每个函数代表一种同步的方法
194- 194+ 
195 (void)StreamSynchronization();195 (void)StreamSynchronization();
196 (void)EventSynchronization();196 (void)EventSynchronization();
197 (void)NotifySynchronization();197 (void)NotifySynchronization();
Mexample/1_basic_features/stream/3_stream_config_query/main.cpp+25-25
@@ -21,7 +21,7 @@ constexpr int32_t kDeviceId = 0;
21constexpr uint32_t kStreamFlags = 0;21constexpr uint32_t kStreamFlags = 0;
22constexpr uint32_t kStreamPriority = 0;22constexpr uint32_t kStreamPriority = 0;
23 23 
24-int CheckAcl(aclError ret, const char *expr)24+int CheckAcl(aclError ret, const char* expr)
25{25{
26 if (ret != ACL_SUCCESS) {26 if (ret != ACL_SUCCESS) {
27 fprintf(stderr, "[ERROR] Operation failed: %s returned error code %d\n", expr, static_cast<int32_t>(ret));27 fprintf(stderr, "[ERROR] Operation failed: %s returned error code %d\n", expr, static_cast<int32_t>(ret));
@@ -31,10 +31,10 @@ int CheckAcl(aclError ret, const char *expr)
31 return 0;31 return 0;
32}32}
33 33 
34-using CreateStreamConfigHandleFunc = aclrtStreamConfigHandle *(*)();34+using CreateStreamConfigHandleFunc = aclrtStreamConfigHandle* (*)();
35-using DestroyStreamConfigHandleFunc = aclError (*)(aclrtStreamConfigHandle *);35+using DestroyStreamConfigHandleFunc = aclError (*)(aclrtStreamConfigHandle*);
36-using SetStreamConfigOptFunc = aclError (*)(aclrtStreamConfigHandle *, aclrtStreamConfigAttr, const void *, size_t);36+using SetStreamConfigOptFunc = aclError (*)(aclrtStreamConfigHandle*, aclrtStreamConfigAttr, const void*, size_t);
37-using CreateStreamV2Func = aclError (*)(aclrtStream *, const aclrtStreamConfigHandle *);37+using CreateStreamV2Func = aclError (*)(aclrtStream*, const aclrtStreamConfigHandle*);
38 38 
39struct StreamConfigApi {39struct StreamConfigApi {
40 CreateStreamConfigHandleFunc createConfigHandle;40 CreateStreamConfigHandleFunc createConfigHandle;
@@ -45,10 +45,7 @@ struct StreamConfigApi {
45 45 
46class StreamConfigHandleGuard {46class StreamConfigHandleGuard {
47public:47public:
48- StreamConfigHandleGuard(const StreamConfigApi &api, aclrtStreamConfigHandle *handle)48+ StreamConfigHandleGuard(const StreamConfigApi& api, aclrtStreamConfigHandle* handle) : api_(api), handle_(handle) {}
49- : api_(api), handle_(handle)
50- {
51- }
52 49 
53 ~StreamConfigHandleGuard()50 ~StreamConfigHandleGuard()
54 {51 {
@@ -57,8 +54,8 @@ public:
57 }54 }
58 }55 }
59 56 
60- StreamConfigHandleGuard(const StreamConfigHandleGuard &) = delete;57+ StreamConfigHandleGuard(const StreamConfigHandleGuard&) = delete;
61- StreamConfigHandleGuard &operator=(const StreamConfigHandleGuard &) = delete;58+ StreamConfigHandleGuard& operator=(const StreamConfigHandleGuard&) = delete;
62 59 
63 aclError Destroy()60 aclError Destroy()
64 {61 {
@@ -74,8 +71,8 @@ public:
74 }71 }
75 72 
76private:73private:
77- const StreamConfigApi &api_;74+ const StreamConfigApi& api_;
78- aclrtStreamConfigHandle *handle_;75+ aclrtStreamConfigHandle* handle_;
79};76};
80 77 
81class StreamGuard {78class StreamGuard {
@@ -89,8 +86,8 @@ public:
89 }86 }
90 }87 }
91 88 
92- StreamGuard(const StreamGuard &) = delete;89+ StreamGuard(const StreamGuard&) = delete;
93- StreamGuard &operator=(const StreamGuard &) = delete;90+ StreamGuard& operator=(const StreamGuard&) = delete;
94 91 
95 aclError Destroy()92 aclError Destroy()
96 {93 {
@@ -110,7 +107,7 @@ private:
110};107};
111 108 
112template <typename Func>109template <typename Func>
113-bool LoadSymbol(void *handle, const char *name, Func *func)110+bool LoadSymbol(void* handle, const char* name, Func* func)
114{111{
115 *func = reinterpret_cast<Func>(dlsym(handle, name));112 *func = reinterpret_cast<Func>(dlsym(handle, name));
116 if (*func == nullptr) {113 if (*func == nullptr) {
@@ -120,7 +117,7 @@ bool LoadSymbol(void *handle, const char *name, Func *func)
120 return true;117 return true;
121}118}
122 119 
123-bool LoadStreamConfigApi(StreamConfigApi *api)120+bool LoadStreamConfigApi(StreamConfigApi* api)
124{121{
125 bool ok = true;122 bool ok = true;
126 ok = LoadSymbol(RTLD_DEFAULT, "aclrtCreateStreamConfigHandle", &api->createConfigHandle) && ok;123 ok = LoadSymbol(RTLD_DEFAULT, "aclrtCreateStreamConfigHandle", &api->createConfigHandle) && ok;
@@ -130,14 +127,16 @@ bool LoadStreamConfigApi(StreamConfigApi *api)
130 return ok;127 return ok;
131}128}
132 129 
133-int SetStreamConfigOptions(const StreamConfigApi &api, aclrtStreamConfigHandle *configHandle)130+int SetStreamConfigOptions(const StreamConfigApi& api, aclrtStreamConfigHandle* configHandle)
134{131{
135- if (CheckAcl(api.setConfigOpt(configHandle, ACL_RT_STREAM_FLAG, &kStreamFlags, sizeof(kStreamFlags)),132+ if (CheckAcl(
136- "aclrtSetStreamConfigOpt(ACL_RT_STREAM_FLAG)") != 0) {133+ api.setConfigOpt(configHandle, ACL_RT_STREAM_FLAG, &kStreamFlags, sizeof(kStreamFlags)),
134+ "aclrtSetStreamConfigOpt(ACL_RT_STREAM_FLAG)") != 0) {
137 return -1;135 return -1;
138 }136 }
139- if (CheckAcl(api.setConfigOpt(configHandle, ACL_RT_STREAM_PRIORITY, &kStreamPriority, sizeof(kStreamPriority)),137+ if (CheckAcl(
140- "aclrtSetStreamConfigOpt(ACL_RT_STREAM_PRIORITY)") != 0) {138+ api.setConfigOpt(configHandle, ACL_RT_STREAM_PRIORITY, &kStreamPriority, sizeof(kStreamPriority)),
139+ "aclrtSetStreamConfigOpt(ACL_RT_STREAM_PRIORITY)") != 0) {
141 return -1;140 return -1;
142 }141 }
143 INFO_LOG("Set stream config flags=%u priority=%u", kStreamFlags, kStreamPriority);142 INFO_LOG("Set stream config flags=%u priority=%u", kStreamFlags, kStreamPriority);
@@ -170,11 +169,12 @@ int RunStreamConfigQuerySample()
170{169{
171 StreamConfigApi api = {};170 StreamConfigApi api = {};
172 if (!LoadStreamConfigApi(&api)) {171 if (!LoadStreamConfigApi(&api)) {
173- INFO_LOG("[SKIP] Stream config query sample skipped because the current Runtime library does not export all stream config APIs.");172+ INFO_LOG("[SKIP] Stream config query sample skipped because the current Runtime library does not export all "
173+ "stream config APIs.");
174 return 0;174 return 0;
175 }175 }
176 176 
177- aclrtStreamConfigHandle *configHandle = api.createConfigHandle();177+ aclrtStreamConfigHandle* configHandle = api.createConfigHandle();
178 if (configHandle == nullptr) {178 if (configHandle == nullptr) {
179 INFO_LOG("[SKIP] Stream config query sample skipped because aclrtCreateStreamConfigHandle returned nullptr.");179 INFO_LOG("[SKIP] Stream config query sample skipped because aclrtCreateStreamConfigHandle returned nullptr.");
180 return 0;180 return 0;
@@ -209,7 +209,7 @@ int RunStreamConfigQuerySample()
209 INFO_LOG("[SUCCESS] Stream config query sample completed successfully");209 INFO_LOG("[SUCCESS] Stream config query sample completed successfully");
210 return 0;210 return 0;
211}211}
212-} // namespace212+} // namespace
213 213 
214int32_t main()214int32_t main()
215{215{
Mexample/2_advanced_features/built_in_task/1_random_num_task/main.cpp+1-1
@@ -75,7 +75,7 @@ int main()
75 75 
76 // 申请存放随机数状态 counter 的device内存 （要求 16Byte)76 // 申请存放随机数状态 counter 的device内存 （要求 16Byte)
77 void* counterAddr = NULL;77 void* counterAddr = NULL;
78- CHECK_ERROR(aclrtMalloc((void**)&counterAddr, 16, ACL_MEM_MALLOC_HUGE_FIRST)); 78+ CHECK_ERROR(aclrtMalloc((void**)&counterAddr, 16, ACL_MEM_MALLOC_HUGE_FIRST));
79 79 
80 printf("Gen normal distribution random num, data type: float \n");80 printf("Gen normal distribution random num, data type: float \n");
81 float mean = 3.0;81 float mean = 3.0;
Mexample/2_advanced_features/callback/0_simple_callback/callback_sample.cpp+12-15
@@ -23,10 +23,7 @@ int32_t CallBackSpace::CallBackSample::deviceId_ = 0;
23 23 
24CallBackSpace::CallBackSample::CallBackSample() = default;24CallBackSpace::CallBackSample::CallBackSample() = default;
25 25 
26-CallBackSpace::CallBackSample::~CallBackSample()26+CallBackSpace::CallBackSample::~CallBackSample() { (void)Destroy(); }
27-{
28- (void)Destroy();
29-}
30 27 
31int CallBackSpace::CallBackSample::Init()28int CallBackSpace::CallBackSample::Init()
32{29{
@@ -38,7 +35,7 @@ int CallBackSpace::CallBackSample::Init()
38 return 0;35 return 0;
39}36}
40 37 
41-void CallBackSpace::CallBackSample::ThreadFunc(void *arg)38+void CallBackSpace::CallBackSample::ThreadFunc(void* arg)
42{39{
43 const int waitTime = 100;40 const int waitTime = 100;
44 aclError ret = aclrtSetCurrentContext(context_);41 aclError ret = aclrtSetCurrentContext(context_);
@@ -52,7 +49,7 @@ void CallBackSpace::CallBackSample::ThreadFunc(void *arg)
52 INFO_LOG("Report callback thread exit");49 INFO_LOG("Report callback thread exit");
53}50}
54 51 
55-void CallBackSpace::CallBackSample::HostFuncThreadFunc(void *arg)52+void CallBackSpace::CallBackSample::HostFuncThreadFunc(void* arg)
56{53{
57 const int waitTime = 100;54 const int waitTime = 100;
58 aclError ret = aclrtSetCurrentContext(context_);55 aclError ret = aclrtSetCurrentContext(context_);
@@ -66,30 +63,30 @@ void CallBackSpace::CallBackSample::HostFuncThreadFunc(void *arg)
66 INFO_LOG("Hostfunc processing thread exit");63 INFO_LOG("Hostfunc processing thread exit");
67}64}
68 65 
69-void CallBackSpace::CallBackSample::CallBackBeforeLaunchFunc(void *arg)66+void CallBackSpace::CallBackSample::CallBackBeforeLaunchFunc(void* arg)
70{67{
71 thread::id tid = std::this_thread::get_id();68 thread::id tid = std::this_thread::get_id();
72 string tidStr = CallbackUtils::GetThreadId(tid);69 string tidStr = CallbackUtils::GetThreadId(tid);
73 INFO_LOG("Report callback thread id is %s", tidStr.c_str());70 INFO_LOG("Report callback thread id is %s", tidStr.c_str());
74- int *data = static_cast<int*>(arg);71+ int* data = static_cast<int*>(arg);
75 INFO_LOG("This callback before task, result: user data is: %d.", *data);72 INFO_LOG("This callback before task, result: user data is: %d.", *data);
76}73}
77 74 
78-void CallBackSpace::CallBackSample::CallBackFunc(void *arg)75+void CallBackSpace::CallBackSample::CallBackFunc(void* arg)
79{76{
80 thread::id tid = std::this_thread::get_id();77 thread::id tid = std::this_thread::get_id();
81 string tidStr = CallbackUtils::GetThreadId(tid);78 string tidStr = CallbackUtils::GetThreadId(tid);
82 INFO_LOG("Report callback thread id is %s", tidStr.c_str());79 INFO_LOG("Report callback thread id is %s", tidStr.c_str());
83- int *data = static_cast<int*>(arg);80+ int* data = static_cast<int*>(arg);
84 INFO_LOG("This callback after task and loop five times, result: user data is: %d.", *data);81 INFO_LOG("This callback after task and loop five times, result: user data is: %d.", *data);
85}82}
86 83 
87-void CallBackSpace::CallBackSample::HostFunc(void *arg)84+void CallBackSpace::CallBackSample::HostFunc(void* arg)
88{85{
89 thread::id tid = std::this_thread::get_id();86 thread::id tid = std::this_thread::get_id();
90 string tidStr = CallbackUtils::GetThreadId(tid);87 string tidStr = CallbackUtils::GetThreadId(tid);
91 INFO_LOG("Hostfunc callback thread id is %s", tidStr.c_str());88 INFO_LOG("Hostfunc callback thread id is %s", tidStr.c_str());
92- int *data = static_cast<int*>(arg);89+ int* data = static_cast<int*>(arg);
93 INFO_LOG("Hostfunc executed in subscribed thread, user data is: %d.", *data);90 INFO_LOG("Hostfunc executed in subscribed thread, user data is: %d.", *data);
94}91}
95 92 
@@ -100,9 +97,9 @@ int CallBackSpace::CallBackSample::Callback()
100 bool isReportLoop = true;97 bool isReportLoop = true;
101 bool isHostFuncLoop = true;98 bool isHostFuncLoop = true;
102 const size_t size = sizeof(uint32_t);99 const size_t size = sizeof(uint32_t);
103- uint32_t *numDevice = nullptr;100+ uint32_t* numDevice = nullptr;
104 const int count = 5;101 const int count = 5;
105- CHECK_ERROR(aclrtMalloc((void **)&numDevice, size, ACL_MEM_MALLOC_HUGE_FIRST));102+ CHECK_ERROR(aclrtMalloc((void**)&numDevice, size, ACL_MEM_MALLOC_HUGE_FIRST));
106 CHECK_ERROR(aclrtMemcpy(numDevice, size, &num, size, ACL_MEMCPY_HOST_TO_DEVICE));103 CHECK_ERROR(aclrtMemcpy(numDevice, size, &num, size, ACL_MEMCPY_HOST_TO_DEVICE));
107 104 
108 thread reportThread(ThreadFunc, &isReportLoop);105 thread reportThread(ThreadFunc, &isReportLoop);
@@ -115,7 +112,7 @@ int CallBackSpace::CallBackSample::Callback()
115 INFO_LOG("The created report thread id is %s", reportThreadId.c_str());112 INFO_LOG("The created report thread id is %s", reportThreadId.c_str());
116 INFO_LOG("The created hostfunc thread id is %s", hostFuncThreadId.c_str());113 INFO_LOG("The created hostfunc thread id is %s", hostFuncThreadId.c_str());
117 114 
118- int *userData = new int(520);115+ int* userData = new int(520);
119 uint64_t reportTidInt = std::stoull(reportThreadId);116 uint64_t reportTidInt = std::stoull(reportThreadId);
120 uint64_t hostFuncTidInt = std::stoull(hostFuncThreadId);117 uint64_t hostFuncTidInt = std::stoull(hostFuncThreadId);
121 CHECK_ERROR(aclrtSubscribeReport(reportTidInt, stream_));118 CHECK_ERROR(aclrtSubscribeReport(reportTidInt, stream_));
Mexample/2_advanced_features/callback/0_simple_callback/callback_sample.h+18-18
@@ -17,24 +17,24 @@
17#include "acl/acl.h"17#include "acl/acl.h"
18 18 
19namespace CallBackSpace {19namespace CallBackSpace {
20- class CallBackSample {20+class CallBackSample {
21- public:21+public:
22- CallBackSample();22+ CallBackSample();
23- virtual ~CallBackSample();23+ virtual ~CallBackSample();
24- int Init();24+ int Init();
25- int Callback();25+ int Callback();
26- int Destroy();26+ int Destroy();
27- static void ThreadFunc(void *arg);27+ static void ThreadFunc(void* arg);
28- static void HostFuncThreadFunc(void *arg);28+ static void HostFuncThreadFunc(void* arg);
29- static void CallBackFunc(void *arg);29+ static void CallBackFunc(void* arg);
30- static void CallBackBeforeLaunchFunc(void *arg);30+ static void CallBackBeforeLaunchFunc(void* arg);
31- static void HostFunc(void *arg);31+ static void HostFunc(void* arg);
32 32 
33- public:33+public:
34- static int32_t deviceId_;34+ static int32_t deviceId_;
35- static aclrtContext context_;35+ static aclrtContext context_;
36- static aclrtStream stream_;36+ static aclrtStream stream_;
37- };37+};
38-}38+} // namespace CallBackSpace
39 39 
40#endif40#endif
Mexample/2_advanced_features/callback/1_callback_hostfunc/main.cpp+4-7
@@ -16,17 +16,14 @@ using namespace std;
16 16 
17// 核函数，让x自增117// 核函数，让x自增1
18namespace {18namespace {
19- void CallBackFunc(void *arg)19+void CallBackFunc(void* arg) { INFO_LOG("Hostfunc callback!!!"); }
20- {20+} // namespace
21- INFO_LOG("Hostfunc callback!!!");
22- }
23-}
24int main()21int main()
25{22{
26 // 初始化23 // 初始化
27 int32_t deviceId = 0;24 int32_t deviceId = 0;
28 uint32_t num = 0;25 uint32_t num = 0;
29- uint32_t *numDevice = nullptr;26+ uint32_t* numDevice = nullptr;
30 size_t size = sizeof(uint32_t);27 size_t size = sizeof(uint32_t);
31 uint32_t blockDim = 1;28 uint32_t blockDim = 1;
32 aclrtStream stream = nullptr;29 aclrtStream stream = nullptr;
@@ -35,7 +32,7 @@ int main()
35 CHECK_ERROR(aclrtSetDevice(deviceId));32 CHECK_ERROR(aclrtSetDevice(deviceId));
36 CHECK_ERROR(aclrtCreateContext(&context, deviceId));33 CHECK_ERROR(aclrtCreateContext(&context, deviceId));
37 // 默认流下发任务,默认流随着context创建而创建，给函数传入空即使用默认流34 // 默认流下发任务,默认流随着context创建而创建，给函数传入空即使用默认流
38- CHECK_ERROR(aclrtMalloc((void **)&numDevice, size, ACL_MEM_MALLOC_HUGE_FIRST));35+ CHECK_ERROR(aclrtMalloc((void**)&numDevice, size, ACL_MEM_MALLOC_HUGE_FIRST));
39 CHECK_ERROR(aclrtMemcpy(numDevice, size, &num, size, ACL_MEMCPY_HOST_TO_DEVICE));36 CHECK_ERROR(aclrtMemcpy(numDevice, size, &num, size, ACL_MEMCPY_HOST_TO_DEVICE));
40 CHECK_ERROR(aclrtCreateStream(&stream));37 CHECK_ERROR(aclrtCreateStream(&stream));
41 // 默认为遇错继续执行，调用该函数设置为遇错即停，当一个任务执行遇到错误，后续任务将不进行38 // 默认为遇错继续执行，调用该函数设置为遇错即停，当一个任务执行遇到错误，后续任务将不进行
Mexample/2_advanced_features/callback/2_callback_exception/exception_callback_sample.cpp+22-31
@@ -23,20 +23,16 @@ aclrtStream ExceptionCallBackSpace::ExceptionCallBackSample::stream_ = nullptr;
23int32_t ExceptionCallBackSpace::ExceptionCallBackSample::deviceId_ = 0;23int32_t ExceptionCallBackSpace::ExceptionCallBackSample::deviceId_ = 0;
24 24 
25namespace {25namespace {
26-const char *SafeString(const char *message)26+const char* SafeString(const char* message) { return message != nullptr ? message : "<null>"; }
27-{
28- return message != nullptr ? message : "<null>";
29-}
30 27 
31-void LogRuntimeErrorState(int32_t deviceId, const char *stage)28+void LogRuntimeErrorState(int32_t deviceId, const char* stage)
32{29{
33 aclrtErrorInfo errorInfo = {};30 aclrtErrorInfo errorInfo = {};
34 aclError verboseRet = aclrtGetErrorVerbose(deviceId, &errorInfo);31 aclError verboseRet = aclrtGetErrorVerbose(deviceId, &errorInfo);
35 if (verboseRet == ACL_SUCCESS) {32 if (verboseRet == ACL_SUCCESS) {
36- INFO_LOG("%s verbose error info: errorType=%d, tryRepair=%u, hasDetail=%u",33+ INFO_LOG(
37- stage,34+ "%s verbose error info: errorType=%d, tryRepair=%u, hasDetail=%u", stage,
38- static_cast<int32_t>(errorInfo.errorType),35+ static_cast<int32_t>(errorInfo.errorType), static_cast<uint32_t>(errorInfo.tryRepair),
39- static_cast<uint32_t>(errorInfo.tryRepair),
40 static_cast<uint32_t>(errorInfo.hasDetail));36 static_cast<uint32_t>(errorInfo.hasDetail));
41 } else {37 } else {
42 WARN_LOG("%s aclrtGetErrorVerbose failed with error code %d", stage, static_cast<int32_t>(verboseRet));38 WARN_LOG("%s aclrtGetErrorVerbose failed with error code %d", stage, static_cast<int32_t>(verboseRet));
@@ -44,21 +40,16 @@ void LogRuntimeErrorState(int32_t deviceId, const char *stage)
44 40 
45 aclError peekError = aclrtPeekAtLastError(ACL_RT_THREAD_LEVEL);41 aclError peekError = aclrtPeekAtLastError(ACL_RT_THREAD_LEVEL);
46 aclError lastError = aclrtGetLastError(ACL_RT_THREAD_LEVEL);42 aclError lastError = aclrtGetLastError(ACL_RT_THREAD_LEVEL);
47- const char *recentErrMsg = aclGetRecentErrMsg();43+ const char* recentErrMsg = aclGetRecentErrMsg();
48- ERROR_LOG("%s runtime diagnostics: peekErr=%d, lastErr=%d, recentErrMsg=%s",44+ ERROR_LOG(
49- stage,45+ "%s runtime diagnostics: peekErr=%d, lastErr=%d, recentErrMsg=%s", stage, static_cast<int32_t>(peekError),
50- static_cast<int32_t>(peekError),46+ static_cast<int32_t>(lastError), SafeString(recentErrMsg));
51- static_cast<int32_t>(lastError),
52- SafeString(recentErrMsg));
53}47}
54} // namespace48} // namespace
55 49 
56ExceptionCallBackSpace::ExceptionCallBackSample::ExceptionCallBackSample() = default;50ExceptionCallBackSpace::ExceptionCallBackSample::ExceptionCallBackSample() = default;
57 51 
58-ExceptionCallBackSpace::ExceptionCallBackSample::~ExceptionCallBackSample()52+ExceptionCallBackSpace::ExceptionCallBackSample::~ExceptionCallBackSample() { (void)Destroy(); }
59-{
60- (void)Destroy();
61-}
62 53 
63int ExceptionCallBackSpace::ExceptionCallBackSample::Init()54int ExceptionCallBackSpace::ExceptionCallBackSample::Init()
64{55{
@@ -70,7 +61,7 @@ int ExceptionCallBackSpace::ExceptionCallBackSample::Init()
70 return 0;61 return 0;
71}62}
72 63 
73-void ExceptionCallBackSpace::ExceptionCallBackSample::ThreadFunc(void *arg)64+void ExceptionCallBackSpace::ExceptionCallBackSample::ThreadFunc(void* arg)
74{65{
75 const int waitTime = 100;66 const int waitTime = 100;
76 aclrtSetCurrentContext(context_);67 aclrtSetCurrentContext(context_);
@@ -80,13 +71,13 @@ void ExceptionCallBackSpace::ExceptionCallBackSample::ThreadFunc(void *arg)
80 INFO_LOG("Thread exit");71 INFO_LOG("Thread exit");
81}72}
82 73 
83-void ExceptionCallBackSpace::ExceptionCallBackSample::CallBackFunc(void *arg)74+void ExceptionCallBackSpace::ExceptionCallBackSample::CallBackFunc(void* arg)
84{75{
85- int *data = static_cast<int*>(arg);76+ int* data = static_cast<int*>(arg);
86 INFO_LOG("After error still callback, the userdata is: %d.", *data);77 INFO_LOG("After error still callback, the userdata is: %d.", *data);
87}78}
88 79 
89-void ExceptionCallBackSpace::ExceptionCallBackSample::ExceptionCallBackFunc(aclrtExceptionInfo *exceptionInfo)80+void ExceptionCallBackSpace::ExceptionCallBackSample::ExceptionCallBackFunc(aclrtExceptionInfo* exceptionInfo)
90{81{
91 INFO_LOG("Exception occurred, callback function.");82 INFO_LOG("Exception occurred, callback function.");
92 uint32_t errorMsg = aclrtGetTaskIdFromExceptionInfo(exceptionInfo);83 uint32_t errorMsg = aclrtGetTaskIdFromExceptionInfo(exceptionInfo);
@@ -100,16 +91,16 @@ void ExceptionCallBackSpace::ExceptionCallBackSample::ExceptionCallBackFunc(aclr
100 errorMsg = aclrtGetErrorCodeFromExceptionInfo(exceptionInfo);91 errorMsg = aclrtGetErrorCodeFromExceptionInfo(exceptionInfo);
101 INFO_LOG("The error code id is %u.", errorMsg);92 INFO_LOG("The error code id is %u.", errorMsg);
102 93 
103- void *devArgsPtr = nullptr;94+ void* devArgsPtr = nullptr;
104 uint32_t devArgsLen = 0;95 uint32_t devArgsLen = 0;
105 aclError argsRet = aclrtGetArgsFromExceptionInfo(exceptionInfo, &devArgsPtr, &devArgsLen);96 aclError argsRet = aclrtGetArgsFromExceptionInfo(exceptionInfo, &devArgsPtr, &devArgsLen);
106- INFO_LOG("Exception args query ret=%d, devArgsPtr=%p, devArgsLen=%u",97+ INFO_LOG(
107- static_cast<int32_t>(argsRet), devArgsPtr, devArgsLen);98+ "Exception args query ret=%d, devArgsPtr=%p, devArgsLen=%u", static_cast<int32_t>(argsRet), devArgsPtr,
99+ devArgsLen);
108 100 
109 aclrtFuncHandle funcHandle = nullptr;101 aclrtFuncHandle funcHandle = nullptr;
110 aclError funcRet = aclrtGetFuncHandleFromExceptionInfo(exceptionInfo, &funcHandle);102 aclError funcRet = aclrtGetFuncHandleFromExceptionInfo(exceptionInfo, &funcHandle);
111- INFO_LOG("Exception func handle query ret=%d, funcHandle=%p",103+ INFO_LOG("Exception func handle query ret=%d, funcHandle=%p", static_cast<int32_t>(funcRet), funcHandle);
112- static_cast<int32_t>(funcRet), funcHandle);
113}104}
114 105 
115int ExceptionCallBackSpace::ExceptionCallBackSample::Callback()106int ExceptionCallBackSpace::ExceptionCallBackSample::Callback()
@@ -118,16 +109,16 @@ int ExceptionCallBackSpace::ExceptionCallBackSample::Callback()
118 const int blockDim = 1;109 const int blockDim = 1;
119 bool isLoop = true;110 bool isLoop = true;
120 const size_t size = sizeof(uint32_t);111 const size_t size = sizeof(uint32_t);
121- uint32_t *numDevice = nullptr;112+ uint32_t* numDevice = nullptr;
122 uint32_t taskId = 0;113 uint32_t taskId = 0;
123- CHECK_ERROR(aclrtMalloc((void **)&numDevice, size, ACL_MEM_MALLOC_HUGE_FIRST));114+ CHECK_ERROR(aclrtMalloc((void**)&numDevice, size, ACL_MEM_MALLOC_HUGE_FIRST));
124 CHECK_ERROR(aclrtMemcpy(numDevice, size, &num, size, ACL_MEMCPY_HOST_TO_DEVICE));115 CHECK_ERROR(aclrtMemcpy(numDevice, size, &num, size, ACL_MEMCPY_HOST_TO_DEVICE));
125 116 
126 thread td(ThreadFunc, &isLoop);117 thread td(ThreadFunc, &isLoop);
127 thread::id tid = td.get_id();118 thread::id tid = td.get_id();
128 ostringstream oss;119 ostringstream oss;
129 oss << tid;120 oss << tid;
130- int *userData = new int(520);121+ int* userData = new int(520);
131 uint64_t tidInt = std::stoull(oss.str());122 uint64_t tidInt = std::stoull(oss.str());
132 CHECK_ERROR(aclrtSubscribeReport(tidInt, stream_));123 CHECK_ERROR(aclrtSubscribeReport(tidInt, stream_));
133 CHECK_ERROR(aclrtSetExceptionInfoCallback(ExceptionCallBackFunc));124 CHECK_ERROR(aclrtSetExceptionInfoCallback(ExceptionCallBackFunc));
Mexample/2_advanced_features/callback/2_callback_exception/exception_callback_sample.h+16-17
@@ -14,23 +14,22 @@
14#include "utils.h"14#include "utils.h"
15#include "acl/acl.h"15#include "acl/acl.h"
16namespace ExceptionCallBackSpace {16namespace ExceptionCallBackSpace {
17- class ExceptionCallBackSample {17+class ExceptionCallBackSample {
18- public:18+public:
19- ExceptionCallBackSample();19+ ExceptionCallBackSample();
20- virtual ~ExceptionCallBackSample();20+ virtual ~ExceptionCallBackSample();
21- int Init();21+ int Init();
22- int Callback();22+ int Callback();
23- int Destroy();23+ int Destroy();
24- static void ThreadFunc(void *arg);24+ static void ThreadFunc(void* arg);
25- static void CallBackFunc(void *arg);25+ static void CallBackFunc(void* arg);
26- static void ExceptionCallBackFunc(aclrtExceptionInfo *exceptionInfo);26+ static void ExceptionCallBackFunc(aclrtExceptionInfo* exceptionInfo);
27- 
28- public:
29- static int32_t deviceId_;
30- static aclrtContext context_;
31- static aclrtStream stream_;
32- };
33-}
34 27 
28+public:
29+ static int32_t deviceId_;
30+ static aclrtContext context_;
31+ static aclrtStream stream_;
32+};
33+} // namespace ExceptionCallBackSpace
35 34 
36#endif35#endif
Mexample/2_advanced_features/callback/callback_utils.cpp+15-15
@@ -13,19 +13,19 @@
13#include "callback_utils.h"13#include "callback_utils.h"
14 14 
15namespace CallbackUtils {15namespace CallbackUtils {
16- bool IsLoopFlag(void *arg)16+bool IsLoopFlag(void* arg)
17- {17+{
18- if (*(static_cast<bool *>(arg))) {18+ if (*(static_cast<bool*>(arg))) {
19- return true;19+ return true;
20- } else {20+ } else {
21- return false;21+ return false;
22- }
23 }22 }
24- std::string GetThreadId(std::thread::id tid)23+}
25- {24+std::string GetThreadId(std::thread::id tid)
26- std::ostringstream oss;25+{
27- oss << tid;26+ std::ostringstream oss;
28- std::string tidStr = oss.str();27+ oss << tid;
29- return tidStr;28+ std::string tidStr = oss.str();
30- }29+ return tidStr;
31-}30+}
31+} // namespace CallbackUtils
Mexample/2_advanced_features/callback/callback_utils.h+3-3
@@ -15,8 +15,8 @@
15#include <string>15#include <string>
16 16 
17namespace CallbackUtils {17namespace CallbackUtils {
18- bool IsLoopFlag(void *arg);18+bool IsLoopFlag(void* arg);
19- std::string GetThreadId(std::thread::id tid);19+std::string GetThreadId(std::thread::id tid);
20-}20+} // namespace CallbackUtils
21 21 
22#endif22#endif
Mexample/2_advanced_features/ipcevent/0_ipcevent/proc_a.cpp+2-1
@@ -24,7 +24,8 @@
24#define EVENT_HANDLE_FILE "./event_handle.bin"24#define EVENT_HANDLE_FILE "./event_handle.bin"
25#define DONE_FILE "./consumer_done.flag"25#define DONE_FILE "./consumer_done.flag"
26 26 
27-int main() {27+int main()
28+{
28 CHECK_ERROR(aclInit(nullptr));29 CHECK_ERROR(aclInit(nullptr));
29 30 
30 int32_t deviceId = 0;31 int32_t deviceId = 0;
Mexample/2_advanced_features/ipcevent/0_ipcevent/proc_b.cpp+3-2
@@ -23,7 +23,8 @@
23#define EVENT_HANDLE_FILE "./event_handle.bin"23#define EVENT_HANDLE_FILE "./event_handle.bin"
24#define DONE_FILE "./consumer_done.flag"24#define DONE_FILE "./consumer_done.flag"
25 25 
26-int32_t main() {26+int32_t main()
27+{
27 CHECK_ERROR(aclInit(nullptr));28 CHECK_ERROR(aclInit(nullptr));
28 29 
29 int32_t deviceId = 0;30 int32_t deviceId = 0;
@@ -66,7 +67,7 @@ int32_t main() {
66 67 
67 // 5. 模拟消费者自身的工作68 // 5. 模拟消费者自身的工作
68 INFO_LOG("Process B: doing some work...");69 INFO_LOG("Process B: doing some work...");
69- usleep(2000000); // 2秒70+ usleep(2000000); // 2秒
70 71 
71 // 6. 记录事件，通知生产者72 // 6. 记录事件，通知生产者
72 CHECK_ERROR(aclrtRecordEvent(ipcEvent, stream));73 CHECK_ERROR(aclrtRecordEvent(ipcEvent, stream));
Mexample/2_advanced_features/ipcevent/1_ipcevent_multi_device/precheck.cpp+2-2
@@ -36,8 +36,8 @@ int main(int argc, char* argv[])
36 36 
37 INFO_LOG("Available device count: %u, required device count: %d", deviceCount, requiredDeviceCount);37 INFO_LOG("Available device count: %u, required device count: %d", deviceCount, requiredDeviceCount);
38 if (deviceCount < static_cast<uint32_t>(requiredDeviceCount)) {38 if (deviceCount < static_cast<uint32_t>(requiredDeviceCount)) {
39- INFO_LOG("[SKIP] Need at least %d devices, but only %u device(s) are available.", requiredDeviceCount,39+ INFO_LOG(
40- deviceCount);40+ "[SKIP] Need at least %d devices, but only %u device(s) are available.", requiredDeviceCount, deviceCount);
41 return 2;41 return 2;
42 }42 }
43 return 0;43 return 0;
Mexample/2_advanced_features/ipcevent/1_ipcevent_multi_device/proc_a.cpp+4-2
@@ -25,7 +25,8 @@
25#define MAX_CONSUMER_NUM 825#define MAX_CONSUMER_NUM 8
26 26 
27// 等待所有消费者完成（通过检查每个消费者创建的完成标志文件）27// 等待所有消费者完成（通过检查每个消费者创建的完成标志文件）
28-int WaitForConsumers(int expectedCount) {28+int WaitForConsumers(int expectedCount)
29+{
29 char flagFile[64];30 char flagFile[64];
30 int completed = 0;31 int completed = 0;
31 int timeoutSec = 30;32 int timeoutSec = 30;
@@ -53,7 +54,8 @@ int WaitForConsumers(int expectedCount) {
53 }54 }
54}55}
55 56 
56-int main(int argc, char* argv[]) {57+int main(int argc, char* argv[])
58+{
57 if (argc != 2) {59 if (argc != 2) {
58 ERROR_LOG("Usage: %s <number_of_consumers>", argv[0]);60 ERROR_LOG("Usage: %s <number_of_consumers>", argv[0]);
59 return -1;61 return -1;
Mexample/2_advanced_features/ipcevent/1_ipcevent_multi_device/proc_b.cpp+3-2
@@ -23,7 +23,8 @@
23 23 
24#define EVENT_HANDLE_FILE "./event_handle.bin"24#define EVENT_HANDLE_FILE "./event_handle.bin"
25 25 
26-int main(int argc, char* argv[]) {26+int main(int argc, char* argv[])
27+{
27 if (argc != 3) {28 if (argc != 3) {
28 ERROR_LOG("Usage: %s <device_id> <consumer_id>", argv[0]);29 ERROR_LOG("Usage: %s <device_id> <consumer_id>", argv[0]);
29 return -1;30 return -1;
@@ -71,7 +72,7 @@ int main(int argc, char* argv[]) {
71 72 
72 // 5. 模拟消费者自身的工作73 // 5. 模拟消费者自身的工作
73 INFO_LOG("Consumer %d: doing some work...", consumerId);74 INFO_LOG("Consumer %d: doing some work...", consumerId);
74- usleep(1000000); // 1秒75+ usleep(1000000); // 1秒
75 76 
76 // 6. 记录事件，通知其他等待者（这里主要是为了演示，实际已无等待者）77 // 6. 记录事件，通知其他等待者（这里主要是为了演示，实际已无等待者）
77 CHECK_ERROR(aclrtRecordEvent(ipcEvent, stream));78 CHECK_ERROR(aclrtRecordEvent(ipcEvent, stream));
Mexample/2_advanced_features/kernel/0_launch_kernel/main.cpp+39-46
@@ -26,12 +26,12 @@
26 26 
27namespace {27namespace {
28struct KernelBuffers {28struct KernelBuffers {
29- uint8_t *xHost = nullptr;29+ uint8_t* xHost = nullptr;
30- uint8_t *yHost = nullptr;30+ uint8_t* yHost = nullptr;
31- uint8_t *zHost = nullptr;31+ uint8_t* zHost = nullptr;
32- uint8_t *xDevice = nullptr;32+ uint8_t* xDevice = nullptr;
33- uint8_t *yDevice = nullptr;33+ uint8_t* yDevice = nullptr;
34- uint8_t *zDevice = nullptr;34+ uint8_t* zDevice = nullptr;
35};35};
36 36 
37struct RuntimeResources {37struct RuntimeResources {
@@ -44,7 +44,7 @@ struct RuntimeResources {
44 bool binLoaded = false;44 bool binLoaded = false;
45};45};
46 46 
47-void UpdateFinalResultOnError(const char *apiName, aclError ret, int32_t &finalResult)47+void UpdateFinalResultOnError(const char* apiName, aclError ret, int32_t& finalResult)
48{48{
49 if (ret == ACL_SUCCESS) {49 if (ret == ACL_SUCCESS) {
50 return;50 return;
@@ -53,7 +53,7 @@ void UpdateFinalResultOnError(const char *apiName, aclError ret, int32_t &finalR
53 finalResult = -1;53 finalResult = -1;
54}54}
55 55 
56-int32_t InitializeRuntime(RuntimeResources *runtime)56+int32_t InitializeRuntime(RuntimeResources* runtime)
57{57{
58 // Initialize ACL and create a stream on device 0.58 // Initialize ACL and create a stream on device 0.
59 CHECK_ERROR(aclInit(nullptr));59 CHECK_ERROR(aclInit(nullptr));
@@ -65,19 +65,19 @@ int32_t InitializeRuntime(RuntimeResources *runtime)
65 return 0;65 return 0;
66}66}
67 67 
68-int32_t AllocateKernelBuffers(size_t inputByteSize, size_t outputByteSize, KernelBuffers *buffers)68+int32_t AllocateKernelBuffers(size_t inputByteSize, size_t outputByteSize, KernelBuffers* buffers)
69{69{
70 // Allocate host and device buffers for kernel inputs and output.70 // Allocate host and device buffers for kernel inputs and output.
71- CHECK_ERROR(aclrtMallocHost(reinterpret_cast<void **>(&buffers->xHost), inputByteSize));71+ CHECK_ERROR(aclrtMallocHost(reinterpret_cast<void**>(&buffers->xHost), inputByteSize));
72- CHECK_ERROR(aclrtMallocHost(reinterpret_cast<void **>(&buffers->yHost), inputByteSize));72+ CHECK_ERROR(aclrtMallocHost(reinterpret_cast<void**>(&buffers->yHost), inputByteSize));
73- CHECK_ERROR(aclrtMallocHost(reinterpret_cast<void **>(&buffers->zHost), outputByteSize));73+ CHECK_ERROR(aclrtMallocHost(reinterpret_cast<void**>(&buffers->zHost), outputByteSize));
74- CHECK_ERROR(aclrtMalloc(reinterpret_cast<void **>(&buffers->xDevice), inputByteSize, ACL_MEM_MALLOC_HUGE_FIRST));74+ CHECK_ERROR(aclrtMalloc(reinterpret_cast<void**>(&buffers->xDevice), inputByteSize, ACL_MEM_MALLOC_HUGE_FIRST));
75- CHECK_ERROR(aclrtMalloc(reinterpret_cast<void **>(&buffers->yDevice), inputByteSize, ACL_MEM_MALLOC_HUGE_FIRST));75+ CHECK_ERROR(aclrtMalloc(reinterpret_cast<void**>(&buffers->yDevice), inputByteSize, ACL_MEM_MALLOC_HUGE_FIRST));
76- CHECK_ERROR(aclrtMalloc(reinterpret_cast<void **>(&buffers->zDevice), outputByteSize, ACL_MEM_MALLOC_HUGE_FIRST));76+ CHECK_ERROR(aclrtMalloc(reinterpret_cast<void**>(&buffers->zDevice), outputByteSize, ACL_MEM_MALLOC_HUGE_FIRST));
77 return 0;77 return 0;
78}78}
79 79 
80-int32_t PrepareInputData(size_t inputByteSize, const KernelBuffers &buffers)80+int32_t PrepareInputData(size_t inputByteSize, const KernelBuffers& buffers)
81{81{
82 // Load generated input files and copy them to device memory.82 // Load generated input files and copy them to device memory.
83 size_t xFileSize = inputByteSize;83 size_t xFileSize = inputByteSize;
@@ -98,14 +98,17 @@ int32_t PrepareInputData(size_t inputByteSize, const KernelBuffers &buffers)
98 return 0;98 return 0;
99}99}
100 100 
101-int32_t AppendCommonKernelArgs(aclrtArgsHandle argsHandle, uint8_t *xDevice, uint8_t *yDevice, uint8_t *zDevice)101+int32_t AppendCommonKernelArgs(aclrtArgsHandle argsHandle, uint8_t* xDevice, uint8_t* yDevice, uint8_t* zDevice)
102{102{
103 aclrtParamHandle paramHandle1 = nullptr;103 aclrtParamHandle paramHandle1 = nullptr;
104 aclrtParamHandle paramHandle2 = nullptr;104 aclrtParamHandle paramHandle2 = nullptr;
105 aclrtParamHandle paramHandle3 = nullptr;105 aclrtParamHandle paramHandle3 = nullptr;
106- CHECK_ERROR(aclrtKernelArgsAppend(argsHandle, reinterpret_cast<void **>(&xDevice), sizeof(uintptr_t), &paramHandle1));106+ CHECK_ERROR(
107- CHECK_ERROR(aclrtKernelArgsAppend(argsHandle, reinterpret_cast<void **>(&yDevice), sizeof(uintptr_t), &paramHandle2));107+ aclrtKernelArgsAppend(argsHandle, reinterpret_cast<void**>(&xDevice), sizeof(uintptr_t), &paramHandle1));
108- CHECK_ERROR(aclrtKernelArgsAppend(argsHandle, reinterpret_cast<void **>(&zDevice), sizeof(uintptr_t), &paramHandle3));108+ CHECK_ERROR(
109+ aclrtKernelArgsAppend(argsHandle, reinterpret_cast<void**>(&yDevice), sizeof(uintptr_t), &paramHandle2));
110+ CHECK_ERROR(
111+ aclrtKernelArgsAppend(argsHandle, reinterpret_cast<void**>(&zDevice), sizeof(uintptr_t), &paramHandle3));
109 return 0;112 return 0;
110}113}
111 114 
@@ -114,17 +117,17 @@ int32_t ConfigurePlaceholderArgs(aclrtArgsHandle argsHandle)
114 constexpr int32_t TOTAL_LENGTH = 8 * 2048;117 constexpr int32_t TOTAL_LENGTH = 8 * 2048;
115 constexpr int32_t TILE_NUM = 8;118 constexpr int32_t TILE_NUM = 8;
116 119 
117- int32_t *lengthHost = nullptr;120+ int32_t* lengthHost = nullptr;
118- int32_t *numHost = nullptr;121+ int32_t* numHost = nullptr;
119 aclrtParamHandle paramHandle4 = nullptr;122 aclrtParamHandle paramHandle4 = nullptr;
120 aclrtParamHandle paramHandle5 = nullptr;123 aclrtParamHandle paramHandle5 = nullptr;
121 124 
122 CHECK_ERROR(aclrtKernelArgsAppendPlaceHolder(argsHandle, &paramHandle4));125 CHECK_ERROR(aclrtKernelArgsAppendPlaceHolder(argsHandle, &paramHandle4));
123 CHECK_ERROR(aclrtKernelArgsAppendPlaceHolder(argsHandle, &paramHandle5));126 CHECK_ERROR(aclrtKernelArgsAppendPlaceHolder(argsHandle, &paramHandle5));
124 CHECK_ERROR(aclrtKernelArgsGetPlaceHolderBuffer(127 CHECK_ERROR(aclrtKernelArgsGetPlaceHolderBuffer(
125- argsHandle, paramHandle4, sizeof(TOTAL_LENGTH), reinterpret_cast<void **>(&lengthHost)));128+ argsHandle, paramHandle4, sizeof(TOTAL_LENGTH), reinterpret_cast<void**>(&lengthHost)));
126 CHECK_ERROR(aclrtKernelArgsGetPlaceHolderBuffer(129 CHECK_ERROR(aclrtKernelArgsGetPlaceHolderBuffer(
127- argsHandle, paramHandle5, sizeof(TILE_NUM), reinterpret_cast<void **>(&numHost)));130+ argsHandle, paramHandle5, sizeof(TILE_NUM), reinterpret_cast<void**>(&numHost)));
128 131 
129 *lengthHost = TOTAL_LENGTH;132 *lengthHost = TOTAL_LENGTH;
130 *numHost = TILE_NUM;133 *numHost = TILE_NUM;
@@ -132,19 +135,13 @@ int32_t ConfigurePlaceholderArgs(aclrtArgsHandle argsHandle)
132}135}
133 136 
134int32_t BuildKernelArgs(137int32_t BuildKernelArgs(
135- const std::string &mode,138+ const std::string& mode, uint8_t* xDevice, uint8_t* yDevice, uint8_t* zDevice, RuntimeResources* runtime,
136- uint8_t *xDevice,139+ aclrtFuncHandle* funcHandle, aclrtArgsHandle* argsHandle)
137- uint8_t *yDevice,
138- uint8_t *zDevice,
139- RuntimeResources *runtime,
140- aclrtFuncHandle *funcHandle,
141- aclrtArgsHandle *argsHandle)
142{140{
143 // Load the selected kernel binary and build the launch argument list.141 // Load the selected kernel binary and build the launch argument list.
144 const bool isPlaceholder = (mode == "placeholder");142 const bool isPlaceholder = (mode == "placeholder");
145- const char *filePath = isPlaceholder143+ const char* filePath = isPlaceholder ? "./out/fatbin/ascendc_kernels_placeholder/ascendc_kernels_placeholder.o" :
146- ? "./out/fatbin/ascendc_kernels_placeholder/ascendc_kernels_placeholder.o"144+ "./out/fatbin/ascendc_kernels_simple/ascendc_kernels_simple.o";
147- : "./out/fatbin/ascendc_kernels_simple/ascendc_kernels_simple.o";
148 145 
149 CHECK_ERROR(aclrtBinaryLoadFromFile(filePath, nullptr, &runtime->binHandle));146 CHECK_ERROR(aclrtBinaryLoadFromFile(filePath, nullptr, &runtime->binHandle));
150 runtime->binLoaded = true;147 runtime->binLoaded = true;
@@ -163,13 +160,8 @@ int32_t BuildKernelArgs(
163}160}
164 161 
165int32_t LaunchKernelAndWriteOutput(162int32_t LaunchKernelAndWriteOutput(
166- aclrtFuncHandle funcHandle,163+ aclrtFuncHandle funcHandle, aclrtArgsHandle argsHandle, uint32_t blockDim, aclrtStream stream, uint8_t* zDevice,
167- aclrtArgsHandle argsHandle,164+ uint8_t* zHost, size_t outputByteSize)
168- uint32_t blockDim,
169- aclrtStream stream,
170- uint8_t *zDevice,
171- uint8_t *zHost,
172- size_t outputByteSize)
173{165{
174 // Launch the kernel, synchronize the stream, and write output for verification.166 // Launch the kernel, synchronize the stream, and write output for verification.
175 CHECK_ERROR(aclrtLaunchKernelWithConfig(funcHandle, blockDim, stream, nullptr, argsHandle, nullptr));167 CHECK_ERROR(aclrtLaunchKernelWithConfig(funcHandle, blockDim, stream, nullptr, argsHandle, nullptr));
@@ -182,7 +174,7 @@ int32_t LaunchKernelAndWriteOutput(
182 return 0;174 return 0;
183}175}
184 176 
185-void ReleaseKernelResources(RuntimeResources &runtime, KernelBuffers &buffers, int32_t &finalResult)177+void ReleaseKernelResources(RuntimeResources& runtime, KernelBuffers& buffers, int32_t& finalResult)
186{178{
187 if (runtime.binLoaded) {179 if (runtime.binLoaded) {
188 UpdateFinalResultOnError(180 UpdateFinalResultOnError(
@@ -219,7 +211,7 @@ void ReleaseKernelResources(RuntimeResources &runtime, KernelBuffers &buffers, i
219 }211 }
220}212}
221 213 
222-int32_t RunKernelLaunchSample(const std::string &mode)214+int32_t RunKernelLaunchSample(const std::string& mode)
223{215{
224 const uint32_t blockDim = 8;216 const uint32_t blockDim = 8;
225 const size_t inputByteSize = 8 * 2048 * sizeof(uint16_t);217 const size_t inputByteSize = 8 * 2048 * sizeof(uint16_t);
@@ -243,11 +235,12 @@ int32_t RunKernelLaunchSample(const std::string &mode)
243 return -1;235 return -1;
244 }236 }
245 if (BuildKernelArgs(237 if (BuildKernelArgs(
246- mode, buffers.xDevice, buffers.yDevice, buffers.zDevice, &runtime, &funcHandle, &argsHandle) != 0) {238+ mode, buffers.xDevice, buffers.yDevice, buffers.zDevice, &runtime, &funcHandle, &argsHandle) != 0) {
247 return -1;239 return -1;
248 }240 }
249 if (LaunchKernelAndWriteOutput(241 if (LaunchKernelAndWriteOutput(
250- funcHandle, argsHandle, blockDim, runtime.stream, buffers.zDevice, buffers.zHost, outputByteSize) != 0) {242+ funcHandle, argsHandle, blockDim, runtime.stream, buffers.zDevice, buffers.zHost, outputByteSize) !=
243+ 0) {
251 return -1;244 return -1;
252 }245 }
253 INFO_LOG("Kernel launch sample runs in %s mode.", mode.c_str());246 INFO_LOG("Kernel launch sample runs in %s mode.", mode.c_str());
@@ -263,7 +256,7 @@ int32_t RunKernelLaunchSample(const std::string &mode)
263}256}
264} // namespace257} // namespace
265 258 
266-int32_t main(int32_t argc, char *argv[])259+int32_t main(int32_t argc, char* argv[])
267{260{
268 const std::string mode = (argc > 1) ? argv[1] : "simple";261 const std::string mode = (argc > 1) ? argv[1] : "simple";
269 if (mode != "simple" && mode != "placeholder") {262 if (mode != "simple" && mode != "placeholder") {
Mexample/2_advanced_features/kernel/1_launch_kernel_with_reslimit/main.cpp+1-1
@@ -18,7 +18,7 @@
18#include "utils.h"18#include "utils.h"
19 19 
20namespace {20namespace {
21-void UpdateFinalResultOnError(const char *apiName, aclError ret, int32_t &finalResult)21+void UpdateFinalResultOnError(const char* apiName, aclError ret, int32_t& finalResult)
22{22{
23 if (ret == ACL_SUCCESS) {23 if (ret == ACL_SUCCESS) {
24 return;24 return;
Mexample/2_advanced_features/kernel/file_ops.cpp+4-5
@@ -17,7 +17,7 @@
17#include "utils.h"17#include "utils.h"
18 18 
19namespace kernel {19namespace kernel {
20-bool ReadFile(const std::string &filePath, size_t &fileSize, void *buffer, size_t bufferSize)20+bool ReadFile(const std::string& filePath, size_t& fileSize, void* buffer, size_t bufferSize)
21{21{
22 // Read data from the target file22 // Read data from the target file
23 struct stat sBuf;23 struct stat sBuf;
@@ -38,7 +38,7 @@ bool ReadFile(const std::string &filePath, size_t &fileSize, void *buffer, size_
38 return false;38 return false;
39 }39 }
40 40 
41- std::filebuf *buf = file.rdbuf();41+ std::filebuf* buf = file.rdbuf();
42 size_t size = buf->pubseekoff(0, std::ios::end, std::ios::in);42 size_t size = buf->pubseekoff(0, std::ios::end, std::ios::in);
43 if (size == 0U) {43 if (size == 0U) {
44 ERROR_LOG("file size is 0");44 ERROR_LOG("file size is 0");
@@ -51,16 +51,15 @@ bool ReadFile(const std::string &filePath, size_t &fileSize, void *buffer, size_
51 return false;51 return false;
52 }52 }
53 buf->pubseekpos(0, std::ios::in);53 buf->pubseekpos(0, std::ios::in);
54- (void)buf->sgetn(static_cast<char *>(buffer), static_cast<std::streamsize>(size));54+ (void)buf->sgetn(static_cast<char*>(buffer), static_cast<std::streamsize>(size));
55 fileSize = size;55 fileSize = size;
56 file.close();56 file.close();
57 return true;57 return true;
58}58}
59} // namespace kernel59} // namespace kernel
60 60 
61- 
62namespace kernel {61namespace kernel {
63-bool WriteFile(const std::string &filePath, const void *buffer, size_t size)62+bool WriteFile(const std::string& filePath, const void* buffer, size_t size)
64{63{
65 // Write data to the target file64 // Write data to the target file
66 if (buffer == nullptr) {65 if (buffer == nullptr) {
Mexample/2_advanced_features/kernel/kernel_utils.h+3-3
@@ -14,7 +14,7 @@
14#include <cstddef>14#include <cstddef>
15 15 
16namespace kernel {16namespace kernel {
17- bool ReadFile(const std::string &filePath, size_t &fileSize, void *buffer, size_t bufferSize);17+bool ReadFile(const std::string& filePath, size_t& fileSize, void* buffer, size_t bufferSize);
18- bool WriteFile(const std::string &filePath, const void *buffer, size_t size);18+bool WriteFile(const std::string& filePath, const void* buffer, size_t size);
19-}19+} // namespace kernel
20#endif20#endif
Mexample/2_advanced_features/label/0_simple_label/main.cpp+4-8
@@ -13,7 +13,7 @@
13#include "utils.h"13#include "utils.h"
14 14 
15namespace {15namespace {
16-void UpdateFinalResultOnError(const char *apiName, aclError ret, int32_t &finalResult)16+void UpdateFinalResultOnError(const char* apiName, aclError ret, int32_t& finalResult)
17{17{
18 if (ret == ACL_SUCCESS) {18 if (ret == ACL_SUCCESS) {
19 return;19 return;
@@ -35,7 +35,7 @@ int main()
35 aclrtStream executeStream = nullptr;35 aclrtStream executeStream = nullptr;
36 aclrtLabel labels[labelCount] = {nullptr, nullptr};36 aclrtLabel labels[labelCount] = {nullptr, nullptr};
37 aclrtLabelList labelList = nullptr;37 aclrtLabelList labelList = nullptr;
38- uint32_t *branchIndexDevice = nullptr;38+ uint32_t* branchIndexDevice = nullptr;
39 aclmdlRI modelRI = nullptr;39 aclmdlRI modelRI = nullptr;
40 40 
41 bool aclInitialized = false;41 bool aclInitialized = false;
@@ -82,16 +82,12 @@ int main()
82 82 
83 // Prepare the branch index on device memory.83 // Prepare the branch index on device memory.
84 CHECK_ERROR(84 CHECK_ERROR(
85- aclrtMalloc(reinterpret_cast<void **>(&branchIndexDevice), sizeof(branchIndex), ACL_MEM_MALLOC_HUGE_FIRST));85+ aclrtMalloc(reinterpret_cast<void**>(&branchIndexDevice), sizeof(branchIndex), ACL_MEM_MALLOC_HUGE_FIRST));
86 branchIndexAllocated = true;86 branchIndexAllocated = true;
87 INFO_LOG("Allocated device memory for branch index.");87 INFO_LOG("Allocated device memory for branch index.");
88 88 
89 CHECK_ERROR(aclrtMemcpy(89 CHECK_ERROR(aclrtMemcpy(
90- branchIndexDevice,90+ branchIndexDevice, sizeof(branchIndex), &branchIndex, sizeof(branchIndex), ACL_MEMCPY_HOST_TO_DEVICE));
91- sizeof(branchIndex),
92- &branchIndex,
93- sizeof(branchIndex),
94- ACL_MEMCPY_HOST_TO_DEVICE));
95 INFO_LOG("Copied branch index %u from host to device.", branchIndex);91 INFO_LOG("Copied branch index %u from host to device.", branchIndex);
96 92 
97 // Create labels and organize them as a label list.93 // Create labels and organize them as a label list.
Mexample/2_advanced_features/model_ri/0_simple_model/main.cpp+20-21
@@ -8,7 +8,6 @@
8 * See LICENSE in the root of the software repository for the full text of the License.8 * See LICENSE in the root of the software repository for the full text of the License.
9 */9 */
10 10 
11- 
12#include <vector>11#include <vector>
13#include <iostream>12#include <iostream>
14#include "utils.h"13#include "utils.h"
@@ -24,17 +23,17 @@ using namespace std;
24int main()23int main()
25{24{
26 int deviceId = 0;25 int deviceId = 0;
27- void *selfDevice = nullptr;26+ void* selfDevice = nullptr;
28- void *otherDevice = nullptr;27+ void* otherDevice = nullptr;
29- void *outDevice = nullptr;28+ void* outDevice = nullptr;
30- void *outTmpDevice = nullptr;29+ void* outTmpDevice = nullptr;
31- void *selfHost = nullptr;30+ void* selfHost = nullptr;
32- void *otherHost = nullptr;31+ void* otherHost = nullptr;
33- aclTensor *self = nullptr;32+ aclTensor* self = nullptr;
34- aclTensor *other = nullptr;33+ aclTensor* other = nullptr;
35- aclScalar *alpha = nullptr;34+ aclScalar* alpha = nullptr;
36- aclTensor *out = nullptr;35+ aclTensor* out = nullptr;
37- aclTensor *outTmp = nullptr;36+ aclTensor* outTmp = nullptr;
38 vector<float> selfHostData = {1, 2, 3, 4, 5, 6, 7, 8};37 vector<float> selfHostData = {1, 2, 3, 4, 5, 6, 7, 8};
39 vector<float> otherHostData = {2, 2, 2, 2, 2, 2, 2, 2};38 vector<float> otherHostData = {2, 2, 2, 2, 2, 2, 2, 2};
40 vector<float> outHostData = {0, 0, 0, 0, 0, 0, 0, 0};39 vector<float> outHostData = {0, 0, 0, 0, 0, 0, 0, 0};
@@ -43,11 +42,11 @@ int main()
43 float alphaValue = 1.1f;42 float alphaValue = 1.1f;
44 uint64_t addWorkspaceSize = 0;43 uint64_t addWorkspaceSize = 0;
45 uint64_t mulWorkspaceSize = 0;44 uint64_t mulWorkspaceSize = 0;
46- aclOpExecutor *addExecutor;45+ aclOpExecutor* addExecutor;
47- aclOpExecutor *mulExecutor;46+ aclOpExecutor* mulExecutor;
48 aclrtContext context;47 aclrtContext context;
49 const int loopCount = 4;48 const int loopCount = 4;
50- int64_t size = ModelUtils::GetShapeSize(shape) * sizeof(float);49+ int64_t size = ModelUtils::GetShapeSize(shape) * sizeof(float);
51 50 
52 CHECK_ERROR(aclInit(NULL));51 CHECK_ERROR(aclInit(NULL));
53 CHECK_ERROR(aclrtSetDevice(deviceId));52 CHECK_ERROR(aclrtSetDevice(deviceId));
@@ -58,16 +57,16 @@ int main()
58 alpha = aclCreateScalar(&alphaValue, aclDataType::ACL_FLOAT);57 alpha = aclCreateScalar(&alphaValue, aclDataType::ACL_FLOAT);
59 ModelUtils::CreateAclTensor(shape, &outDevice, aclDataType::ACL_FLOAT, &out);58 ModelUtils::CreateAclTensor(shape, &outDevice, aclDataType::ACL_FLOAT, &out);
60 ModelUtils::CreateAclTensor(shape, &outTmpDevice, aclDataType::ACL_FLOAT, &outTmp);59 ModelUtils::CreateAclTensor(shape, &outTmpDevice, aclDataType::ACL_FLOAT, &outTmp);
61- 60+ 
62 // 创建workspace来执行算子，不同算子的workspace可能不同，该处计算outTmp = self + other * alpha61 // 创建workspace来执行算子，不同算子的workspace可能不同，该处计算outTmp = self + other * alpha
63 aclnnAddGetWorkspaceSize(self, other, alpha, outTmp, &addWorkspaceSize, &addExecutor);62 aclnnAddGetWorkspaceSize(self, other, alpha, outTmp, &addWorkspaceSize, &addExecutor);
64- void *addWorkspaceAddr = nullptr;63+ void* addWorkspaceAddr = nullptr;
65 if (addWorkspaceSize > 0) {64 if (addWorkspaceSize > 0) {
66 CHECK_ERROR(aclrtMalloc(&addWorkspaceAddr, addWorkspaceSize, ACL_MEM_MALLOC_HUGE_FIRST));65 CHECK_ERROR(aclrtMalloc(&addWorkspaceAddr, addWorkspaceSize, ACL_MEM_MALLOC_HUGE_FIRST));
67 }66 }
68 // 该算子计算out = outTmp * other，outTmp是上一个add算子的输出。67 // 该算子计算out = outTmp * other，outTmp是上一个add算子的输出。
69 aclnnMulGetWorkspaceSize(outTmp, other, out, &mulWorkspaceSize, &mulExecutor);68 aclnnMulGetWorkspaceSize(outTmp, other, out, &mulWorkspaceSize, &mulExecutor);
70- void *mulWorkspaceAddr = nullptr;69+ void* mulWorkspaceAddr = nullptr;
71 if (mulWorkspaceSize > 0) {70 if (mulWorkspaceSize > 0) {
72 CHECK_ERROR(aclrtMalloc(&mulWorkspaceAddr, mulWorkspaceSize, ACL_MEM_MALLOC_HUGE_FIRST));71 CHECK_ERROR(aclrtMalloc(&mulWorkspaceAddr, mulWorkspaceSize, ACL_MEM_MALLOC_HUGE_FIRST));
73 }72 }
@@ -77,7 +76,7 @@ int main()
77 CHECK_ERROR(aclrtCreateStream(&stream));76 CHECK_ERROR(aclrtCreateStream(&stream));
78 CHECK_ERROR(aclrtMallocHost(&selfHost, size));77 CHECK_ERROR(aclrtMallocHost(&selfHost, size));
79 CHECK_ERROR(aclrtMemcpy(selfHost, size, selfHostData.data(), size, ACL_MEMCPY_HOST_TO_HOST));78 CHECK_ERROR(aclrtMemcpy(selfHost, size, selfHostData.data(), size, ACL_MEMCPY_HOST_TO_HOST));
80- 79+ 
81 // 标志着model的开始直到end。此处设置的mode为ACL_MODEL_RI_CAPTURE_MODE_GLOBAL，禁止执行非安全的函数80 // 标志着model的开始直到end。此处设置的mode为ACL_MODEL_RI_CAPTURE_MODE_GLOBAL，禁止执行非安全的函数
82 // 该model执行了将内存从host复制到device侧，然后调用了一个add算子和一个mul算子81 // 该model执行了将内存从host复制到device侧，然后调用了一个add算子和一个mul算子
83 CHECK_ERROR(aclmdlRICaptureBegin(stream, ACL_MODEL_RI_CAPTURE_MODE_GLOBAL));82 CHECK_ERROR(aclmdlRICaptureBegin(stream, ACL_MODEL_RI_CAPTURE_MODE_GLOBAL));
@@ -96,9 +95,9 @@ int main()
96 // 标志model创建结束，且保存为modelRI,后续调用可以通过modelRI进行95 // 标志model创建结束，且保存为modelRI,后续调用可以通过modelRI进行
97 CHECK_ERROR(aclmdlRICaptureEnd(stream, &modelRI));96 CHECK_ERROR(aclmdlRICaptureEnd(stream, &modelRI));
98 // 打印信息可以在日志中查找97 // 打印信息可以在日志中查找
99- const char *jsonPath = "./modelRI.json";98+ const char* jsonPath = "./modelRI.json";
100 CHECK_ERROR(aclmdlRIDebugJsonPrint(modelRI, jsonPath, 0));99 CHECK_ERROR(aclmdlRIDebugJsonPrint(modelRI, jsonPath, 0));
101- 100+ 
102 // 循环执行刚才创建的modelRI101 // 循环执行刚才创建的modelRI
103 for (int i = 0; i < loopCount; i++) {102 for (int i = 0; i < loopCount; i++) {
104 INFO_LOG("execute model, loop count: %d.", i + 1);103 INFO_LOG("execute model, loop count: %d.", i + 1);
Mexample/2_advanced_features/model_ri/1_model_update/main.cpp+18-18
@@ -7,7 +7,7 @@
7 * INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE.7 * INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE.
8 * See LICENSE in the root of the software repository for the full text of the License.8 * See LICENSE in the root of the software repository for the full text of the License.
9 */9 */
10- 10+ 
11#include <vector>11#include <vector>
12#include <iostream>12#include <iostream>
13#include "utils.h"13#include "utils.h"
@@ -22,16 +22,16 @@ using namespace std;
22int main()22int main()
23{23{
24 int deviceId = 0;24 int deviceId = 0;
25- void *selfDevice = nullptr;25+ void* selfDevice = nullptr;
26- void *otherDevice = nullptr;26+ void* otherDevice = nullptr;
27- void *outDevice = nullptr;27+ void* outDevice = nullptr;
28- void *outTmpDevice = nullptr;28+ void* outTmpDevice = nullptr;
29- aclTensor *self = nullptr;29+ aclTensor* self = nullptr;
30- aclTensor *other = nullptr;30+ aclTensor* other = nullptr;
31- aclScalar *alpha = nullptr;31+ aclScalar* alpha = nullptr;
32- aclScalar *updateAlpha = nullptr;32+ aclScalar* updateAlpha = nullptr;
33- aclTensor *out = nullptr;33+ aclTensor* out = nullptr;
34- aclTensor *outTmp = nullptr;34+ aclTensor* outTmp = nullptr;
35 vector<float> selfHostData = {1, 2, 3, 4, 5, 6, 7, 8};35 vector<float> selfHostData = {1, 2, 3, 4, 5, 6, 7, 8};
36 vector<float> otherHostData = {2, 2, 2, 2, 2, 2, 2, 2};36 vector<float> otherHostData = {2, 2, 2, 2, 2, 2, 2, 2};
37 vector<float> outHostData = {0, 0, 0, 0, 0, 0, 0, 0};37 vector<float> outHostData = {0, 0, 0, 0, 0, 0, 0, 0};
@@ -42,9 +42,9 @@ int main()
42 uint64_t firstAddWorkspaceSize = 0;42 uint64_t firstAddWorkspaceSize = 0;
43 uint64_t secondAddWorkspaceSize = 0;43 uint64_t secondAddWorkspaceSize = 0;
44 uint64_t updateAddWorkspaceSize = 0;44 uint64_t updateAddWorkspaceSize = 0;
45- aclOpExecutor *firstAddExecutor;45+ aclOpExecutor* firstAddExecutor;
46- aclOpExecutor *secondAddExecutor;46+ aclOpExecutor* secondAddExecutor;
47- aclOpExecutor *updateAddExecutor;47+ aclOpExecutor* updateAddExecutor;
48 aclrtContext context;48 aclrtContext context;
49 const int loopCount = 2;49 const int loopCount = 2;
50 int64_t size = ModelUtils::GetShapeSize(shape) * sizeof(float);50 int64_t size = ModelUtils::GetShapeSize(shape) * sizeof(float);
@@ -59,23 +59,23 @@ int main()
59 updateAlpha = aclCreateScalar(&updateAlphaValue, aclDataType::ACL_FLOAT);59 updateAlpha = aclCreateScalar(&updateAlphaValue, aclDataType::ACL_FLOAT);
60 ModelUtils::CreateAclTensor(shape, &outDevice, aclDataType::ACL_FLOAT, &out);60 ModelUtils::CreateAclTensor(shape, &outDevice, aclDataType::ACL_FLOAT, &out);
61 ModelUtils::CreateAclTensor(shape, &outTmpDevice, aclDataType::ACL_FLOAT, &outTmp);61 ModelUtils::CreateAclTensor(shape, &outTmpDevice, aclDataType::ACL_FLOAT, &outTmp);
62- 62+ 
63 // 该算子计算outTmp = self + other * alpha63 // 该算子计算outTmp = self + other * alpha
64 aclnnAddGetWorkspaceSize(self, other, alpha, outTmp, &firstAddWorkspaceSize, &firstAddExecutor);64 aclnnAddGetWorkspaceSize(self, other, alpha, outTmp, &firstAddWorkspaceSize, &firstAddExecutor);
65- void *firstAddWorkspaceAddr = nullptr;65+ void* firstAddWorkspaceAddr = nullptr;
66 if (firstAddWorkspaceSize > 0) {66 if (firstAddWorkspaceSize > 0) {
67 CHECK_ERROR(aclrtMalloc(&firstAddWorkspaceAddr, firstAddWorkspaceSize, ACL_MEM_MALLOC_HUGE_FIRST));67 CHECK_ERROR(aclrtMalloc(&firstAddWorkspaceAddr, firstAddWorkspaceSize, ACL_MEM_MALLOC_HUGE_FIRST));
68 }68 }
69 69 
70 // 该算子计算outTmp = outTmp + other * alpha70 // 该算子计算outTmp = outTmp + other * alpha
71 aclnnAddGetWorkspaceSize(outTmp, other, alpha, out, &secondAddWorkspaceSize, &secondAddExecutor);71 aclnnAddGetWorkspaceSize(outTmp, other, alpha, out, &secondAddWorkspaceSize, &secondAddExecutor);
72- void *secondAddWorkspaceAddr = nullptr;72+ void* secondAddWorkspaceAddr = nullptr;
73 if (secondAddWorkspaceSize > 0) {73 if (secondAddWorkspaceSize > 0) {
74 CHECK_ERROR(aclrtMalloc(&secondAddWorkspaceAddr, secondAddWorkspaceSize, ACL_MEM_MALLOC_HUGE_FIRST));74 CHECK_ERROR(aclrtMalloc(&secondAddWorkspaceAddr, secondAddWorkspaceSize, ACL_MEM_MALLOC_HUGE_FIRST));
75 }75 }
76 // 该算子计算outTmp = outTmp + other * updateAlpha76 // 该算子计算outTmp = outTmp + other * updateAlpha
77 aclnnAddGetWorkspaceSize(outTmp, other, updateAlpha, out, &updateAddWorkspaceSize, &updateAddExecutor);77 aclnnAddGetWorkspaceSize(outTmp, other, updateAlpha, out, &updateAddWorkspaceSize, &updateAddExecutor);
78- void *updateAddWorkspaceAddr = nullptr;78+ void* updateAddWorkspaceAddr = nullptr;
79 if (updateAddWorkspaceSize > 0) {79 if (updateAddWorkspaceSize > 0) {
80 CHECK_ERROR(aclrtMalloc(&updateAddWorkspaceAddr, updateAddWorkspaceSize, ACL_MEM_MALLOC_HUGE_FIRST));80 CHECK_ERROR(aclrtMalloc(&updateAddWorkspaceAddr, updateAddWorkspaceSize, ACL_MEM_MALLOC_HUGE_FIRST));
81 }81 }
Mexample/2_advanced_features/model_ri/2_model_switch/main.cpp+34-32
@@ -23,26 +23,26 @@ int main()
23 int deviceId = 0;23 int deviceId = 0;
24 int blockDim = 1;24 int blockDim = 1;
25 // 初始化数据，包括三个算子任务所需要的输入输出25 // 初始化数据，包括三个算子任务所需要的输入输出
26- uint32_t *numDevice = nullptr;26+ uint32_t* numDevice = nullptr;
27- void *selfDevice1 = nullptr;27+ void* selfDevice1 = nullptr;
28- void *otherDevice1 = nullptr;28+ void* otherDevice1 = nullptr;
29- void *outDevice1 = nullptr;29+ void* outDevice1 = nullptr;
30- void *selfDevice2 = nullptr;30+ void* selfDevice2 = nullptr;
31- void *otherDevice2 = nullptr;31+ void* otherDevice2 = nullptr;
32- void *outDevice2 = nullptr;32+ void* outDevice2 = nullptr;
33- void *selfDevice3 = nullptr;33+ void* selfDevice3 = nullptr;
34- void *otherDevice3 = nullptr;34+ void* otherDevice3 = nullptr;
35- void *outDevice3 = nullptr;35+ void* outDevice3 = nullptr;
36- aclTensor *self1 = nullptr;36+ aclTensor* self1 = nullptr;
37- aclTensor *other1 = nullptr;37+ aclTensor* other1 = nullptr;
38- aclScalar *alpha = nullptr;38+ aclScalar* alpha = nullptr;
39- aclTensor *out1 = nullptr;39+ aclTensor* out1 = nullptr;
40- aclTensor *self2 = nullptr;40+ aclTensor* self2 = nullptr;
41- aclTensor *other2 = nullptr;41+ aclTensor* other2 = nullptr;
42- aclTensor *out2 = nullptr;42+ aclTensor* out2 = nullptr;
43- aclTensor *self3 = nullptr;43+ aclTensor* self3 = nullptr;
44- aclTensor *other3 = nullptr;44+ aclTensor* other3 = nullptr;
45- aclTensor *out3 = nullptr;45+ aclTensor* out3 = nullptr;
46 vector<float> selfHostData1 = {1, 1, 1, 1, 1, 1, 1, 1};46 vector<float> selfHostData1 = {1, 1, 1, 1, 1, 1, 1, 1};
47 vector<float> otherHostData1 = {2, 2, 2, 2, 2, 2, 2, 2};47 vector<float> otherHostData1 = {2, 2, 2, 2, 2, 2, 2, 2};
48 vector<float> outHostData1 = {0, 0, 0, 0, 0, 0, 0, 0};48 vector<float> outHostData1 = {0, 0, 0, 0, 0, 0, 0, 0};
@@ -57,9 +57,9 @@ int main()
57 uint64_t addWorkspaceSize1 = 0;57 uint64_t addWorkspaceSize1 = 0;
58 uint64_t addWorkspaceSize2 = 0;58 uint64_t addWorkspaceSize2 = 0;
59 uint64_t addWorkspaceSize3 = 0;59 uint64_t addWorkspaceSize3 = 0;
60- aclOpExecutor *addExecutor1;60+ aclOpExecutor* addExecutor1;
61- aclOpExecutor *addExecutor2;61+ aclOpExecutor* addExecutor2;
62- aclOpExecutor *addExecutor3;62+ aclOpExecutor* addExecutor3;
63 aclrtContext context;63 aclrtContext context;
64 int64_t size = ModelUtils::GetShapeSize(shape) * sizeof(float);64 int64_t size = ModelUtils::GetShapeSize(shape) * sizeof(float);
65 CHECK_ERROR(aclInit(NULL));65 CHECK_ERROR(aclInit(NULL));
@@ -80,37 +80,37 @@ int main()
80 ModelUtils::CreateAclTensor(shape, &outDevice2, aclDataType::ACL_FLOAT, &out2);80 ModelUtils::CreateAclTensor(shape, &outDevice2, aclDataType::ACL_FLOAT, &out2);
81 ModelUtils::CreateAclTensor(shape, &outDevice3, aclDataType::ACL_FLOAT, &out3);81 ModelUtils::CreateAclTensor(shape, &outDevice3, aclDataType::ACL_FLOAT, &out3);
82 aclnnAddGetWorkspaceSize(self1, other1, alpha, out1, &addWorkspaceSize1, &addExecutor1);82 aclnnAddGetWorkspaceSize(self1, other1, alpha, out1, &addWorkspaceSize1, &addExecutor1);
83- void *addWorkspaceAddr1 = nullptr;83+ void* addWorkspaceAddr1 = nullptr;
84 if (addWorkspaceSize1 > 0) {84 if (addWorkspaceSize1 > 0) {
85 CHECK_ERROR(aclrtMalloc(&addWorkspaceAddr1, addWorkspaceSize1, ACL_MEM_MALLOC_HUGE_FIRST));85 CHECK_ERROR(aclrtMalloc(&addWorkspaceAddr1, addWorkspaceSize1, ACL_MEM_MALLOC_HUGE_FIRST));
86 }86 }
87 aclnnAddGetWorkspaceSize(self2, other2, alpha, out2, &addWorkspaceSize2, &addExecutor2);87 aclnnAddGetWorkspaceSize(self2, other2, alpha, out2, &addWorkspaceSize2, &addExecutor2);
88- void *addWorkspaceAddr2 = nullptr;88+ void* addWorkspaceAddr2 = nullptr;
89 if (addWorkspaceSize2 > 0) {89 if (addWorkspaceSize2 > 0) {
90 CHECK_ERROR(aclrtMalloc(&addWorkspaceAddr2, addWorkspaceSize2, ACL_MEM_MALLOC_HUGE_FIRST));90 CHECK_ERROR(aclrtMalloc(&addWorkspaceAddr2, addWorkspaceSize2, ACL_MEM_MALLOC_HUGE_FIRST));
91 }91 }
92 aclnnAddGetWorkspaceSize(self3, other3, alpha, out3, &addWorkspaceSize3, &addExecutor3);92 aclnnAddGetWorkspaceSize(self3, other3, alpha, out3, &addWorkspaceSize3, &addExecutor3);
93- void *addWorkspaceAddr3 = nullptr;93+ void* addWorkspaceAddr3 = nullptr;
94 if (addWorkspaceSize3 > 0) {94 if (addWorkspaceSize3 > 0) {
95 CHECK_ERROR(aclrtMalloc(&addWorkspaceAddr3, addWorkspaceSize3, ACL_MEM_MALLOC_HUGE_FIRST));95 CHECK_ERROR(aclrtMalloc(&addWorkspaceAddr3, addWorkspaceSize3, ACL_MEM_MALLOC_HUGE_FIRST));
96 }96 }
97 // switchstream函数需要的参数为device侧数据的地址，这里创建两个device侧的地址用于比较，一个为1，一个为297 // switchstream函数需要的参数为device侧数据的地址，这里创建两个device侧的地址用于比较，一个为1，一个为2
98 int32_t rightValue1 = 1;98 int32_t rightValue1 = 1;
99 int32_t rightValue2 = 2;99 int32_t rightValue2 = 2;
100- void *rightDevice1 = nullptr;100+ void* rightDevice1 = nullptr;
101- void *rightDevice2 = nullptr;101+ void* rightDevice2 = nullptr;
102 aclrtCondition condition = ACL_RT_EQUAL;102 aclrtCondition condition = ACL_RT_EQUAL;
103 aclrtCompareDataType dataType = ACL_RT_SWITCH_INT32;103 aclrtCompareDataType dataType = ACL_RT_SWITCH_INT32;
104 CHECK_ERROR(aclrtMalloc(&rightDevice1, sizeof(int32_t), ACL_MEM_MALLOC_HUGE_FIRST));104 CHECK_ERROR(aclrtMalloc(&rightDevice1, sizeof(int32_t), ACL_MEM_MALLOC_HUGE_FIRST));
105 CHECK_ERROR(aclrtMemcpy(rightDevice1, sizeof(int32_t), &rightValue1, sizeof(int32_t), ACL_MEMCPY_HOST_TO_DEVICE));105 CHECK_ERROR(aclrtMemcpy(rightDevice1, sizeof(int32_t), &rightValue1, sizeof(int32_t), ACL_MEMCPY_HOST_TO_DEVICE));
106 CHECK_ERROR(aclrtMalloc(&rightDevice2, sizeof(int32_t), ACL_MEM_MALLOC_HUGE_FIRST));106 CHECK_ERROR(aclrtMalloc(&rightDevice2, sizeof(int32_t), ACL_MEM_MALLOC_HUGE_FIRST));
107 CHECK_ERROR(aclrtMemcpy(rightDevice2, sizeof(int32_t), &rightValue2, sizeof(int32_t), ACL_MEMCPY_HOST_TO_DEVICE));107 CHECK_ERROR(aclrtMemcpy(rightDevice2, sizeof(int32_t), &rightValue2, sizeof(int32_t), ACL_MEMCPY_HOST_TO_DEVICE));
108- CHECK_ERROR(aclrtMalloc((void **)&numDevice, sizeof(uint32_t), ACL_MEM_MALLOC_HUGE_FIRST));108+ CHECK_ERROR(aclrtMalloc((void**)&numDevice, sizeof(uint32_t), ACL_MEM_MALLOC_HUGE_FIRST));
109 uint32_t numInit = 0;109 uint32_t numInit = 0;
110 CHECK_ERROR(aclrtMemcpy(numDevice, sizeof(uint32_t), &numInit, sizeof(uint32_t), ACL_MEMCPY_HOST_TO_DEVICE));110 CHECK_ERROR(aclrtMemcpy(numDevice, sizeof(uint32_t), &numInit, sizeof(uint32_t), ACL_MEMCPY_HOST_TO_DEVICE));
111 // 通过 DeviceToDevice 异步拷贝将其置为 1，供 SwitchStream 分支验证。111 // 通过 DeviceToDevice 异步拷贝将其置为 1，供 SwitchStream 分支验证。
112 uint32_t val1 = 1;112 uint32_t val1 = 1;
113- void *targetValDev_1 = nullptr;113+ void* targetValDev_1 = nullptr;
114 CHECK_ERROR(aclrtMalloc(&targetValDev_1, sizeof(uint32_t), ACL_MEM_MALLOC_HUGE_FIRST));114 CHECK_ERROR(aclrtMalloc(&targetValDev_1, sizeof(uint32_t), ACL_MEM_MALLOC_HUGE_FIRST));
115 CHECK_ERROR(aclrtMemcpy(targetValDev_1, sizeof(uint32_t), &val1, sizeof(uint32_t), ACL_MEMCPY_HOST_TO_DEVICE));115 CHECK_ERROR(aclrtMemcpy(targetValDev_1, sizeof(uint32_t), &val1, sizeof(uint32_t), ACL_MEMCPY_HOST_TO_DEVICE));
116 // 创建五个流：四个工作流 + 一个汇聚流endStream116 // 创建五个流：四个工作流 + 一个汇聚流endStream
@@ -143,7 +143,8 @@ int main()
143 CHECK_ERROR(aclrtMemcpyAsync(otherDevice2, size, otherHostData2.data(), size, ACL_MEMCPY_HOST_TO_DEVICE, stream2));143 CHECK_ERROR(aclrtMemcpyAsync(otherDevice2, size, otherHostData2.data(), size, ACL_MEMCPY_HOST_TO_DEVICE, stream2));
144 CHECK_ERROR(aclrtMemcpyAsync(otherDevice3, size, otherHostData3.data(), size, ACL_MEMCPY_HOST_TO_DEVICE, stream3));144 CHECK_ERROR(aclrtMemcpyAsync(otherDevice3, size, otherHostData3.data(), size, ACL_MEMCPY_HOST_TO_DEVICE, stream3));
145 aclnnAdd(addWorkspaceAddr1, addWorkspaceSize1, addExecutor1, stream1);145 aclnnAdd(addWorkspaceAddr1, addWorkspaceSize1, addExecutor1, stream1);
146- CHECK_ERROR(aclrtMemcpyAsync(numDevice, sizeof(uint32_t), targetValDev_1, sizeof(uint32_t), ACL_MEMCPY_DEVICE_TO_DEVICE, stream1));146+ CHECK_ERROR(aclrtMemcpyAsync(
147+ numDevice, sizeof(uint32_t), targetValDev_1, sizeof(uint32_t), ACL_MEMCPY_DEVICE_TO_DEVICE, stream1));
147 // stream2的任务为下发一个aclnn算子，完成后激活endStream148 // stream2的任务为下发一个aclnn算子，完成后激活endStream
148 aclnnAdd(addWorkspaceAddr2, addWorkspaceSize2, addExecutor2, stream2);149 aclnnAdd(addWorkspaceAddr2, addWorkspaceSize2, addExecutor2, stream2);
149 CHECK_ERROR(aclrtActiveStream(endStream, stream2));150 CHECK_ERROR(aclrtActiveStream(endStream, stream2));
@@ -181,7 +182,8 @@ int main()
181 // 以便让图里面录入的那句 DeviceToDevice 复制时把 2 写进 numDevice。182 // 以便让图里面录入的那句 DeviceToDevice 复制时把 2 写进 numDevice。
182 uint32_t val2 = 2;183 uint32_t val2 = 2;
183 // 使用独立的 copyStream 拷贝 2 到 targetValDev_1 中184 // 使用独立的 copyStream 拷贝 2 到 targetValDev_1 中
184- CHECK_ERROR(aclrtMemcpyAsync(targetValDev_1, sizeof(uint32_t), &val2, sizeof(uint32_t), ACL_MEMCPY_HOST_TO_DEVICE, copyStream));185+ CHECK_ERROR(aclrtMemcpyAsync(
186+ targetValDev_1, sizeof(uint32_t), &val2, sizeof(uint32_t), ACL_MEMCPY_HOST_TO_DEVICE, copyStream));
185 CHECK_ERROR(aclrtSynchronizeStream(copyStream));187 CHECK_ERROR(aclrtSynchronizeStream(copyStream));
186 // 将上一次的主机结果清零，以免混淆188 // 将上一次的主机结果清零，以免混淆
187 outHostData1.assign(outHostData1.size(), 0);189 outHostData1.assign(outHostData1.size(), 0);
Mexample/2_advanced_features/model_ri/4_model_sync_external/main.cpp+9-10
@@ -12,12 +12,11 @@
12#include "utils.h"12#include "utils.h"
13#include "acl/acl.h"13#include "acl/acl.h"
14 14 
15-static bool CheckResult(const std::vector<int32_t> &actual, const std::vector<int32_t> &expected)15+static bool CheckResult(const std::vector<int32_t>& actual, const std::vector<int32_t>& expected)
16{16{
17 for (size_t i = 0; i < actual.size(); ++i) {17 for (size_t i = 0; i < actual.size(); ++i) {
18 if (actual[i] != expected[i]) {18 if (actual[i] != expected[i]) {
19- ERROR_LOG("Check result failed at index %zu: actual=%d, expected=%d",19+ ERROR_LOG("Check result failed at index %zu: actual=%d, expected=%d", i, actual[i], expected[i]);
20- i, actual[i], expected[i]);
21 return false;20 return false;
22 }21 }
23 }22 }
@@ -42,8 +41,8 @@ static int Scenario1StreamRecordGraphWaitExternal()
42 aclrtEvent event;41 aclrtEvent event;
43 CHECK_ERROR(aclrtCreateEventExWithFlag(&event, ACL_EVENT_SYNC));42 CHECK_ERROR(aclrtCreateEventExWithFlag(&event, ACL_EVENT_SYNC));
44 43 
45- void *srcDevice = nullptr;44+ void* srcDevice = nullptr;
46- void *dstDevice = nullptr;45+ void* dstDevice = nullptr;
47 CHECK_ERROR(aclrtMalloc(&srcDevice, size, ACL_MEM_MALLOC_HUGE_FIRST));46 CHECK_ERROR(aclrtMalloc(&srcDevice, size, ACL_MEM_MALLOC_HUGE_FIRST));
48 CHECK_ERROR(aclrtMalloc(&dstDevice, size, ACL_MEM_MALLOC_HUGE_FIRST));47 CHECK_ERROR(aclrtMalloc(&dstDevice, size, ACL_MEM_MALLOC_HUGE_FIRST));
49 48 
@@ -91,8 +90,8 @@ static int Scenario2GraphRecordExternalStreamWait()
91 aclrtEvent event;90 aclrtEvent event;
92 CHECK_ERROR(aclrtCreateEventExWithFlag(&event, ACL_EVENT_SYNC));91 CHECK_ERROR(aclrtCreateEventExWithFlag(&event, ACL_EVENT_SYNC));
93 92 
94- void *srcDevice = nullptr;93+ void* srcDevice = nullptr;
95- void *dstDevice = nullptr;94+ void* dstDevice = nullptr;
96 CHECK_ERROR(aclrtMalloc(&srcDevice, size, ACL_MEM_MALLOC_HUGE_FIRST));95 CHECK_ERROR(aclrtMalloc(&srcDevice, size, ACL_MEM_MALLOC_HUGE_FIRST));
97 CHECK_ERROR(aclrtMalloc(&dstDevice, size, ACL_MEM_MALLOC_HUGE_FIRST));96 CHECK_ERROR(aclrtMalloc(&dstDevice, size, ACL_MEM_MALLOC_HUGE_FIRST));
98 97 
@@ -141,9 +140,9 @@ static int Scenario3Graph1RecordGraph2WaitExternal()
141 aclrtEvent event;140 aclrtEvent event;
142 CHECK_ERROR(aclrtCreateEventExWithFlag(&event, ACL_EVENT_SYNC));141 CHECK_ERROR(aclrtCreateEventExWithFlag(&event, ACL_EVENT_SYNC));
143 142 
144- void *srcDevice = nullptr;143+ void* srcDevice = nullptr;
145- void *midDevice = nullptr;144+ void* midDevice = nullptr;
146- void *dstDevice = nullptr;145+ void* dstDevice = nullptr;
147 CHECK_ERROR(aclrtMalloc(&srcDevice, size, ACL_MEM_MALLOC_HUGE_FIRST));146 CHECK_ERROR(aclrtMalloc(&srcDevice, size, ACL_MEM_MALLOC_HUGE_FIRST));
148 CHECK_ERROR(aclrtMalloc(&midDevice, size, ACL_MEM_MALLOC_HUGE_FIRST));147 CHECK_ERROR(aclrtMalloc(&midDevice, size, ACL_MEM_MALLOC_HUGE_FIRST));
149 CHECK_ERROR(aclrtMalloc(&dstDevice, size, ACL_MEM_MALLOC_HUGE_FIRST));148 CHECK_ERROR(aclrtMalloc(&dstDevice, size, ACL_MEM_MALLOC_HUGE_FIRST));
Mexample/2_advanced_features/model_ri/model_utils.cpp+37-44
@@ -19,47 +19,40 @@
19#include "model_utils.h"19#include "model_utils.h"
20 20 
21namespace ModelUtils {21namespace ModelUtils {
22- int64_t GetShapeSize(const std::vector<int64_t> &shape)22+int64_t GetShapeSize(const std::vector<int64_t>& shape)
23- {23+{
24- int64_t shapeSize = 1;24+ int64_t shapeSize = 1;
25- for (auto i : shape) {25+ for (auto i : shape) {
26- shapeSize *= i;26+ shapeSize *= i;
27- }27+ }
28- return shapeSize;28+ return shapeSize;
29- }29+}
30- int CreateAclTensor(const std::vector<int64_t> &shape, void **deviceAddr, aclDataType dataType,30+int CreateAclTensor(const std::vector<int64_t>& shape, void** deviceAddr, aclDataType dataType, aclTensor** tensor)
31- aclTensor **tensor)31+{
32- {32+ auto size = GetShapeSize(shape) * sizeof(float);
33- auto size = GetShapeSize(shape) * sizeof(float);33+ // 申请Device侧内存
34- // 申请Device侧内存34+ auto ret = aclrtMalloc(deviceAddr, size, ACL_MEM_MALLOC_HUGE_FIRST);
35- auto ret = aclrtMalloc(deviceAddr, size, ACL_MEM_MALLOC_HUGE_FIRST);35+ // 计算连续tensor的stride
36- // 计算连续tensor的stride36+ std::vector<int64_t> strides(shape.size(), 1);
37- std::vector<int64_t> strides(shape.size(), 1);37+ for (int64_t i = shape.size() - 2; i >= 0; i--) {
38- for (int64_t i = shape.size() - 2; i >= 0; i--) {38+ strides[i] = shape[i + 1] * strides[i + 1];
39- strides[i] = shape[i + 1] * strides[i + 1];39+ }
40- }40+ // 调用aclCreateTensor接口创建aclTensor
41- // 调用aclCreateTensor接口创建aclTensor41+ *tensor = aclCreateTensor(
42- *tensor = aclCreateTensor(shape.data(),42+ shape.data(), shape.size(), dataType, strides.data(), 0, aclFormat::ACL_FORMAT_ND, shape.data(), shape.size(),
43- shape.size(),43+ *deviceAddr);
44- dataType,44+ return 0;
45- strides.data(),45+}
46- 0,46+void PrintArray(const std::vector<float>& data)
47- aclFormat::ACL_FORMAT_ND,47+{
48- shape.data(),48+ const int size = data.size();
49- shape.size(),49+ const int precision = 4;
50- *deviceAddr);50+ std::ostringstream oss;
51- return 0;51+ oss << std::fixed << std::setprecision(precision);
52- }52+ oss << "The vector data is: ";
53- void PrintArray(const std::vector<float>& data)53+ for (int j = 0; j < size; j++) {
54- {54+ oss << data[j] << " ";
55- const int size = data.size();55+ }
56- const int precision = 4;56+ INFO_LOG("%s", oss.str().c_str());
57- std::ostringstream oss;57+}
58- oss << std::fixed << std::setprecision(precision);58+} // namespace ModelUtils
59- oss << "The vector data is: ";
60- for (int j = 0; j < size; j++) {
61- oss << data[j] <<" ";
62- }
63- INFO_LOG("%s", oss.str().c_str());
64- }
65-}
Mexample/2_advanced_features/model_ri/model_utils.h+5-6
@@ -19,12 +19,11 @@
19#include "aclnnop/aclnn_add.h"19#include "aclnnop/aclnn_add.h"
20 20 
21namespace ModelUtils {21namespace ModelUtils {
22- int64_t GetShapeSize(const std::vector<int64_t> &shape);22+int64_t GetShapeSize(const std::vector<int64_t>& shape);
23 23 
24- int CreateAclTensor(const std::vector<int64_t> &shape, void **deviceAddr, aclDataType dataType,24+int CreateAclTensor(const std::vector<int64_t>& shape, void** deviceAddr, aclDataType dataType, aclTensor** tensor);
25- aclTensor **tensor);25+ 
26- 26+void PrintArray(const std::vector<float>& data);
27- void PrintArray(const std::vector<float>& data);27+} // namespace ModelUtils
28-}
29 28 
30#endif29#endif
Mexample/2_advanced_features/notify/0_ipc_notify_withpid/proc_a.cpp+8-4
@@ -9,7 +9,7 @@
9 */9 */
10 10 
11/*11/*
12- * This sample demonstrates Inter-Process Communication (IPC) notify for 12+ * This sample demonstrates Inter-Process Communication (IPC) notify for
13 * task synchronization across multiple devices and processes.13 * task synchronization across multiple devices and processes.
14 * This sample enables process whitelist verification.14 * This sample enables process whitelist verification.
15 */15 */
@@ -42,14 +42,16 @@ int32_t main()
42 42 
43 // Allocate memory on the device43 // Allocate memory on the device
44 uint64_t size = 1 * 1024 * 1024;44 uint64_t size = 1 * 1024 * 1024;
45- int *devPtr = nullptr;45+ int* devPtr = nullptr;
46 CHECK_ERROR(aclrtMalloc((void**)&devPtr, size, ACL_MEM_MALLOC_HUGE_FIRST));46 CHECK_ERROR(aclrtMalloc((void**)&devPtr, size, ACL_MEM_MALLOC_HUGE_FIRST));
47 47 
48 // Export a shareable identifier for IPC memory sharing48 // Export a shareable identifier for IPC memory sharing
49 char memName[256] = "name";49 char memName[256] = "name";
50 size_t lenName = 65;50 size_t lenName = 65;
51 CHECK_ERROR(aclrtIpcMemGetExportKey(devPtr, size, memName, lenName, 1));51 CHECK_ERROR(aclrtIpcMemGetExportKey(devPtr, size, memName, lenName, 1));
52- INFO_LOG("Process A: get a shareable identifier for IPC memory sharing successfully, shareable identifier = %s", memName);52+ INFO_LOG(
53+ "Process A: get a shareable identifier for IPC memory sharing successfully, shareable identifier = %s",
54+ memName);
53 55 
54 // Transfer the shareable identifier to Process B by writing it to the file56 // Transfer the shareable identifier to Process B by writing it to the file
55 memory::WriteFile("file/memName.bin", "file/memName.bin.done", memName, strlen(memName));57 memory::WriteFile("file/memName.bin", "file/memName.bin.done", memName, strlen(memName));
@@ -61,7 +63,9 @@ int32_t main()
61 // Export a shareable identifier for IPC notify sharing63 // Export a shareable identifier for IPC notify sharing
62 char notifyName[256] = "name";64 char notifyName[256] = "name";
63 CHECK_ERROR(aclrtNotifyGetExportKey(notify, notifyName, lenName, 0));65 CHECK_ERROR(aclrtNotifyGetExportKey(notify, notifyName, lenName, 0));
64- INFO_LOG("Process A: get a shareable identifier for IPC notify sharing successfully, shareable identifier = %s", notifyName);66+ INFO_LOG(
67+ "Process A: get a shareable identifier for IPC notify sharing successfully, shareable identifier = %s",
68+ notifyName);
65 69 
66 // Read Process B's pid from the file70 // Read Process B's pid from the file
67 int32_t pid = 0;71 int32_t pid = 0;
Mexample/2_advanced_features/notify/0_ipc_notify_withpid/proc_b.cpp+10-6
@@ -9,7 +9,7 @@
9 */9 */
10 10 
11/*11/*
12- * This sample demonstrates Inter-Process Communication (IPC) notify for 12+ * This sample demonstrates Inter-Process Communication (IPC) notify for
13 * task synchronization across multiple devices and processes.13 * task synchronization across multiple devices and processes.
14 * This sample enables process whitelist verification.14 * This sample enables process whitelist verification.
15 */15 */
@@ -44,10 +44,12 @@ int32_t main()
44 // Get the shareable key from the file44 // Get the shareable key from the file
45 char memName[256] = "name";45 char memName[256] = "name";
46 memory::ReadFile("file/memName.bin", "file/memName.bin.done", memName, sizeof(memName));46 memory::ReadFile("file/memName.bin", "file/memName.bin.done", memName, sizeof(memName));
47- INFO_LOG("Process B: get the shareable identifier for IPC memory sharing successfully, shareable identifier = %s", memName);47+ INFO_LOG(
48+ "Process B: get the shareable identifier for IPC memory sharing successfully, shareable identifier = %s",
49+ memName);
48 50 
49 // Return a device memory pointer accessible to Process B51 // Return a device memory pointer accessible to Process B
50- void *devPtr = nullptr;52+ void* devPtr = nullptr;
51 uint64_t size = 1 * 1024 * 1024;53 uint64_t size = 1 * 1024 * 1024;
52 CHECK_ERROR(aclrtIpcMemImportByKey(&devPtr, memName, 0));54 CHECK_ERROR(aclrtIpcMemImportByKey(&devPtr, memName, 0));
53 55 
@@ -58,11 +60,13 @@ int32_t main()
58 60 
59 // Write Process B's pid to the file61 // Write Process B's pid to the file
60 memory::WriteFile("file/pid.bin", "file/pid.bin.done", &pid, sizeof(pid));62 memory::WriteFile("file/pid.bin", "file/pid.bin.done", &pid, sizeof(pid));
61- 63+ 
62 // Get the shareable key from the file64 // Get the shareable key from the file
63 char notifyName[256] = "name";65 char notifyName[256] = "name";
64 memory::ReadFile("file/notifyName.bin", "file/notifyName.bin.done", notifyName, sizeof(notifyName));66 memory::ReadFile("file/notifyName.bin", "file/notifyName.bin.done", notifyName, sizeof(notifyName));
65- INFO_LOG("Process B: get the shareable identifier for IPC notify sharing successfully, shareable identifier = %s", notifyName);67+ INFO_LOG(
68+ "Process B: get the shareable identifier for IPC notify sharing successfully, shareable identifier = %s",
69+ notifyName);
66 70 
67 // Get key information and return the notify accessible to Process B71 // Get key information and return the notify accessible to Process B
68 aclrtNotify notify = nullptr;72 aclrtNotify notify = nullptr;
@@ -84,7 +88,7 @@ int32_t main()
84 // Release notify and memory resources88 // Release notify and memory resources
85 CHECK_ERROR(aclrtSynchronizeStream(stream));89 CHECK_ERROR(aclrtSynchronizeStream(stream));
86 CHECK_ERROR(aclrtIpcMemClose(memName));90 CHECK_ERROR(aclrtIpcMemClose(memName));
87- CHECK_ERROR(aclrtDestroyNotify(notify)); 91+ CHECK_ERROR(aclrtDestroyNotify(notify));
88 } else {92 } else {
89 ERROR_LOG("The target devices do not support data interaction");93 ERROR_LOG("The target devices do not support data interaction");
90 }94 }
Mexample/2_advanced_features/notify/1_ipc_notify_withoutpid/proc_a.cpp+8-4
@@ -9,7 +9,7 @@
9 */9 */
10 10 
11/*11/*
12- * This sample demonstrates Inter-Process Communication (IPC) notify for 12+ * This sample demonstrates Inter-Process Communication (IPC) notify for
13 * task synchronization across multiple devices and processes.13 * task synchronization across multiple devices and processes.
14 * This sample does not enable whitelist verification.14 * This sample does not enable whitelist verification.
15 */15 */
@@ -42,14 +42,16 @@ int32_t main()
42 42 
43 // Allocate memory on the device43 // Allocate memory on the device
44 uint64_t size = 1 * 1024 * 1024;44 uint64_t size = 1 * 1024 * 1024;
45- int *devPtr = nullptr;45+ int* devPtr = nullptr;
46 CHECK_ERROR(aclrtMalloc((void**)&devPtr, size, ACL_MEM_MALLOC_HUGE_FIRST));46 CHECK_ERROR(aclrtMalloc((void**)&devPtr, size, ACL_MEM_MALLOC_HUGE_FIRST));
47 47 
48 // Export a shareable identifier for IPC memory sharing48 // Export a shareable identifier for IPC memory sharing
49 char memName[256] = "name";49 char memName[256] = "name";
50 size_t lenName = 65;50 size_t lenName = 65;
51 CHECK_ERROR(aclrtIpcMemGetExportKey(devPtr, size, memName, lenName, 1));51 CHECK_ERROR(aclrtIpcMemGetExportKey(devPtr, size, memName, lenName, 1));
52- INFO_LOG("Process A: get a shareable identifier for IPC memory sharing successfully, shareable identifier = %s", memName);52+ INFO_LOG(
53+ "Process A: get a shareable identifier for IPC memory sharing successfully, shareable identifier = %s",
54+ memName);
53 55 
54 // Transfer the shareable identifier to Process B by writing it to the file56 // Transfer the shareable identifier to Process B by writing it to the file
55 memory::WriteFile("file/memName.bin", "file/memName.bin.done", memName, strlen(memName));57 memory::WriteFile("file/memName.bin", "file/memName.bin.done", memName, strlen(memName));
@@ -62,7 +64,9 @@ int32_t main()
62 char notifyName[256] = "name";64 char notifyName[256] = "name";
63 uint64_t flag = 2;65 uint64_t flag = 2;
64 CHECK_ERROR(aclrtNotifyGetExportKey(notify, notifyName, lenName, flag));66 CHECK_ERROR(aclrtNotifyGetExportKey(notify, notifyName, lenName, flag));
65- INFO_LOG("Process A: get a shareable identifier for IPC notify sharing successfully, shareable identifier = %s", notifyName);67+ INFO_LOG(
68+ "Process A: get a shareable identifier for IPC notify sharing successfully, shareable identifier = %s",
69+ notifyName);
66 70 
67 // Transfer the shareable identifier to Process B by writing it to the file71 // Transfer the shareable identifier to Process B by writing it to the file
68 memory::WriteFile("file/notifyName.bin", "file/notifyName.bin.done", notifyName, strlen(notifyName));72 memory::WriteFile("file/notifyName.bin", "file/notifyName.bin.done", notifyName, strlen(notifyName));
Mexample/2_advanced_features/notify/1_ipc_notify_withoutpid/proc_b.cpp+11-7
@@ -9,7 +9,7 @@
9 */9 */
10 10 
11/*11/*
12- * This sample demonstrates Inter-Process Communication (IPC) notify for 12+ * This sample demonstrates Inter-Process Communication (IPC) notify for
13 * task synchronization across multiple devices and processes.13 * task synchronization across multiple devices and processes.
14 * This sample does not enable whitelist verification.14 * This sample does not enable whitelist verification.
15 */15 */
@@ -24,7 +24,7 @@
24#include "memory_runtime_utils.h"24#include "memory_runtime_utils.h"
25 25 
26namespace {26namespace {
27-int CheckAndEnablePeerAccess(int32_t deviceId, int32_t peerDeviceId, bool *isSupported)27+int CheckAndEnablePeerAccess(int32_t deviceId, int32_t peerDeviceId, bool* isSupported)
28{28{
29 int32_t canAccessPeer = 0;29 int32_t canAccessPeer = 0;
30 aclError ret = aclrtDeviceCanAccessPeer(&canAccessPeer, deviceId, peerDeviceId);30 aclError ret = aclrtDeviceCanAccessPeer(&canAccessPeer, deviceId, peerDeviceId);
@@ -61,17 +61,21 @@ int32_t main()
61 // Get the shareable identifier from the file61 // Get the shareable identifier from the file
62 char memName[256] = "name";62 char memName[256] = "name";
63 memory::ReadFile("file/memName.bin", "file/memName.bin.done", memName, sizeof(memName));63 memory::ReadFile("file/memName.bin", "file/memName.bin.done", memName, sizeof(memName));
64- INFO_LOG("Process B: get the shareable identifier for IPC memory sharing successfully, shareable identifier = %s", memName);64+ INFO_LOG(
65+ "Process B: get the shareable identifier for IPC memory sharing successfully, shareable identifier = %s",
66+ memName);
65 67 
66 // Get identifier information and return a device memory pointer accessible to Process B68 // Get identifier information and return a device memory pointer accessible to Process B
67- void *devPtr = nullptr;69+ void* devPtr = nullptr;
68 uint64_t size = 1 * 1024 * 1024;70 uint64_t size = 1 * 1024 * 1024;
69 CHECK_ERROR(aclrtIpcMemImportByKey(&devPtr, memName, 0));71 CHECK_ERROR(aclrtIpcMemImportByKey(&devPtr, memName, 0));
70- 72+ 
71 // Get the shareable identifier from the file73 // Get the shareable identifier from the file
72 char notifyName[256] = "name";74 char notifyName[256] = "name";
73 memory::ReadFile("file/notifyName.bin", "file/notifyName.bin.done", notifyName, sizeof(notifyName));75 memory::ReadFile("file/notifyName.bin", "file/notifyName.bin.done", notifyName, sizeof(notifyName));
74- INFO_LOG("Process B: get the shareable identifier for IPC notify sharing successfully, shareable identifier = %s", notifyName);76+ INFO_LOG(
77+ "Process B: get the shareable identifier for IPC notify sharing successfully, shareable identifier = %s",
78+ notifyName);
75 79 
76 // Return the notify accessible to Process B80 // Return the notify accessible to Process B
77 aclrtNotify notify = nullptr;81 aclrtNotify notify = nullptr;
@@ -93,7 +97,7 @@ int32_t main()
93 // Release notify and memory resources97 // Release notify and memory resources
94 CHECK_ERROR(aclrtSynchronizeStream(stream));98 CHECK_ERROR(aclrtSynchronizeStream(stream));
95 CHECK_ERROR(aclrtIpcMemClose(memName));99 CHECK_ERROR(aclrtIpcMemClose(memName));
96- CHECK_ERROR(aclrtDestroyNotify(notify)); 100+ CHECK_ERROR(aclrtDestroyNotify(notify));
97 } else {101 } else {
98 ERROR_LOG("The target devices do not support data interaction");102 ERROR_LOG("The target devices do not support data interaction");
99 }103 }
Mexample/2_advanced_features/tdt_channel/0_simple_channel/main.cpp+12-17
@@ -16,7 +16,7 @@
16#include "utils.h"16#include "utils.h"
17 17 
18namespace {18namespace {
19-int DumpDataset(const acltdtDataset *dataset)19+int DumpDataset(const acltdtDataset* dataset)
20{20{
21 const size_t datasetSize = acltdtGetDatasetSize(dataset);21 const size_t datasetSize = acltdtGetDatasetSize(dataset);
22 INFO_LOG("Dataset size: %zu", datasetSize);22 INFO_LOG("Dataset size: %zu", datasetSize);
@@ -24,7 +24,7 @@ int DumpDataset(const acltdtDataset *dataset)
24 return 0;24 return 0;
25 }25 }
26 26 
27- acltdtDataItem *item = acltdtGetDataItem(dataset, 0);27+ acltdtDataItem* item = acltdtGetDataItem(dataset, 0);
28 if (item == nullptr) {28 if (item == nullptr) {
29 ERROR_LOG("acltdtGetDataItem returned nullptr");29 ERROR_LOG("acltdtGetDataItem returned nullptr");
30 return -1;30 return -1;
@@ -36,15 +36,12 @@ int DumpDataset(const acltdtDataset *dataset)
36 CHECK_ERROR(acltdtGetDimsFromItem(item, dims.data(), dimNum));36 CHECK_ERROR(acltdtGetDimsFromItem(item, dims.data(), dimNum));
37 }37 }
38 38 
39- float *tensorData = static_cast<float *>(acltdtGetDataAddrFromItem(item));39+ float* tensorData = static_cast<float*>(acltdtGetDataAddrFromItem(item));
40 INFO_LOG(40 INFO_LOG(
41 "Tensor type=%d, data type=%d, bytes=%zu, dims=(%lld, %lld), firstValue=%.3f",41 "Tensor type=%d, data type=%d, bytes=%zu, dims=(%lld, %lld), firstValue=%.3f",
42- static_cast<int32_t>(acltdtGetTensorTypeFromItem(item)),42+ static_cast<int32_t>(acltdtGetTensorTypeFromItem(item)), static_cast<int32_t>(acltdtGetDataTypeFromItem(item)),
43- static_cast<int32_t>(acltdtGetDataTypeFromItem(item)),43+ acltdtGetDataSizeFromItem(item), static_cast<long long>(dimNum > 0 ? dims[0] : 0),
44- acltdtGetDataSizeFromItem(item),44+ static_cast<long long>(dimNum > 1 ? dims[1] : 0), tensorData == nullptr ? 0.0F : tensorData[0]);
45- static_cast<long long>(dimNum > 0 ? dims[0] : 0),
46- static_cast<long long>(dimNum > 1 ? dims[1] : 0),
47- tensorData == nullptr ? 0.0F : tensorData[0]);
48 return 0;45 return 0;
49}46}
50 47 
@@ -56,9 +53,9 @@ int main()
56 constexpr size_t kChannelCapacity = 2;53 constexpr size_t kChannelCapacity = 2;
57 constexpr int32_t kTimeoutMs = 1000;54 constexpr int32_t kTimeoutMs = 1000;
58 std::vector<float> hostTensor = {1.0F, 2.0F, 3.0F, 4.0F};55 std::vector<float> hostTensor = {1.0F, 2.0F, 3.0F, 4.0F};
59- acltdtDataset *sendDataset = nullptr;56+ acltdtDataset* sendDataset = nullptr;
60- acltdtDataset *recvDataset = nullptr;57+ acltdtDataset* recvDataset = nullptr;
61- acltdtChannelHandle *channel = nullptr;58+ acltdtChannelHandle* channel = nullptr;
62 bool aclInitialized = false;59 bool aclInitialized = false;
63 bool deviceSet = false;60 bool deviceSet = false;
64 61 
@@ -70,9 +67,8 @@ int main()
70 67 
71 channel = acltdtCreateChannelWithCapacity(deviceId, "simple_tdt_channel", kChannelCapacity);68 channel = acltdtCreateChannelWithCapacity(deviceId, "simple_tdt_channel", kChannelCapacity);
72 if (channel == nullptr) {69 if (channel == nullptr) {
73- WARN_LOG(70+ WARN_LOG("acltdtCreateChannelWithCapacity returned nullptr: this sample needs a queue-backed TDT channel "
74- "acltdtCreateChannelWithCapacity returned nullptr: this sample needs a queue-backed TDT channel "71+ "so it can send and receive within one host process");
75- "so it can send and receive within one host process");
76 return 0;72 return 0;
77 }73 }
78 74 
@@ -120,8 +116,7 @@ int main()
120 if (deviceSet) {116 if (deviceSet) {
121 tdt::UpdateFinalResultOnError(117 tdt::UpdateFinalResultOnError(
122 "aclrtResetDeviceForce(static_cast<int32_t>(deviceId))",118 "aclrtResetDeviceForce(static_cast<int32_t>(deviceId))",
123- aclrtResetDeviceForce(static_cast<int32_t>(deviceId)),119+ aclrtResetDeviceForce(static_cast<int32_t>(deviceId)), finalResult);
124- finalResult);
125 }120 }
126 if (aclInitialized) {121 if (aclInitialized) {
127 tdt::UpdateFinalResultOnError("aclFinalize()", aclFinalize(), finalResult);122 tdt::UpdateFinalResultOnError("aclFinalize()", aclFinalize(), finalResult);
Mexample/2_advanced_features/tdt_channel/1_channel_capacity/main.cpp+16-17
@@ -26,10 +26,10 @@ int main()
26 std::vector<float> firstTensor = {10.0F, 20.0F};26 std::vector<float> firstTensor = {10.0F, 20.0F};
27 std::vector<float> secondTensor = {30.0F, 40.0F};27 std::vector<float> secondTensor = {30.0F, 40.0F};
28 std::vector<float> thirdTensor = {50.0F, 60.0F};28 std::vector<float> thirdTensor = {50.0F, 60.0F};
29- acltdtDataset *firstDataset = nullptr;29+ acltdtDataset* firstDataset = nullptr;
30- acltdtDataset *secondDataset = nullptr;30+ acltdtDataset* secondDataset = nullptr;
31- acltdtDataset *thirdDataset = nullptr;31+ acltdtDataset* thirdDataset = nullptr;
32- acltdtChannelHandle *channel = nullptr;32+ acltdtChannelHandle* channel = nullptr;
33 bool aclInitialized = false;33 bool aclInitialized = false;
34 bool deviceSet = false;34 bool deviceSet = false;
35 35 
@@ -41,9 +41,8 @@ int main()
41 41 
42 channel = acltdtCreateChannelWithCapacity(deviceId, "capacity_tdt_channel", kChannelCapacity);42 channel = acltdtCreateChannelWithCapacity(deviceId, "capacity_tdt_channel", kChannelCapacity);
43 if (channel == nullptr) {43 if (channel == nullptr) {
44- WARN_LOG(44+ WARN_LOG("acltdtCreateChannelWithCapacity returned nullptr: the current runtime likely does not enable "
45- "acltdtCreateChannelWithCapacity returned nullptr: the current runtime likely does not enable "45+ "capacity-limited TDT channels on this product/build");
46- "capacity-limited TDT channels on this product/build");
47 return 0;46 return 0;
48 }47 }
49 48 
@@ -60,7 +59,7 @@ int main()
60 return -1;59 return -1;
61 }60 }
62 61 
63- acltdtDataItem *firstItem = acltdtGetDataItem(firstDataset, 0);62+ acltdtDataItem* firstItem = acltdtGetDataItem(firstDataset, 0);
64 if (!tdt::CheckNotNull(firstItem, "acltdtGetDataItem")) {63 if (!tdt::CheckNotNull(firstItem, "acltdtGetDataItem")) {
65 return -1;64 return -1;
66 }65 }
@@ -70,9 +69,7 @@ int main()
70 aclError sliceRet = acltdtGetSliceInfoFromItem(firstItem, &sliceNum, &sliceId);69 aclError sliceRet = acltdtGetSliceInfoFromItem(firstItem, &sliceNum, &sliceId);
71 INFO_LOG(70 INFO_LOG(
72 "Slice info ret=%d, sliceNum=%zu, sliceId=%zu, tensorType=%d, datasetName=%s",71 "Slice info ret=%d, sliceNum=%zu, sliceId=%zu, tensorType=%d, datasetName=%s",
73- static_cast<int32_t>(sliceRet),72+ static_cast<int32_t>(sliceRet), sliceNum, sliceId,
74- sliceNum,
75- sliceId,
76 static_cast<int32_t>(acltdtGetTensorTypeFromItem(firstItem)),73 static_cast<int32_t>(acltdtGetTensorTypeFromItem(firstItem)),
77 acltdtGetDatasetName(firstDataset) == nullptr ? "<null>" : acltdtGetDatasetName(firstDataset));74 acltdtGetDatasetName(firstDataset) == nullptr ? "<null>" : acltdtGetDatasetName(firstDataset));
78 75 
@@ -93,9 +90,7 @@ int main()
93 INFO_LOG(90 INFO_LOG(
94 "Second send reached capacity pressure immediately: ret=%d, queried channel size=%zu, "91 "Second send reached capacity pressure immediately: ret=%d, queried channel size=%zu, "
95 "configured capacity=%zu",92 "configured capacity=%zu",
96- static_cast<int32_t>(secondSendRet),93+ static_cast<int32_t>(secondSendRet), channelSize, kChannelCapacity);
97- channelSize,
98- kChannelCapacity);
99 } else {94 } else {
100 ERROR_LOG(95 ERROR_LOG(
101 "Operation failed: acltdtSendTensor(channel, secondDataset, 0) returned error code %d",96 "Operation failed: acltdtSendTensor(channel, secondDataset, 0) returned error code %d",
@@ -109,17 +104,21 @@ int main()
109 if (channel != nullptr) {104 if (channel != nullptr) {
110 const aclError cleanRet = acltdtCleanChannel(channel);105 const aclError cleanRet = acltdtCleanChannel(channel);
111 if (cleanRet != ACL_SUCCESS) {106 if (cleanRet != ACL_SUCCESS) {
112- ERROR_LOG("Operation failed: acltdtCleanChannel(channel) returned error code %d", static_cast<int32_t>(cleanRet));107+ ERROR_LOG(
108+ "Operation failed: acltdtCleanChannel(channel) returned error code %d", static_cast<int32_t>(cleanRet));
113 finalResult = -1;109 finalResult = -1;
114 }110 }
115 const aclError stopRet = acltdtStopChannel(channel);111 const aclError stopRet = acltdtStopChannel(channel);
116 if (stopRet != ACL_SUCCESS) {112 if (stopRet != ACL_SUCCESS) {
117- ERROR_LOG("Operation failed: acltdtStopChannel(channel) returned error code %d", static_cast<int32_t>(stopRet));113+ ERROR_LOG(
114+ "Operation failed: acltdtStopChannel(channel) returned error code %d", static_cast<int32_t>(stopRet));
118 finalResult = -1;115 finalResult = -1;
119 }116 }
120 const aclError destroyRet = acltdtDestroyChannel(channel);117 const aclError destroyRet = acltdtDestroyChannel(channel);
121 if (destroyRet != ACL_SUCCESS) {118 if (destroyRet != ACL_SUCCESS) {
122- ERROR_LOG("Operation failed: acltdtDestroyChannel(channel) returned error code %d", static_cast<int32_t>(destroyRet));119+ ERROR_LOG(
120+ "Operation failed: acltdtDestroyChannel(channel) returned error code %d",
121+ static_cast<int32_t>(destroyRet));
123 finalResult = -1;122 finalResult = -1;
124 }123 }
125 }124 }
Mexample/2_advanced_features/tdt_common_utils.h+9-13
@@ -18,7 +18,7 @@
18#include "utils.h"18#include "utils.h"
19 19 
20namespace tdt {20namespace tdt {
21-inline bool CheckNotNull(const void *ptr, const char *name)21+inline bool CheckNotNull(const void* ptr, const char* name)
22{22{
23 if (ptr == nullptr) {23 if (ptr == nullptr) {
24 ERROR_LOG("%s is nullptr", name);24 ERROR_LOG("%s is nullptr", name);
@@ -27,21 +27,17 @@ inline bool CheckNotNull(const void *ptr, const char *name)
27 return true;27 return true;
28}28}
29 29 
30-inline acltdtDataset *CreateFloatDataset(std::vector<float> &values)30+inline acltdtDataset* CreateFloatDataset(std::vector<float>& values)
31{31{
32 int64_t dims[] = {1, static_cast<int64_t>(values.size())};32 int64_t dims[] = {1, static_cast<int64_t>(values.size())};
33- acltdtDataItem *item = acltdtCreateDataItem(33+ acltdtDataItem* item = acltdtCreateDataItem(
34- ACL_TENSOR_DATA_TENSOR,34+ ACL_TENSOR_DATA_TENSOR, dims, sizeof(dims) / sizeof(dims[0]), ACL_FLOAT, values.data(),
35- dims,
36- sizeof(dims) / sizeof(dims[0]),
37- ACL_FLOAT,
38- values.data(),
39 values.size() * sizeof(float));35 values.size() * sizeof(float));
40 if (item == nullptr) {36 if (item == nullptr) {
41 return nullptr;37 return nullptr;
42 }38 }
43 39 
44- acltdtDataset *dataset = acltdtCreateDataset();40+ acltdtDataset* dataset = acltdtCreateDataset();
45 if (dataset == nullptr) {41 if (dataset == nullptr) {
46 (void)acltdtDestroyDataItem(item);42 (void)acltdtDestroyDataItem(item);
47 return nullptr;43 return nullptr;
@@ -55,7 +51,7 @@ inline acltdtDataset *CreateFloatDataset(std::vector<float> &values)
55 return dataset;51 return dataset;
56}52}
57 53 
58-inline void DestroyDatasetAndItems(acltdtDataset *dataset)54+inline void DestroyDatasetAndItems(acltdtDataset* dataset)
59{55{
60 if (dataset == nullptr) {56 if (dataset == nullptr) {
61 return;57 return;
@@ -63,7 +59,7 @@ inline void DestroyDatasetAndItems(acltdtDataset *dataset)
63 59 
64 const size_t datasetSize = acltdtGetDatasetSize(dataset);60 const size_t datasetSize = acltdtGetDatasetSize(dataset);
65 for (size_t i = 0; i < datasetSize; ++i) {61 for (size_t i = 0; i < datasetSize; ++i) {
66- acltdtDataItem *item = acltdtGetDataItem(dataset, i);62+ acltdtDataItem* item = acltdtGetDataItem(dataset, i);
67 if (item != nullptr) {63 if (item != nullptr) {
68 (void)acltdtDestroyDataItem(item);64 (void)acltdtDestroyDataItem(item);
69 }65 }
@@ -71,7 +67,7 @@ inline void DestroyDatasetAndItems(acltdtDataset *dataset)
71 (void)acltdtDestroyDataset(dataset);67 (void)acltdtDestroyDataset(dataset);
72}68}
73 69 
74-inline void DestroyDataset(acltdtDataset *dataset)70+inline void DestroyDataset(acltdtDataset* dataset)
75{71{
76 if (dataset == nullptr) {72 if (dataset == nullptr) {
77 return;73 return;
@@ -79,7 +75,7 @@ inline void DestroyDataset(acltdtDataset *dataset)
79 (void)acltdtDestroyDataset(dataset);75 (void)acltdtDestroyDataset(dataset);
80}76}
81 77 
82-inline void UpdateFinalResultOnError(const char *apiName, aclError ret, int32_t &finalResult)78+inline void UpdateFinalResultOnError(const char* apiName, aclError ret, int32_t& finalResult)
83{79{
84 if (ret == ACL_SUCCESS) {80 if (ret == ACL_SUCCESS) {
85 return;81 return;
Mexample/2_advanced_features/tdt_queue/0_simple_queue/main.cpp+6-5
@@ -18,12 +18,12 @@ int main()
18{18{
19 const int32_t deviceId = 0;19 const int32_t deviceId = 0;
20 const char queueName[] = "simple_queue";20 const char queueName[] = "simple_queue";
21- const char *queueNamePtr = queueName;21+ const char* queueNamePtr = queueName;
22- constexpr size_t kQueueNamePtrSize = sizeof(const char *);22+ constexpr size_t kQueueNamePtrSize = sizeof(const char*);
23 const uint32_t depth = 4;23 const uint32_t depth = 4;
24 24 
25 uint32_t qid = 0;25 uint32_t qid = 0;
26- acltdtQueueAttr *attr = nullptr;26+ acltdtQueueAttr* attr = nullptr;
27 bool aclInitialized = false;27 bool aclInitialized = false;
28 bool deviceSet = false;28 bool deviceSet = false;
29 bool queueCreated = false;29 bool queueCreated = false;
@@ -42,11 +42,12 @@ int main()
42 CHECK_ERROR(acltdtSetQueueAttr(attr, ACL_TDT_QUEUE_NAME_PTR, kQueueNamePtrSize, &queueNamePtr));42 CHECK_ERROR(acltdtSetQueueAttr(attr, ACL_TDT_QUEUE_NAME_PTR, kQueueNamePtrSize, &queueNamePtr));
43 CHECK_ERROR(acltdtSetQueueAttr(attr, ACL_TDT_QUEUE_DEPTH_UINT32, sizeof(depth), &depth));43 CHECK_ERROR(acltdtSetQueueAttr(attr, ACL_TDT_QUEUE_DEPTH_UINT32, sizeof(depth), &depth));
44 44 
45- const char *queriedName = nullptr;45+ const char* queriedName = nullptr;
46 uint32_t queriedDepth = 0;46 uint32_t queriedDepth = 0;
47 size_t realLen = 0;47 size_t realLen = 0;
48 CHECK_ERROR(acltdtGetQueueAttr(attr, ACL_TDT_QUEUE_NAME_PTR, kQueueNamePtrSize, &realLen, &queriedName));48 CHECK_ERROR(acltdtGetQueueAttr(attr, ACL_TDT_QUEUE_NAME_PTR, kQueueNamePtrSize, &realLen, &queriedName));
49- CHECK_ERROR(acltdtGetQueueAttr(attr, ACL_TDT_QUEUE_DEPTH_UINT32, sizeof(queriedDepth), &realLen, &queriedDepth));49+ CHECK_ERROR(
50+ acltdtGetQueueAttr(attr, ACL_TDT_QUEUE_DEPTH_UINT32, sizeof(queriedDepth), &realLen, &queriedDepth));
50 INFO_LOG("QueueAttr name=%s, depth=%u", queriedName == nullptr ? "<null>" : queriedName, queriedDepth);51 INFO_LOG("QueueAttr name=%s, depth=%u", queriedName == nullptr ? "<null>" : queriedName, queriedDepth);
51 52 
52 CHECK_ERROR(acltdtCreateQueue(attr, &qid));53 CHECK_ERROR(acltdtCreateQueue(attr, &qid));
Mexample/2_advanced_features/tdt_queue/1_queue_route/main.cpp+39-70
@@ -15,7 +15,7 @@
15#include "utils.h"15#include "utils.h"
16 16 
17namespace {17namespace {
18-constexpr size_t kQueueNamePtrSize = sizeof(const char *);18+constexpr size_t kQueueNamePtrSize = sizeof(const char*);
19 19 
20struct QueueRouteSample {20struct QueueRouteSample {
21 static constexpr int32_t kDeviceId = 0;21 static constexpr int32_t kDeviceId = 0;
@@ -29,19 +29,19 @@ struct QueueRouteSample {
29 bool dstQueueCreated = false;29 bool dstQueueCreated = false;
30 bool routeAdded = false;30 bool routeAdded = false;
31 bool bindAttempted = false;31 bool bindAttempted = false;
32- acltdtQueueRoute *route = nullptr;32+ acltdtQueueRoute* route = nullptr;
33- acltdtQueueRouteList *bindList = nullptr;33+ acltdtQueueRouteList* bindList = nullptr;
34- acltdtQueueRouteQueryInfo *queryInfo = nullptr;34+ acltdtQueueRouteQueryInfo* queryInfo = nullptr;
35- acltdtQueueRouteList *queryList = nullptr;35+ acltdtQueueRouteList* queryList = nullptr;
36};36};
37 37 
38-aclError CreateQueueWithAttr(const char *name, uint32_t depth, uint32_t *qid)38+aclError CreateQueueWithAttr(const char* name, uint32_t depth, uint32_t* qid)
39{39{
40 if (qid == nullptr) {40 if (qid == nullptr) {
41 return ACL_ERROR_INVALID_PARAM;41 return ACL_ERROR_INVALID_PARAM;
42 }42 }
43 43 
44- acltdtQueueAttr *attr = acltdtCreateQueueAttr();44+ acltdtQueueAttr* attr = acltdtCreateQueueAttr();
45 if (attr == nullptr) {45 if (attr == nullptr) {
46 return ACL_ERROR_INVALID_PARAM;46 return ACL_ERROR_INVALID_PARAM;
47 }47 }
@@ -49,7 +49,7 @@ aclError CreateQueueWithAttr(const char *name, uint32_t depth, uint32_t *qid)
49 aclError ret = acltdtSetQueueAttr(attr, ACL_TDT_QUEUE_DEPTH_UINT32, sizeof(depth), &depth);49 aclError ret = acltdtSetQueueAttr(attr, ACL_TDT_QUEUE_DEPTH_UINT32, sizeof(depth), &depth);
50 if ((ret == ACL_SUCCESS) && (name != nullptr)) {50 if ((ret == ACL_SUCCESS) && (name != nullptr)) {
51 // ACL_TDT_QUEUE_NAME_PTR expects a pointer-sized payload that points to the queue name.51 // ACL_TDT_QUEUE_NAME_PTR expects a pointer-sized payload that points to the queue name.
52- const char *namePtr = name;52+ const char* namePtr = name;
53 ret = acltdtSetQueueAttr(attr, ACL_TDT_QUEUE_NAME_PTR, kQueueNamePtrSize, &namePtr);53 ret = acltdtSetQueueAttr(attr, ACL_TDT_QUEUE_NAME_PTR, kQueueNamePtrSize, &namePtr);
54 }54 }
55 if (ret == ACL_SUCCESS) {55 if (ret == ACL_SUCCESS) {
@@ -60,7 +60,7 @@ aclError CreateQueueWithAttr(const char *name, uint32_t depth, uint32_t *qid)
60 return ret;60 return ret;
61}61}
62 62 
63-aclError CreateQueue(const char *name, uint32_t depth, uint32_t *qid)63+aclError CreateQueue(const char* name, uint32_t depth, uint32_t* qid)
64{64{
65 aclError ret = CreateQueueWithAttr(name, depth, qid);65 aclError ret = CreateQueueWithAttr(name, depth, qid);
66 if ((ret != ACL_ERROR_INVALID_PARAM) || (name == nullptr)) {66 if ((ret != ACL_ERROR_INVALID_PARAM) || (name == nullptr)) {
@@ -75,7 +75,7 @@ aclError CreateQueue(const char *name, uint32_t depth, uint32_t *qid)
75 return CreateQueueWithAttr(nullptr, depth, qid);75 return CreateQueueWithAttr(nullptr, depth, qid);
76}76}
77 77 
78-int32_t CreateQueues(QueueRouteSample &sample)78+int32_t CreateQueues(QueueRouteSample& sample)
79{79{
80 aclError ret = CreateQueue("route_src_queue", QueueRouteSample::kQueueDepth, &sample.srcQid);80 aclError ret = CreateQueue("route_src_queue", QueueRouteSample::kQueueDepth, &sample.srcQid);
81 if (ret != ACL_SUCCESS) {81 if (ret != ACL_SUCCESS) {
@@ -99,7 +99,7 @@ int32_t CreateQueues(QueueRouteSample &sample)
99 return 0;99 return 0;
100}100}
101 101 
102-int32_t BindQueueRoute(QueueRouteSample &sample)102+int32_t BindQueueRoute(QueueRouteSample& sample)
103{103{
104 sample.route = acltdtCreateQueueRoute(sample.srcQid, sample.dstQid);104 sample.route = acltdtCreateQueueRoute(sample.srcQid, sample.dstQid);
105 if (!tdt::CheckNotNull(sample.route, "acltdtCreateQueueRoute")) {105 if (!tdt::CheckNotNull(sample.route, "acltdtCreateQueueRoute")) {
@@ -117,7 +117,7 @@ int32_t BindQueueRoute(QueueRouteSample &sample)
117 return 0;117 return 0;
118}118}
119 119 
120-int32_t PrepareRouteQuery(QueueRouteSample &sample)120+int32_t PrepareRouteQuery(QueueRouteSample& sample)
121{121{
122 sample.queryInfo = acltdtCreateQueueRouteQueryInfo();122 sample.queryInfo = acltdtCreateQueueRouteQueryInfo();
123 if (!tdt::CheckNotNull(sample.queryInfo, "acltdtCreateQueueRouteQueryInfo")) {123 if (!tdt::CheckNotNull(sample.queryInfo, "acltdtCreateQueueRouteQueryInfo")) {
@@ -125,21 +125,12 @@ int32_t PrepareRouteQuery(QueueRouteSample &sample)
125 }125 }
126 126 
127 acltdtQueueRouteQueryMode mode = ACL_TDT_QUEUE_ROUTE_QUERY_SRC_AND_DST;127 acltdtQueueRouteQueryMode mode = ACL_TDT_QUEUE_ROUTE_QUERY_SRC_AND_DST;
128+ CHECK_ERROR(
129+ acltdtSetQueueRouteQueryInfo(sample.queryInfo, ACL_TDT_QUEUE_ROUTE_QUERY_MODE_ENUM, sizeof(mode), &mode));
128 CHECK_ERROR(acltdtSetQueueRouteQueryInfo(130 CHECK_ERROR(acltdtSetQueueRouteQueryInfo(
129- sample.queryInfo,131+ sample.queryInfo, ACL_TDT_QUEUE_ROUTE_QUERY_SRC_ID_UINT32, sizeof(sample.srcQid), &sample.srcQid));
130- ACL_TDT_QUEUE_ROUTE_QUERY_MODE_ENUM,
131- sizeof(mode),
132- &mode));
133 CHECK_ERROR(acltdtSetQueueRouteQueryInfo(132 CHECK_ERROR(acltdtSetQueueRouteQueryInfo(
134- sample.queryInfo,133+ sample.queryInfo, ACL_TDT_QUEUE_ROUTE_QUERY_DST_ID_UINT32, sizeof(sample.dstQid), &sample.dstQid));
135- ACL_TDT_QUEUE_ROUTE_QUERY_SRC_ID_UINT32,
136- sizeof(sample.srcQid),
137- &sample.srcQid));
138- CHECK_ERROR(acltdtSetQueueRouteQueryInfo(
139- sample.queryInfo,
140- ACL_TDT_QUEUE_ROUTE_QUERY_DST_ID_UINT32,
141- sizeof(sample.dstQid),
142- &sample.dstQid));
143 134 
144 sample.queryList = acltdtCreateQueueRouteList();135 sample.queryList = acltdtCreateQueueRouteList();
145 if (!tdt::CheckNotNull(sample.queryList, "acltdtCreateQueueRouteList(query)")) {136 if (!tdt::CheckNotNull(sample.queryList, "acltdtCreateQueueRouteList(query)")) {
@@ -150,11 +141,8 @@ int32_t PrepareRouteQuery(QueueRouteSample &sample)
150}141}
151 142 
152aclError GetQueueRouteParamValue(143aclError GetQueueRouteParamValue(
153- acltdtQueueRoute *route,144+ acltdtQueueRoute* route, decltype(ACL_TDT_QUEUE_ROUTE_SRC_UINT32) paramType, const char* apiName, size_t valueSize,
154- decltype(ACL_TDT_QUEUE_ROUTE_SRC_UINT32) paramType,145+ void* value)
155- const char *apiName,
156- size_t valueSize,
157- void *value)
158{146{
159 size_t paramSize = 0;147 size_t paramSize = 0;
160 const aclError ret = acltdtGetQueueRouteParam(route, paramType, valueSize, &paramSize, value);148 const aclError ret = acltdtGetQueueRouteParam(route, paramType, valueSize, &paramSize, value);
@@ -164,7 +152,7 @@ aclError GetQueueRouteParamValue(
164 return ret;152 return ret;
165}153}
166 154 
167-int32_t LogQueriedRoute(const QueueRouteSample &sample)155+int32_t LogQueriedRoute(const QueueRouteSample& sample)
168{156{
169 const size_t routeNum = acltdtGetQueueRouteNum(sample.queryList);157 const size_t routeNum = acltdtGetQueueRouteNum(sample.queryList);
170 INFO_LOG("Queried route count: %zu", routeNum);158 INFO_LOG("Queried route count: %zu", routeNum);
@@ -172,7 +160,7 @@ int32_t LogQueriedRoute(const QueueRouteSample &sample)
172 return 0;160 return 0;
173 }161 }
174 162 
175- acltdtQueueRoute *routeView = acltdtCreateQueueRoute(sample.srcQid, sample.dstQid);163+ acltdtQueueRoute* routeView = acltdtCreateQueueRoute(sample.srcQid, sample.dstQid);
176 if (!tdt::CheckNotNull(routeView, "routeView")) {164 if (!tdt::CheckNotNull(routeView, "routeView")) {
177 return -1;165 return -1;
178 }166 }
@@ -190,26 +178,20 @@ int32_t LogQueriedRoute(const QueueRouteSample &sample)
190 uint32_t queriedDst = 0;178 uint32_t queriedDst = 0;
191 int32_t queriedStatus = 0;179 int32_t queriedStatus = 0;
192 if (GetQueueRouteParamValue(180 if (GetQueueRouteParamValue(
193- routeView,181+ routeView, ACL_TDT_QUEUE_ROUTE_SRC_UINT32,
194- ACL_TDT_QUEUE_ROUTE_SRC_UINT32,
195 "acltdtGetQueueRouteParam(routeView, ACL_TDT_QUEUE_ROUTE_SRC_UINT32, sizeof(queriedSrc), &paramSize, "182 "acltdtGetQueueRouteParam(routeView, ACL_TDT_QUEUE_ROUTE_SRC_UINT32, sizeof(queriedSrc), &paramSize, "
196 "&queriedSrc)",183 "&queriedSrc)",
197- sizeof(queriedSrc),184+ sizeof(queriedSrc), &queriedSrc) != ACL_SUCCESS ||
198- &queriedSrc) != ACL_SUCCESS ||
199 GetQueueRouteParamValue(185 GetQueueRouteParamValue(
200- routeView,186+ routeView, ACL_TDT_QUEUE_ROUTE_DST_UINT32,
201- ACL_TDT_QUEUE_ROUTE_DST_UINT32,
202 "acltdtGetQueueRouteParam(routeView, ACL_TDT_QUEUE_ROUTE_DST_UINT32, sizeof(queriedDst), &paramSize, "187 "acltdtGetQueueRouteParam(routeView, ACL_TDT_QUEUE_ROUTE_DST_UINT32, sizeof(queriedDst), &paramSize, "
203 "&queriedDst)",188 "&queriedDst)",
204- sizeof(queriedDst),189+ sizeof(queriedDst), &queriedDst) != ACL_SUCCESS ||
205- &queriedDst) != ACL_SUCCESS ||
206 GetQueueRouteParamValue(190 GetQueueRouteParamValue(
207- routeView,191+ routeView, ACL_TDT_QUEUE_ROUTE_STATUS_INT32,
208- ACL_TDT_QUEUE_ROUTE_STATUS_INT32,
209 "acltdtGetQueueRouteParam(routeView, ACL_TDT_QUEUE_ROUTE_STATUS_INT32, sizeof(queriedStatus), "192 "acltdtGetQueueRouteParam(routeView, ACL_TDT_QUEUE_ROUTE_STATUS_INT32, sizeof(queriedStatus), "
210 "&paramSize, &queriedStatus)",193 "&paramSize, &queriedStatus)",
211- sizeof(queriedStatus),194+ sizeof(queriedStatus), &queriedStatus) != ACL_SUCCESS) {
212- &queriedStatus) != ACL_SUCCESS) {
213 (void)acltdtDestroyQueueRoute(routeView);195 (void)acltdtDestroyQueueRoute(routeView);
214 return -1;196 return -1;
215 }197 }
@@ -219,7 +201,7 @@ int32_t LogQueriedRoute(const QueueRouteSample &sample)
219 return 0;201 return 0;
220}202}
221 203 
222-int32_t RunSample(QueueRouteSample &sample)204+int32_t RunSample(QueueRouteSample& sample)
223{205{
224 CHECK_ERROR(aclInit(nullptr));206 CHECK_ERROR(aclInit(nullptr));
225 sample.aclInitialized = true;207 sample.aclInitialized = true;
@@ -238,62 +220,49 @@ int32_t RunSample(QueueRouteSample &sample)
238 return LogQueriedRoute(sample);220 return LogQueriedRoute(sample);
239}221}
240 222 
241-void CleanupRouteResources(QueueRouteSample &sample, int32_t &finalResult)223+void CleanupRouteResources(QueueRouteSample& sample, int32_t& finalResult)
242{224{
243 if (sample.bindAttempted && sample.routeAdded && (sample.bindList != nullptr)) {225 if (sample.bindAttempted && sample.routeAdded && (sample.bindList != nullptr)) {
244 tdt::UpdateFinalResultOnError(226 tdt::UpdateFinalResultOnError(
245- "acltdtUnbindQueueRoutes(sample.bindList)",227+ "acltdtUnbindQueueRoutes(sample.bindList)", acltdtUnbindQueueRoutes(sample.bindList), finalResult);
246- acltdtUnbindQueueRoutes(sample.bindList),
247- finalResult);
248 }228 }
249 if (sample.queryList != nullptr) {229 if (sample.queryList != nullptr) {
250 tdt::UpdateFinalResultOnError(230 tdt::UpdateFinalResultOnError(
251- "acltdtDestroyQueueRouteList(sample.queryList)",231+ "acltdtDestroyQueueRouteList(sample.queryList)", acltdtDestroyQueueRouteList(sample.queryList),
252- acltdtDestroyQueueRouteList(sample.queryList),
253 finalResult);232 finalResult);
254 }233 }
255 if (sample.queryInfo != nullptr) {234 if (sample.queryInfo != nullptr) {
256 tdt::UpdateFinalResultOnError(235 tdt::UpdateFinalResultOnError(
257- "acltdtDestroyQueueRouteQueryInfo(sample.queryInfo)",236+ "acltdtDestroyQueueRouteQueryInfo(sample.queryInfo)", acltdtDestroyQueueRouteQueryInfo(sample.queryInfo),
258- acltdtDestroyQueueRouteQueryInfo(sample.queryInfo),
259 finalResult);237 finalResult);
260 }238 }
261 if (sample.bindList != nullptr) {239 if (sample.bindList != nullptr) {
262 tdt::UpdateFinalResultOnError(240 tdt::UpdateFinalResultOnError(
263- "acltdtDestroyQueueRouteList(sample.bindList)",241+ "acltdtDestroyQueueRouteList(sample.bindList)", acltdtDestroyQueueRouteList(sample.bindList), finalResult);
264- acltdtDestroyQueueRouteList(sample.bindList),
265- finalResult);
266 }242 }
267 if (sample.route != nullptr) {243 if (sample.route != nullptr) {
268 tdt::UpdateFinalResultOnError(244 tdt::UpdateFinalResultOnError(
269- "acltdtDestroyQueueRoute(sample.route)",245+ "acltdtDestroyQueueRoute(sample.route)", acltdtDestroyQueueRoute(sample.route), finalResult);
270- acltdtDestroyQueueRoute(sample.route),
271- finalResult);
272 }246 }
273}247}
274 248 
275-void CleanupQueues(QueueRouteSample &sample, int32_t &finalResult)249+void CleanupQueues(QueueRouteSample& sample, int32_t& finalResult)
276{250{
277 if (sample.dstQueueCreated) {251 if (sample.dstQueueCreated) {
278 tdt::UpdateFinalResultOnError(252 tdt::UpdateFinalResultOnError(
279- "acltdtDestroyQueue(sample.dstQid)",253+ "acltdtDestroyQueue(sample.dstQid)", acltdtDestroyQueue(sample.dstQid), finalResult);
280- acltdtDestroyQueue(sample.dstQid),
281- finalResult);
282 }254 }
283 if (sample.srcQueueCreated) {255 if (sample.srcQueueCreated) {
284 tdt::UpdateFinalResultOnError(256 tdt::UpdateFinalResultOnError(
285- "acltdtDestroyQueue(sample.srcQid)",257+ "acltdtDestroyQueue(sample.srcQid)", acltdtDestroyQueue(sample.srcQid), finalResult);
286- acltdtDestroyQueue(sample.srcQid),
287- finalResult);
288 }258 }
289}259}
290 260 
291-void CleanupAclRuntime(int32_t &finalResult, const QueueRouteSample &sample)261+void CleanupAclRuntime(int32_t& finalResult, const QueueRouteSample& sample)
292{262{
293 if (sample.deviceSet) {263 if (sample.deviceSet) {
294 tdt::UpdateFinalResultOnError(264 tdt::UpdateFinalResultOnError(
295- "aclrtResetDeviceForce(QueueRouteSample::kDeviceId)",265+ "aclrtResetDeviceForce(QueueRouteSample::kDeviceId)", aclrtResetDeviceForce(QueueRouteSample::kDeviceId),
296- aclrtResetDeviceForce(QueueRouteSample::kDeviceId),
297 finalResult);266 finalResult);
298 }267 }
299 if (sample.aclInitialized) {268 if (sample.aclInitialized) {
@@ -301,7 +270,7 @@ void CleanupAclRuntime(int32_t &finalResult, const QueueRouteSample &sample)
301 }270 }
302}271}
303 272 
304-void CleanupSample(QueueRouteSample &sample, int32_t &finalResult)273+void CleanupSample(QueueRouteSample& sample, int32_t& finalResult)
305{274{
306 CleanupRouteResources(sample, finalResult);275 CleanupRouteResources(sample, finalResult);
307 CleanupQueues(sample, finalResult);276 CleanupQueues(sample, finalResult);
Mexample/3_memory_advanced/allocator/0_custom_allocator/main.cpp+89-27
@@ -17,56 +17,118 @@
17#include "utils.h"17#include "utils.h"
18 18 
19namespace {19namespace {
20-struct CustomBlock { void *addr; size_t size; };20+struct CustomBlock {
21-struct CustomAllocatorState { uint32_t allocCount; uint32_t freeCount; };21+ void* addr;
22+ size_t size;
23+};
24+struct CustomAllocatorState {
25+ uint32_t allocCount;
26+ uint32_t freeCount;
27+};
22 28 
23-void *CustomAlloc(aclrtAllocator allocator, size_t size)29+void* CustomAlloc(aclrtAllocator allocator, size_t size)
24{30{
25- auto *state = static_cast<CustomAllocatorState *>(allocator);31+ auto* state = static_cast<CustomAllocatorState*>(allocator);
26- auto *block = static_cast<CustomBlock *>(std::malloc(sizeof(CustomBlock)));32+ auto* block = static_cast<CustomBlock*>(std::malloc(sizeof(CustomBlock)));
27- if (block == nullptr) { return nullptr; }33+ if (block == nullptr) {
34+ return nullptr;
35+ }
28 block->addr = std::malloc(size);36 block->addr = std::malloc(size);
29- if (block->addr == nullptr) { std::free(block); return nullptr; }37+ if (block->addr == nullptr) {
38+ std::free(block);
39+ return nullptr;
40+ }
30 block->size = size;41 block->size = size;
31- if (state != nullptr) { ++state->allocCount; }42+ if (state != nullptr) {
43+ ++state->allocCount;
44+ }
32 INFO_LOG("CustomAlloc size=%zu", size);45 INFO_LOG("CustomAlloc size=%zu", size);
33 return block;46 return block;
34}47}
35 48 
36void CustomFree(aclrtAllocator allocator, aclrtAllocatorBlock block)49void CustomFree(aclrtAllocator allocator, aclrtAllocatorBlock block)
37{50{
38- auto *state = static_cast<CustomAllocatorState *>(allocator);51+ auto* state = static_cast<CustomAllocatorState*>(allocator);
39- auto *typedBlock = static_cast<CustomBlock *>(block);52+ auto* typedBlock = static_cast<CustomBlock*>(block);
40- if (typedBlock != nullptr) { std::free(typedBlock->addr); std::free(typedBlock); }53+ if (typedBlock != nullptr) {
41- if (state != nullptr) { ++state->freeCount; }54+ std::free(typedBlock->addr);
55+ std::free(typedBlock);
56+ }
57+ if (state != nullptr) {
58+ ++state->freeCount;
59+ }
42 INFO_LOG("CustomFree invoked");60 INFO_LOG("CustomFree invoked");
43}61}
44 62 
45-void *CustomAllocAdvise(aclrtAllocator allocator, size_t size, aclrtAllocatorAddr addr)63+void* CustomAllocAdvise(aclrtAllocator allocator, size_t size, aclrtAllocatorAddr addr)
46{64{
47 (void)addr;65 (void)addr;
48 INFO_LOG("CustomAllocAdvise size=%zu", size);66 INFO_LOG("CustomAllocAdvise size=%zu", size);
49 return CustomAlloc(allocator, size);67 return CustomAlloc(allocator, size);
50}68}
51 69 
52-void *CustomGetAddrFromBlock(aclrtAllocatorBlock block)70+void* CustomGetAddrFromBlock(aclrtAllocatorBlock block)
53{71{
54- auto *typedBlock = static_cast<CustomBlock *>(block);72+ auto* typedBlock = static_cast<CustomBlock*>(block);
55 return typedBlock == nullptr ? nullptr : typedBlock->addr;73 return typedBlock == nullptr ? nullptr : typedBlock->addr;
56}74}
57-}75+} // namespace
58-#define CHECK_NOT_NULL(ptr, name) do { if ((ptr) == nullptr) { ERROR_LOG("%s is nullptr", name); return -1; } } while (0)76+#define CHECK_NOT_NULL(ptr, name) \
77+ do { \
78+ if ((ptr) == nullptr) { \
79+ ERROR_LOG("%s is nullptr", name); \
80+ return -1; \
81+ } \
82+ } while (0)
59int main()83int main()
60{84{
61- const int32_t deviceId = 0; aclrtContext context = nullptr; aclrtStream stream = nullptr; CustomAllocatorState state = {0, 0};85+ const int32_t deviceId = 0;
62- CHECK_ERROR(aclInit(nullptr)); CHECK_ERROR(aclrtSetDevice(deviceId)); CHECK_ERROR(aclrtCreateContext(&context, deviceId)); CHECK_ERROR(aclrtCreateStream(&stream));86+ aclrtContext context = nullptr;
63- aclrtAllocatorDesc desc = aclrtAllocatorCreateDesc(); CHECK_NOT_NULL(desc, "aclrtAllocatorCreateDesc");87+ aclrtStream stream = nullptr;
64- CHECK_ERROR(aclrtAllocatorSetObjToDesc(desc, reinterpret_cast<aclrtAllocator>(&state))); CHECK_ERROR(aclrtAllocatorSetAllocFuncToDesc(desc, CustomAlloc)); CHECK_ERROR(aclrtAllocatorSetFreeFuncToDesc(desc, CustomFree)); CHECK_ERROR(aclrtAllocatorSetAllocAdviseFuncToDesc(desc, CustomAllocAdvise)); CHECK_ERROR(aclrtAllocatorSetGetAddrFromBlockFuncToDesc(desc, CustomGetAddrFromBlock));88+ CustomAllocatorState state = {0, 0};
89+ CHECK_ERROR(aclInit(nullptr));
90+ CHECK_ERROR(aclrtSetDevice(deviceId));
91+ CHECK_ERROR(aclrtCreateContext(&context, deviceId));
92+ CHECK_ERROR(aclrtCreateStream(&stream));
93+ aclrtAllocatorDesc desc = aclrtAllocatorCreateDesc();
94+ CHECK_NOT_NULL(desc, "aclrtAllocatorCreateDesc");
95+ CHECK_ERROR(aclrtAllocatorSetObjToDesc(desc, reinterpret_cast<aclrtAllocator>(&state)));
96+ CHECK_ERROR(aclrtAllocatorSetAllocFuncToDesc(desc, CustomAlloc));
97+ CHECK_ERROR(aclrtAllocatorSetFreeFuncToDesc(desc, CustomFree));
98+ CHECK_ERROR(aclrtAllocatorSetAllocAdviseFuncToDesc(desc, CustomAllocAdvise));
99+ CHECK_ERROR(aclrtAllocatorSetGetAddrFromBlockFuncToDesc(desc, CustomGetAddrFromBlock));
65 CHECK_ERROR(aclrtAllocatorRegister(stream, desc));100 CHECK_ERROR(aclrtAllocatorRegister(stream, desc));
66- aclrtAllocatorDesc descOut = nullptr; aclrtAllocator allocatorOut = nullptr; aclrtAllocatorAllocFunc allocFunc = nullptr; aclrtAllocatorFreeFunc freeFunc = nullptr; aclrtAllocatorAllocAdviseFunc allocAdviseFunc = nullptr; aclrtAllocatorGetAddrFromBlockFunc getAddrFunc = nullptr;101+ aclrtAllocatorDesc descOut = nullptr;
67- CHECK_ERROR(aclrtAllocatorGetByStream(stream, &descOut, &allocatorOut, &allocFunc, &freeFunc, &allocAdviseFunc, &getAddrFunc)); CHECK_NOT_NULL(allocFunc, "allocFunc"); CHECK_NOT_NULL(freeFunc, "freeFunc"); CHECK_NOT_NULL(getAddrFunc, "getAddrFunc");102+ aclrtAllocator allocatorOut = nullptr;
68- aclrtAllocatorBlock block = allocFunc(allocatorOut, 256); CHECK_NOT_NULL(block, "Custom block"); void *addr = getAddrFunc(block); CHECK_NOT_NULL(addr, "Custom block addr"); std::fill_n(static_cast<unsigned char *>(addr), 256, static_cast<unsigned char>(0xAB));103+ aclrtAllocatorAllocFunc allocFunc = nullptr;
69- aclrtAllocatorBlock advisedBlock = nullptr; if (allocAdviseFunc != nullptr) { advisedBlock = allocAdviseFunc(allocatorOut, 128, nullptr); if (advisedBlock != nullptr) { freeFunc(allocatorOut, advisedBlock); } }104+ aclrtAllocatorFreeFunc freeFunc = nullptr;
70- freeFunc(allocatorOut, block); INFO_LOG("Allocator desc queried: descOut=%p allocCount=%u freeCount=%u", descOut, state.allocCount, state.freeCount);105+ aclrtAllocatorAllocAdviseFunc allocAdviseFunc = nullptr;
71- CHECK_ERROR(aclrtAllocatorUnregister(stream)); CHECK_ERROR(aclrtAllocatorDestroyDesc(desc)); CHECK_ERROR(aclrtDestroyStream(stream)); CHECK_ERROR(aclrtDestroyContext(context)); CHECK_ERROR(aclrtResetDeviceForce(deviceId)); CHECK_ERROR(aclFinalize()); return 0;106+ aclrtAllocatorGetAddrFromBlockFunc getAddrFunc = nullptr;
107+ CHECK_ERROR(aclrtAllocatorGetByStream(
108+ stream, &descOut, &allocatorOut, &allocFunc, &freeFunc, &allocAdviseFunc, &getAddrFunc));
109+ CHECK_NOT_NULL(allocFunc, "allocFunc");
110+ CHECK_NOT_NULL(freeFunc, "freeFunc");
111+ CHECK_NOT_NULL(getAddrFunc, "getAddrFunc");
112+ aclrtAllocatorBlock block = allocFunc(allocatorOut, 256);
113+ CHECK_NOT_NULL(block, "Custom block");
114+ void* addr = getAddrFunc(block);
115+ CHECK_NOT_NULL(addr, "Custom block addr");
116+ std::fill_n(static_cast<unsigned char*>(addr), 256, static_cast<unsigned char>(0xAB));
117+ aclrtAllocatorBlock advisedBlock = nullptr;
118+ if (allocAdviseFunc != nullptr) {
119+ advisedBlock = allocAdviseFunc(allocatorOut, 128, nullptr);
120+ if (advisedBlock != nullptr) {
121+ freeFunc(allocatorOut, advisedBlock);
122+ }
123+ }
124+ freeFunc(allocatorOut, block);
125+ INFO_LOG(
126+ "Allocator desc queried: descOut=%p allocCount=%u freeCount=%u", descOut, state.allocCount, state.freeCount);
127+ CHECK_ERROR(aclrtAllocatorUnregister(stream));
128+ CHECK_ERROR(aclrtAllocatorDestroyDesc(desc));
129+ CHECK_ERROR(aclrtDestroyStream(stream));
130+ CHECK_ERROR(aclrtDestroyContext(context));
131+ CHECK_ERROR(aclrtResetDeviceForce(deviceId));
132+ CHECK_ERROR(aclFinalize());
133+ return 0;
72}134}
Mexample/3_memory_advanced/managed_memory/0_uvm_allocate/main.cpp+27-36
@@ -9,8 +9,8 @@
9 */9 */
10 10 
11/*11/*
12- * This example leverages the Unified Virtual Memory (UVM) mechanism to allocate memory 12+ * This example leverages the Unified Virtual Memory (UVM) mechanism to allocate memory
13- * for kernel inputs and outputs via UVM allocation APIs, thereby eliminating the need 13+ * for kernel inputs and outputs via UVM allocation APIs, thereby eliminating the need
14 * for explicit data transfers during parameter passing and result write-back.14 * for explicit data transfers during parameter passing and result write-back.
15 */15 */
16 16 
@@ -23,9 +23,9 @@
23 23 
24namespace {24namespace {
25struct KernelBuffers {25struct KernelBuffers {
26- uint8_t *xPtr = nullptr;26+ uint8_t* xPtr = nullptr;
27- uint8_t *yPtr = nullptr;27+ uint8_t* yPtr = nullptr;
28- uint8_t *zPtr = nullptr;28+ uint8_t* zPtr = nullptr;
29};29};
30 30 
31struct RuntimeResources {31struct RuntimeResources {
@@ -38,7 +38,7 @@ struct RuntimeResources {
38 bool binLoaded = false;38 bool binLoaded = false;
39};39};
40 40 
41-void UpdateFinalResultOnError(const char *apiName, aclError ret, int32_t &finalResult)41+void UpdateFinalResultOnError(const char* apiName, aclError ret, int32_t& finalResult)
42{42{
43 if (ret == ACL_SUCCESS) {43 if (ret == ACL_SUCCESS) {
44 return;44 return;
@@ -47,7 +47,7 @@ void UpdateFinalResultOnError(const char *apiName, aclError ret, int32_t &finalR
47 finalResult = -1;47 finalResult = -1;
48}48}
49 49 
50-int32_t InitializeRuntime(RuntimeResources *runtime)50+int32_t InitializeRuntime(RuntimeResources* runtime)
51{51{
52 // Initialize ACL and create a stream on device 0.52 // Initialize ACL and create a stream on device 0.
53 CHECK_ERROR(aclInit(nullptr));53 CHECK_ERROR(aclInit(nullptr));
@@ -59,16 +59,19 @@ int32_t InitializeRuntime(RuntimeResources *runtime)
59 return 0;59 return 0;
60}60}
61 61 
62-int32_t AllocateKernelBuffers(size_t inputByteSize, size_t outputByteSize, KernelBuffers *buffers)62+int32_t AllocateKernelBuffers(size_t inputByteSize, size_t outputByteSize, KernelBuffers* buffers)
63{63{
64 // Allocate uvm memory for kernel inputs and output.64 // Allocate uvm memory for kernel inputs and output.
65- CHECK_ERROR(aclrtMemAllocManaged(reinterpret_cast<void **>(&buffers->xPtr), inputByteSize, ACL_RT_MEM_ATTACH_GLOBAL));65+ CHECK_ERROR(
66- CHECK_ERROR(aclrtMemAllocManaged(reinterpret_cast<void **>(&buffers->yPtr), inputByteSize, ACL_RT_MEM_ATTACH_GLOBAL));66+ aclrtMemAllocManaged(reinterpret_cast<void**>(&buffers->xPtr), inputByteSize, ACL_RT_MEM_ATTACH_GLOBAL));
67- CHECK_ERROR(aclrtMemAllocManaged(reinterpret_cast<void **>(&buffers->zPtr), outputByteSize, ACL_RT_MEM_ATTACH_GLOBAL));67+ CHECK_ERROR(
68+ aclrtMemAllocManaged(reinterpret_cast<void**>(&buffers->yPtr), inputByteSize, ACL_RT_MEM_ATTACH_GLOBAL));
69+ CHECK_ERROR(
70+ aclrtMemAllocManaged(reinterpret_cast<void**>(&buffers->zPtr), outputByteSize, ACL_RT_MEM_ATTACH_GLOBAL));
68 return 0;71 return 0;
69}72}
70 73 
71-int32_t PrepareInputData(size_t inputByteSize, const KernelBuffers &buffers)74+int32_t PrepareInputData(size_t inputByteSize, const KernelBuffers& buffers)
72{75{
73 // Load generated input files.76 // Load generated input files.
74 size_t xFileSize = inputByteSize;77 size_t xFileSize = inputByteSize;
@@ -87,27 +90,23 @@ int32_t PrepareInputData(size_t inputByteSize, const KernelBuffers &buffers)
87 return 0;90 return 0;
88}91}
89 92 
90-int32_t AppendCommonKernelArgs(aclrtArgsHandle argsHandle, uint8_t *xPtr, uint8_t *yPtr, uint8_t *zPtr)93+int32_t AppendCommonKernelArgs(aclrtArgsHandle argsHandle, uint8_t* xPtr, uint8_t* yPtr, uint8_t* zPtr)
91{94{
92 aclrtParamHandle paramHandle1 = nullptr;95 aclrtParamHandle paramHandle1 = nullptr;
93 aclrtParamHandle paramHandle2 = nullptr;96 aclrtParamHandle paramHandle2 = nullptr;
94 aclrtParamHandle paramHandle3 = nullptr;97 aclrtParamHandle paramHandle3 = nullptr;
95- CHECK_ERROR(aclrtKernelArgsAppend(argsHandle, reinterpret_cast<void **>(&xPtr), sizeof(uintptr_t), &paramHandle1));98+ CHECK_ERROR(aclrtKernelArgsAppend(argsHandle, reinterpret_cast<void**>(&xPtr), sizeof(uintptr_t), &paramHandle1));
96- CHECK_ERROR(aclrtKernelArgsAppend(argsHandle, reinterpret_cast<void **>(&yPtr), sizeof(uintptr_t), &paramHandle2));99+ CHECK_ERROR(aclrtKernelArgsAppend(argsHandle, reinterpret_cast<void**>(&yPtr), sizeof(uintptr_t), &paramHandle2));
97- CHECK_ERROR(aclrtKernelArgsAppend(argsHandle, reinterpret_cast<void **>(&zPtr), sizeof(uintptr_t), &paramHandle3));100+ CHECK_ERROR(aclrtKernelArgsAppend(argsHandle, reinterpret_cast<void**>(&zPtr), sizeof(uintptr_t), &paramHandle3));
98 return 0;101 return 0;
99}102}
100 103 
101int32_t BuildKernelArgs(104int32_t BuildKernelArgs(
102- uint8_t *xPtr,105+ uint8_t* xPtr, uint8_t* yPtr, uint8_t* zPtr, RuntimeResources* runtime, aclrtFuncHandle* funcHandle,
103- uint8_t *yPtr,106+ aclrtArgsHandle* argsHandle)
104- uint8_t *zPtr,
105- RuntimeResources *runtime,
106- aclrtFuncHandle *funcHandle,
107- aclrtArgsHandle *argsHandle)
108{107{
109 // Load the selected kernel binary and build the launch argument list.108 // Load the selected kernel binary and build the launch argument list.
110- const char *filePath = "./out/fatbin/ascendc_kernels_simple/ascendc_kernels_simple.o";109+ const char* filePath = "./out/fatbin/ascendc_kernels_simple/ascendc_kernels_simple.o";
111 110 
112 CHECK_ERROR(aclrtBinaryLoadFromFile(filePath, nullptr, &runtime->binHandle));111 CHECK_ERROR(aclrtBinaryLoadFromFile(filePath, nullptr, &runtime->binHandle));
113 runtime->binLoaded = true;112 runtime->binLoaded = true;
@@ -123,11 +122,7 @@ int32_t BuildKernelArgs(
123}122}
124 123 
125int32_t LaunchKernelAndWriteOutput(124int32_t LaunchKernelAndWriteOutput(
126- aclrtFuncHandle funcHandle,125+ aclrtFuncHandle funcHandle, aclrtArgsHandle argsHandle, uint32_t blockDim, aclrtStream stream, uint8_t* zPtr,
127- aclrtArgsHandle argsHandle,
128- uint32_t blockDim,
129- aclrtStream stream,
130- uint8_t *zPtr,
131 size_t outputByteSize)126 size_t outputByteSize)
132{127{
133 // Launch the kernel, synchronize the stream, and write output for verification.128 // Launch the kernel, synchronize the stream, and write output for verification.
@@ -140,7 +135,7 @@ int32_t LaunchKernelAndWriteOutput(
140 return 0;135 return 0;
141}136}
142 137 
143-void ReleaseKernelResources(RuntimeResources &runtime, KernelBuffers &buffers, int32_t &finalResult)138+void ReleaseKernelResources(RuntimeResources& runtime, KernelBuffers& buffers, int32_t& finalResult)
144{139{
145 if (runtime.binLoaded) {140 if (runtime.binLoaded) {
146 UpdateFinalResultOnError(141 UpdateFinalResultOnError(
@@ -191,12 +186,11 @@ int32_t RunKernelLaunchSample()
191 if (PrepareInputData(inputByteSize, buffers) != 0) {186 if (PrepareInputData(inputByteSize, buffers) != 0) {
192 return -1;187 return -1;
193 }188 }
194- if (BuildKernelArgs(189+ if (BuildKernelArgs(buffers.xPtr, buffers.yPtr, buffers.zPtr, &runtime, &funcHandle, &argsHandle) != 0) {
195- buffers.xPtr, buffers.yPtr, buffers.zPtr, &runtime, &funcHandle, &argsHandle) != 0) {
196 return -1;190 return -1;
197 }191 }
198 if (LaunchKernelAndWriteOutput(192 if (LaunchKernelAndWriteOutput(
199- funcHandle, argsHandle, blockDim, runtime.stream, buffers.zPtr, outputByteSize) != 0) {193+ funcHandle, argsHandle, blockDim, runtime.stream, buffers.zPtr, outputByteSize) != 0) {
200 return -1;194 return -1;
201 }195 }
202 return 0;196 return 0;
@@ -211,7 +205,4 @@ int32_t RunKernelLaunchSample()
211}205}
212} // namespace206} // namespace
213 207 
214-int32_t main(int32_t argc, char *argv[])208+int32_t main(int32_t argc, char* argv[]) { return RunKernelLaunchSample(); }
215-{
216- return RunKernelLaunchSample();
217-}
Mexample/3_memory_advanced/managed_memory/file_ops.cpp+4-5
@@ -17,7 +17,7 @@
17#include "utils.h"17#include "utils.h"
18 18 
19namespace kernel {19namespace kernel {
20-bool ReadFile(const std::string &filePath, size_t &fileSize, void *buffer, size_t bufferSize)20+bool ReadFile(const std::string& filePath, size_t& fileSize, void* buffer, size_t bufferSize)
21{21{
22 // Read data from the target file22 // Read data from the target file
23 struct stat sBuf;23 struct stat sBuf;
@@ -38,7 +38,7 @@ bool ReadFile(const std::string &filePath, size_t &fileSize, void *buffer, size_
38 return false;38 return false;
39 }39 }
40 40 
41- std::filebuf *buf = file.rdbuf();41+ std::filebuf* buf = file.rdbuf();
42 size_t size = buf->pubseekoff(0, std::ios::end, std::ios::in);42 size_t size = buf->pubseekoff(0, std::ios::end, std::ios::in);
43 if (size == 0U) {43 if (size == 0U) {
44 ERROR_LOG("file size is 0");44 ERROR_LOG("file size is 0");
@@ -51,16 +51,15 @@ bool ReadFile(const std::string &filePath, size_t &fileSize, void *buffer, size_
51 return false;51 return false;
52 }52 }
53 buf->pubseekpos(0, std::ios::in);53 buf->pubseekpos(0, std::ios::in);
54- (void)buf->sgetn(static_cast<char *>(buffer), static_cast<std::streamsize>(size));54+ (void)buf->sgetn(static_cast<char*>(buffer), static_cast<std::streamsize>(size));
55 fileSize = size;55 fileSize = size;
56 file.close();56 file.close();
57 return true;57 return true;
58}58}
59} // namespace kernel59} // namespace kernel
60 60 
61- 
62namespace kernel {61namespace kernel {
63-bool WriteFile(const std::string &filePath, const void *buffer, size_t size)62+bool WriteFile(const std::string& filePath, const void* buffer, size_t size)
64{63{
65 // Write data to the target file64 // Write data to the target file
66 if (buffer == nullptr) {65 if (buffer == nullptr) {
Mexample/3_memory_advanced/managed_memory/kernel_utils.h+3-3
@@ -14,7 +14,7 @@
14#include <cstddef>14#include <cstddef>
15 15 
16namespace kernel {16namespace kernel {
17- bool ReadFile(const std::string &filePath, size_t &fileSize, void *buffer, size_t bufferSize);17+bool ReadFile(const std::string& filePath, size_t& fileSize, void* buffer, size_t bufferSize);
18- bool WriteFile(const std::string &filePath, const void *buffer, size_t size);18+bool WriteFile(const std::string& filePath, const void* buffer, size_t size);
19-}19+} // namespace kernel
20#endif20#endif
Mexample/4_reliability/overflow_detection/0_overflow_detection/main.cpp+17-31
@@ -17,14 +17,14 @@
17namespace {17namespace {
18constexpr size_t kOverflowStatusBufferSize = 64;18constexpr size_t kOverflowStatusBufferSize = 64;
19 19 
20-uint32_t ReadOverflowFlag(const uint8_t *statusBuffer)20+uint32_t ReadOverflowFlag(const uint8_t* statusBuffer)
21{21{
22 uint32_t overflowFlag = 0;22 uint32_t overflowFlag = 0;
23- std::copy_n(statusBuffer, sizeof(overflowFlag), reinterpret_cast<unsigned char *>(&overflowFlag));23+ std::copy_n(statusBuffer, sizeof(overflowFlag), reinterpret_cast<unsigned char*>(&overflowFlag));
24 return overflowFlag;24 return overflowFlag;
25}25}
26 26 
27-const char *OverflowModeToString(aclrtFloatOverflowMode mode)27+const char* OverflowModeToString(aclrtFloatOverflowMode mode)
28{28{
29 switch (mode) {29 switch (mode) {
30 case ACL_RT_OVERFLOW_MODE_SATURATION:30 case ACL_RT_OVERFLOW_MODE_SATURATION:
@@ -38,17 +38,14 @@ const char *OverflowModeToString(aclrtFloatOverflowMode mode)
38 }38 }
39}39}
40 40 
41-int32_t HandleOptionalOverflowRet(const char *apiName, aclError ret, const char *reason)41+int32_t HandleOptionalOverflowRet(const char* apiName, aclError ret, const char* reason)
42{42{
43 if (ret == ACL_SUCCESS) {43 if (ret == ACL_SUCCESS) {
44 return 0;44 return 0;
45 }45 }
46 if (ret == ACL_ERROR_RT_FEATURE_NOT_SUPPORT) {46 if (ret == ACL_ERROR_RT_FEATURE_NOT_SUPPORT) {
47 WARN_LOG(47 WARN_LOG(
48- "%s is unavailable in the current environment (ret=%d): %s",48+ "%s is unavailable in the current environment (ret=%d): %s", apiName, static_cast<int32_t>(ret), reason);
49- apiName,
50- static_cast<int32_t>(ret),
51- reason);
52 return 1;49 return 1;
53 }50 }
54 ERROR_LOG("Operation failed: %s returned error code %d", apiName, static_cast<int32_t>(ret));51 ERROR_LOG("Operation failed: %s returned error code %d", apiName, static_cast<int32_t>(ret));
@@ -61,7 +58,7 @@ int main()
61 const int32_t deviceId = 0;58 const int32_t deviceId = 0;
62 aclrtContext context = nullptr;59 aclrtContext context = nullptr;
63 aclrtStream stream = nullptr;60 aclrtStream stream = nullptr;
64- uint8_t *statusDevice = nullptr;61+ uint8_t* statusDevice = nullptr;
65 aclrtFloatOverflowMode originalMode = ACL_RT_OVERFLOW_MODE_UNDEF;62 aclrtFloatOverflowMode originalMode = ACL_RT_OVERFLOW_MODE_UNDEF;
66 aclrtFloatOverflowMode currentMode = ACL_RT_OVERFLOW_MODE_UNDEF;63 aclrtFloatOverflowMode currentMode = ACL_RT_OVERFLOW_MODE_UNDEF;
67 bool aclInitialized = false;64 bool aclInitialized = false;
@@ -82,8 +79,7 @@ int main()
82 contextCreated = true;79 contextCreated = true;
83 80 
84 const int32_t getModeStatus = HandleOptionalOverflowRet(81 const int32_t getModeStatus = HandleOptionalOverflowRet(
85- "aclrtGetDeviceSatMode(&originalMode)",82+ "aclrtGetDeviceSatMode(&originalMode)", aclrtGetDeviceSatMode(&originalMode),
86- aclrtGetDeviceSatMode(&originalMode),
87 "the current device/runtime does not expose device saturation mode");83 "the current device/runtime does not expose device saturation mode");
88 if (getModeStatus < 0) {84 if (getModeStatus < 0) {
89 return -1;85 return -1;
@@ -96,7 +92,8 @@ int main()
96 const int32_t setModeStatus = HandleOptionalOverflowRet(92 const int32_t setModeStatus = HandleOptionalOverflowRet(
97 "aclrtSetDeviceSatMode(ACL_RT_OVERFLOW_MODE_SATURATION)",93 "aclrtSetDeviceSatMode(ACL_RT_OVERFLOW_MODE_SATURATION)",
98 aclrtSetDeviceSatMode(ACL_RT_OVERFLOW_MODE_SATURATION),94 aclrtSetDeviceSatMode(ACL_RT_OVERFLOW_MODE_SATURATION),
99- "stream overflow detection requires saturation mode and is only supported on specific products/runtime builds");95+ "stream overflow detection requires saturation mode and is only supported on specific products/runtime "
96+ "builds");
100 if (setModeStatus < 0) {97 if (setModeStatus < 0) {
101 return -1;98 return -1;
102 }99 }
@@ -108,16 +105,14 @@ int main()
108 105 
109 CHECK_ERROR(aclrtGetDeviceSatMode(&currentMode));106 CHECK_ERROR(aclrtGetDeviceSatMode(&currentMode));
110 INFO_LOG(107 INFO_LOG(
111- "Device saturation mode switched from %s to %s.",108+ "Device saturation mode switched from %s to %s.", OverflowModeToString(originalMode),
112- OverflowModeToString(originalMode),
113 OverflowModeToString(currentMode));109 OverflowModeToString(currentMode));
114 110 
115 CHECK_ERROR(aclrtCreateStream(&stream));111 CHECK_ERROR(aclrtCreateStream(&stream));
116 streamCreated = true;112 streamCreated = true;
117 113 
118 const int32_t setOverflowSwitchStatus = HandleOptionalOverflowRet(114 const int32_t setOverflowSwitchStatus = HandleOptionalOverflowRet(
119- "aclrtSetStreamOverflowSwitch(stream, 1)",115+ "aclrtSetStreamOverflowSwitch(stream, 1)", aclrtSetStreamOverflowSwitch(stream, 1),
120- aclrtSetStreamOverflowSwitch(stream, 1),
121 "the current environment does not expose stream-level overflow detection even in saturation mode");116 "the current environment does not expose stream-level overflow detection even in saturation mode");
122 if (setOverflowSwitchStatus < 0) {117 if (setOverflowSwitchStatus < 0) {
123 return -1;118 return -1;
@@ -132,10 +127,8 @@ int main()
132 CHECK_ERROR(aclrtGetStreamOverflowSwitch(stream, &queriedSwitch));127 CHECK_ERROR(aclrtGetStreamOverflowSwitch(stream, &queriedSwitch));
133 INFO_LOG("Overflow switch=%u", queriedSwitch);128 INFO_LOG("Overflow switch=%u", queriedSwitch);
134 129 
135- CHECK_ERROR(aclrtMalloc(130+ CHECK_ERROR(
136- reinterpret_cast<void **>(&statusDevice),131+ aclrtMalloc(reinterpret_cast<void**>(&statusDevice), kOverflowStatusBufferSize, ACL_MEM_MALLOC_HUGE_FIRST));
137- kOverflowStatusBufferSize,
138- ACL_MEM_MALLOC_HUGE_FIRST));
139 statusAllocated = true;132 statusAllocated = true;
140 133 
141 uint8_t statusHost[kOverflowStatusBufferSize] = {};134 uint8_t statusHost[kOverflowStatusBufferSize] = {};
@@ -144,11 +137,7 @@ int main()
144 CHECK_ERROR(aclrtGetOverflowStatus(statusDevice, kOverflowStatusBufferSize, stream));137 CHECK_ERROR(aclrtGetOverflowStatus(statusDevice, kOverflowStatusBufferSize, stream));
145 CHECK_ERROR(aclrtSynchronizeStream(stream));138 CHECK_ERROR(aclrtSynchronizeStream(stream));
146 CHECK_ERROR(aclrtMemcpy(139 CHECK_ERROR(aclrtMemcpy(
147- statusHost,140+ statusHost, sizeof(statusHost), statusDevice, kOverflowStatusBufferSize, ACL_MEMCPY_DEVICE_TO_HOST));
148- sizeof(statusHost),
149- statusDevice,
150- kOverflowStatusBufferSize,
151- ACL_MEMCPY_DEVICE_TO_HOST));
152 overflowFlag = ReadOverflowFlag(statusHost);141 overflowFlag = ReadOverflowFlag(statusHost);
153 INFO_LOG("Overflow status before reset=%u", overflowFlag);142 INFO_LOG("Overflow status before reset=%u", overflowFlag);
154 143 
@@ -160,11 +149,7 @@ int main()
160 CHECK_ERROR(aclrtGetOverflowStatus(statusDevice, kOverflowStatusBufferSize, stream));149 CHECK_ERROR(aclrtGetOverflowStatus(statusDevice, kOverflowStatusBufferSize, stream));
161 CHECK_ERROR(aclrtSynchronizeStream(stream));150 CHECK_ERROR(aclrtSynchronizeStream(stream));
162 CHECK_ERROR(aclrtMemcpy(151 CHECK_ERROR(aclrtMemcpy(
163- statusHost,152+ statusHost, sizeof(statusHost), statusDevice, kOverflowStatusBufferSize, ACL_MEMCPY_DEVICE_TO_HOST));
164- sizeof(statusHost),
165- statusDevice,
166- kOverflowStatusBufferSize,
167- ACL_MEMCPY_DEVICE_TO_HOST));
168 overflowFlag = ReadOverflowFlag(statusHost);153 overflowFlag = ReadOverflowFlag(statusHost);
169 INFO_LOG("Overflow status after reset=%u", overflowFlag);154 INFO_LOG("Overflow status after reset=%u", overflowFlag);
170 return 0;155 return 0;
@@ -173,7 +158,8 @@ int main()
173 if (statusAllocated) {158 if (statusAllocated) {
174 const aclError freeRet = aclrtFree(statusDevice);159 const aclError freeRet = aclrtFree(statusDevice);
175 if (freeRet != ACL_SUCCESS) {160 if (freeRet != ACL_SUCCESS) {
176- ERROR_LOG("Operation failed: aclrtFree(statusDevice) returned error code %d", static_cast<int32_t>(freeRet));161+ ERROR_LOG(
162+ "Operation failed: aclrtFree(statusDevice) returned error code %d", static_cast<int32_t>(freeRet));
177 finalResult = -1;163 finalResult = -1;
178 }164 }
179 }165 }
Mexample/5_performance/adump/0_adump_args/main.cpp+27-24
@@ -18,18 +18,18 @@
18using namespace std;18using namespace std;
19 19 
20namespace {20namespace {
21- void LogDumpPath(acldumpType dumpType, const char *fallbackPath)21+void LogDumpPath(acldumpType dumpType, const char* fallbackPath)
22- {22+{
23- const char *dumpPath = acldumpGetPath(dumpType);23+ const char* dumpPath = acldumpGetPath(dumpType);
24- if (dumpPath != nullptr) {24+ if (dumpPath != nullptr) {
25- INFO_LOG("acldumpGetPath returned dump path: %s", dumpPath);25+ INFO_LOG("acldumpGetPath returned dump path: %s", dumpPath);
26- return;26+ return;
27- }
28- WARN_LOG("acldumpGetPath returned null, fallback dump path is %s", fallbackPath);
29 }27 }
30- 28+ WARN_LOG("acldumpGetPath returned null, fallback dump path is %s", fallbackPath);
31}29}
32 30 
31+} // namespace
32+ 
33int main()33int main()
34{34{
35 // The device id35 // The device id
@@ -51,13 +51,13 @@ int main()
51 std::vector<int64_t> selfShape{4, 2};51 std::vector<int64_t> selfShape{4, 2};
52 std::vector<int64_t> otherShape{4, 2};52 std::vector<int64_t> otherShape{4, 2};
53 std::vector<int64_t> outShape{4, 2};53 std::vector<int64_t> outShape{4, 2};
54- void *selfDeviceAddr = nullptr;54+ void* selfDeviceAddr = nullptr;
55- void *otherDeviceAddr = nullptr;55+ void* otherDeviceAddr = nullptr;
56- void *outDeviceAddr = nullptr;56+ void* outDeviceAddr = nullptr;
57- aclTensor *self = nullptr;57+ aclTensor* self = nullptr;
58- aclTensor *other = nullptr;58+ aclTensor* other = nullptr;
59- aclScalar *alpha = nullptr;59+ aclScalar* alpha = nullptr;
60- aclTensor *out = nullptr;60+ aclTensor* out = nullptr;
61 std::vector<float> selfHostData = {0, 1, 2, 3, 4, 5, 6, 7};61 std::vector<float> selfHostData = {0, 1, 2, 3, 4, 5, 6, 7};
62 std::vector<float> otherHostData = {1, 1, 1, 2, 2, 2, 3, 3};62 std::vector<float> otherHostData = {1, 1, 1, 2, 2, 2, 3, 3};
63 std::vector<float> outHostData = {0, 0, 0, 0, 0, 0, 0, 0};63 std::vector<float> outHostData = {0, 0, 0, 0, 0, 0, 0, 0};
@@ -77,11 +77,11 @@ int main()
77 77 
78 // 3. Call the CANN operator library API(Custom Implementation)78 // 3. Call the CANN operator library API(Custom Implementation)
79 uint64_t workspaceSize = 0;79 uint64_t workspaceSize = 0;
80- aclOpExecutor *executor;80+ aclOpExecutor* executor;
81 CHECK_ERROR(aclnnAddGetWorkspaceSize(self, other, alpha, out, &workspaceSize, &executor));81 CHECK_ERROR(aclnnAddGetWorkspaceSize(self, other, alpha, out, &workspaceSize, &executor));
82 82 
83 // Allocate device memory based on the calculation results83 // Allocate device memory based on the calculation results
84- void *workspaceAddr = nullptr;84+ void* workspaceAddr = nullptr;
85 if (workspaceSize > 0lu) {85 if (workspaceSize > 0lu) {
86 CHECK_ERROR(aclrtMalloc(&workspaceAddr, workspaceSize, ACL_MEM_MALLOC_HUGE_FIRST));86 CHECK_ERROR(aclrtMalloc(&workspaceAddr, workspaceSize, ACL_MEM_MALLOC_HUGE_FIRST));
87 }87 }
@@ -94,11 +94,12 @@ int main()
94 // 5. Obtain the execution result of the operator and copy the result from the device memory to the host94 // 5. Obtain the execution result of the operator and copy the result from the device memory to the host
95 auto size = adump::GetShapeSize(outShape);95 auto size = adump::GetShapeSize(outShape);
96 std::vector<float> resultData(size, 0);96 std::vector<float> resultData(size, 0);
97- CHECK_ERROR(aclrtMemcpy(resultData.data(), resultData.size() * sizeof(resultData[0]), outDeviceAddr,97+ CHECK_ERROR(aclrtMemcpy(
98- size * sizeof(float), ACL_MEMCPY_DEVICE_TO_HOST));98+ resultData.data(), resultData.size() * sizeof(resultData[0]), outDeviceAddr, size * sizeof(float),
99+ ACL_MEMCPY_DEVICE_TO_HOST));
99 100 
100 for (int64_t i = 0; i < size; i++) {101 for (int64_t i = 0; i < size; i++) {
101- INFO_LOG("result[%ld] is: %f", i, resultData[i]);102+ INFO_LOG("result[%ld] is: %f", i, resultData[i]);
102 }103 }
103 104 
104 // 6. Release the resources(Custom Destruction)105 // 6. Release the resources(Custom Destruction)
@@ -109,7 +110,7 @@ int main()
109 CHECK_ERROR(aclrtFree(otherDeviceAddr));110 CHECK_ERROR(aclrtFree(otherDeviceAddr));
110 CHECK_ERROR(aclrtFree(outDeviceAddr));111 CHECK_ERROR(aclrtFree(outDeviceAddr));
111 if (workspaceSize > 0lu) {112 if (workspaceSize > 0lu) {
112- CHECK_ERROR(aclrtFree(workspaceAddr));113+ CHECK_ERROR(aclrtFree(workspaceAddr));
113 }114 }
114 // 7. AsendCL Destroy115 // 7. AsendCL Destroy
115 // Disable Dump Args Function116 // Disable Dump Args Function
@@ -118,7 +119,9 @@ int main()
118 CHECK_ERROR(aclrtDestroyStream(stream));119 CHECK_ERROR(aclrtDestroyStream(stream));
119 CHECK_ERROR(aclrtResetDeviceForce(deviceId));120 CHECK_ERROR(aclrtResetDeviceForce(deviceId));
120 CHECK_ERROR(aclFinalize());121 CHECK_ERROR(aclFinalize());
121- INFO_LOG("Run the device_normal sample successfully. "122+ INFO_LOG(
122- "please make sure dump data has been in path: %s", dumpPath);123+ "Run the device_normal sample successfully. "
124+ "please make sure dump data has been in path: %s",
125+ dumpPath);
123 return 0;126 return 0;
124}127}
Mexample/5_performance/adump/1_adump_callback/main.cpp+71-68
@@ -24,71 +24,73 @@
24using namespace std;24using namespace std;
25 25 
26namespace {26namespace {
27- std::mutex gCallbackMutex;27+std::mutex gCallbackMutex;
28- std::map<std::string, int32_t> gChunkCountByFile;28+std::map<std::string, int32_t> gChunkCountByFile;
29- int32_t gTotalChunkCount = 0;29+int32_t gTotalChunkCount = 0;
30- int64_t gTotalBytes = 0;30+int64_t gTotalBytes = 0;
31 31 
32- void LogDumpPath(acldumpType dumpType, const char *fallbackPath)32+void LogDumpPath(acldumpType dumpType, const char* fallbackPath)
33- {33+{
34- const char *dumpPath = acldumpGetPath(dumpType);34+ const char* dumpPath = acldumpGetPath(dumpType);
35- if (dumpPath != nullptr) {35+ if (dumpPath != nullptr) {
36- INFO_LOG("acldumpGetPath returned dump path: %s", dumpPath);36+ INFO_LOG("acldumpGetPath returned dump path: %s", dumpPath);
37- return;37+ return;
38- }
39- WARN_LOG("acldumpGetPath returned null, fallback dump path is %s", fallbackPath);
40 }38 }
39+ WARN_LOG("acldumpGetPath returned null, fallback dump path is %s", fallbackPath);
40+}
41 41 
42- std::string BuildChunkPreview(const acldumpChunk *data)42+std::string BuildChunkPreview(const acldumpChunk* data)
43- {43+{
44- const uint32_t previewLen = std::min<uint32_t>(data->bufLen, 16U);44+ const uint32_t previewLen = std::min<uint32_t>(data->bufLen, 16U);
45- std::ostringstream preview;45+ std::ostringstream preview;
46- preview << std::hex << std::setfill('0');46+ preview << std::hex << std::setfill('0');
47- for (uint32_t i = 0; i < previewLen; ++i) {47+ for (uint32_t i = 0; i < previewLen; ++i) {
48- if (i != 0) {48+ if (i != 0) {
49- preview << " ";49+ preview << " ";
50- }
51- preview << std::setw(2) << static_cast<uint32_t>(data->dataBuf[i]);
52 }50 }
53- if (data->bufLen > previewLen) {51+ preview << std::setw(2) << static_cast<uint32_t>(data->dataBuf[i]);
54- preview << " ...";
55- }
56- return preview.str();
57 }52 }
58- 53+ if (data->bufLen > previewLen) {
59- void PrintCallbackSummary()54+ preview << " ...";
60- {
61- std::lock_guard<std::mutex> lock(gCallbackMutex);
62- INFO_LOG("Dump callback summary: total chunks=%d, total bytes=%ld, total files=%d",
63- gTotalChunkCount, gTotalBytes, static_cast<int32_t>(gChunkCountByFile.size()));
64- for (const auto &entry : gChunkCountByFile) {
65- INFO_LOG(" file=%s, chunks=%d", entry.first.c_str(), entry.second);
66- }
67 }55 }
56+ return preview.str();
57+}
68 58 
69- int32_t DumpTensorCallback(const acldumpChunk *data, int32_t len)59+void PrintCallbackSummary()
70- {60+{
71- if (data == nullptr) {61+ std::lock_guard<std::mutex> lock(gCallbackMutex);
72- ERROR_LOG("Callback data is null!");62+ INFO_LOG(
73- return -1;63+ "Dump callback summary: total chunks=%d, total bytes=%ld, total files=%d", gTotalChunkCount, gTotalBytes,
74- }64+ static_cast<int32_t>(gChunkCountByFile.size()));
75- if ((sizeof(acldumpChunk) + data->bufLen) != len) {65+ for (const auto& entry : gChunkCountByFile) {
76- ERROR_LOG("Callback data is invalid. bufLen: %d, callback len: %d", data->bufLen, len);66+ INFO_LOG(" file=%s, chunks=%d", entry.first.c_str(), entry.second);
77- return -1;
78- }
79- const std::string preview = BuildChunkPreview(data);
80- {
81- std::lock_guard<std::mutex> lock(gCallbackMutex);
82- ++gTotalChunkCount;
83- gTotalBytes += data->bufLen;
84- ++gChunkCountByFile[data->fileName];
85- }
86- INFO_LOG("Receive dump tensor data success. file=%s, bufLen=%u, isLastChunk=%u, offset=%ld, flag=%d, preview=%s",
87- data->fileName, data->bufLen, data->isLastChunk, data->offset, data->flag, preview.c_str());
88- return 0;
89 }67 }
90}68}
91 69 
70+int32_t DumpTensorCallback(const acldumpChunk* data, int32_t len)
71+{
72+ if (data == nullptr) {
73+ ERROR_LOG("Callback data is null!");
74+ return -1;
75+ }
76+ if ((sizeof(acldumpChunk) + data->bufLen) != len) {
77+ ERROR_LOG("Callback data is invalid. bufLen: %d, callback len: %d", data->bufLen, len);
78+ return -1;
79+ }
80+ const std::string preview = BuildChunkPreview(data);
81+ {
82+ std::lock_guard<std::mutex> lock(gCallbackMutex);
83+ ++gTotalChunkCount;
84+ gTotalBytes += data->bufLen;
85+ ++gChunkCountByFile[data->fileName];
86+ }
87+ INFO_LOG(
88+ "Receive dump tensor data success. file=%s, bufLen=%u, isLastChunk=%u, offset=%ld, flag=%d, preview=%s",
89+ data->fileName, data->bufLen, data->isLastChunk, data->offset, data->flag, preview.c_str());
90+ return 0;
91+}
92+} // namespace
93+ 
92int main()94int main()
93{95{
94 // The device id96 // The device id
@@ -107,13 +109,13 @@ int main()
107 std::vector<int64_t> selfShape{4, 2};109 std::vector<int64_t> selfShape{4, 2};
108 std::vector<int64_t> otherShape{4, 2};110 std::vector<int64_t> otherShape{4, 2};
109 std::vector<int64_t> outShape{4, 2};111 std::vector<int64_t> outShape{4, 2};
110- void *selfDeviceAddr = nullptr;112+ void* selfDeviceAddr = nullptr;
111- void *otherDeviceAddr = nullptr;113+ void* otherDeviceAddr = nullptr;
112- void *outDeviceAddr = nullptr;114+ void* outDeviceAddr = nullptr;
113- aclTensor *self = nullptr;115+ aclTensor* self = nullptr;
114- aclTensor *out = nullptr;116+ aclTensor* out = nullptr;
115- aclTensor *other = nullptr;117+ aclTensor* other = nullptr;
116- aclScalar *alpha = nullptr;118+ aclScalar* alpha = nullptr;
117 std::vector<float> selfHostData = {0, 1, 2, 3, 4, 5, 6, 7};119 std::vector<float> selfHostData = {0, 1, 2, 3, 4, 5, 6, 7};
118 std::vector<float> outHostData = {0, 0, 0, 0, 0, 0, 0, 0};120 std::vector<float> outHostData = {0, 0, 0, 0, 0, 0, 0, 0};
119 std::vector<float> otherHostData = {1, 1, 1, 2, 2, 2, 3, 3};121 std::vector<float> otherHostData = {1, 1, 1, 2, 2, 2, 3, 3};
@@ -134,11 +136,11 @@ int main()
134 INFO_LOG("Get workspace size...");136 INFO_LOG("Get workspace size...");
135 // 3. Call the CANN operator library API(Custom Implementation)137 // 3. Call the CANN operator library API(Custom Implementation)
136 uint64_t workspaceSize = 0;138 uint64_t workspaceSize = 0;
137- aclOpExecutor *executor;139+ aclOpExecutor* executor;
138 CHECK_ERROR(aclnnAddGetWorkspaceSize(self, other, alpha, out, &workspaceSize, &executor));140 CHECK_ERROR(aclnnAddGetWorkspaceSize(self, other, alpha, out, &workspaceSize, &executor));
139 141 
140 // Allocate device memory based on the calculation results.142 // Allocate device memory based on the calculation results.
141- void *workspaceAddr = nullptr;143+ void* workspaceAddr = nullptr;
142 if (workspaceSize > 0lu) {144 if (workspaceSize > 0lu) {
143 CHECK_ERROR(aclrtMalloc(&workspaceAddr, workspaceSize, ACL_MEM_MALLOC_HUGE_FIRST));145 CHECK_ERROR(aclrtMalloc(&workspaceAddr, workspaceSize, ACL_MEM_MALLOC_HUGE_FIRST));
144 }146 }
@@ -152,11 +154,12 @@ int main()
152 // 5. Obtain the execution result of the operator and copy the result from the device memory to the host154 // 5. Obtain the execution result of the operator and copy the result from the device memory to the host
153 auto size = adump::GetShapeSize(outShape);155 auto size = adump::GetShapeSize(outShape);
154 std::vector<float> resultData(size, 0);156 std::vector<float> resultData(size, 0);
155- CHECK_ERROR(aclrtMemcpy(resultData.data(), resultData.size() * sizeof(resultData[0]), outDeviceAddr,157+ CHECK_ERROR(aclrtMemcpy(
156- size * sizeof(float), ACL_MEMCPY_DEVICE_TO_HOST));158+ resultData.data(), resultData.size() * sizeof(resultData[0]), outDeviceAddr, size * sizeof(float),
159+ ACL_MEMCPY_DEVICE_TO_HOST));
157 160 
158 for (int64_t i = 0; i < size; i++) {161 for (int64_t i = 0; i < size; i++) {
159- INFO_LOG("result[%ld] is: %f ", i, resultData[i]);162+ INFO_LOG("result[%ld] is: %f ", i, resultData[i]);
160 }163 }
161 PrintCallbackSummary();164 PrintCallbackSummary();
162 165 
@@ -168,7 +171,7 @@ int main()
168 CHECK_ERROR(aclrtFree(outDeviceAddr));171 CHECK_ERROR(aclrtFree(outDeviceAddr));
169 CHECK_ERROR(aclrtFree(otherDeviceAddr));172 CHECK_ERROR(aclrtFree(otherDeviceAddr));
170 if (workspaceSize > 0lu) {173 if (workspaceSize > 0lu) {
171- CHECK_ERROR(aclrtFree(workspaceAddr));174+ CHECK_ERROR(aclrtFree(workspaceAddr));
172 }175 }
173 // 7. AsendCL Destroy.176 // 7. AsendCL Destroy.
174 // Unregister callback for dump tensor data.177 // Unregister callback for dump tensor data.
Mexample/5_performance/adump/2_model_dump_config/main.cpp+14-13
@@ -20,14 +20,14 @@ using namespace std;
20int main()20int main()
21{21{
22 const int32_t deviceId = 0;22 const int32_t deviceId = 0;
23- const char *dumpCfgPath = "./acl.json";23+ const char* dumpCfgPath = "./acl.json";
24 aclrtStream stream = nullptr;24 aclrtStream stream = nullptr;
25 25 
26 CHECK_ERROR(adump::InitRuntime(deviceId, &stream));26 CHECK_ERROR(adump::InitRuntime(deviceId, &stream));
27 CHECK_ERROR(aclmdlInitDump());27 CHECK_ERROR(aclmdlInitDump());
28 CHECK_ERROR(aclmdlSetDump(dumpCfgPath));28 CHECK_ERROR(aclmdlSetDump(dumpCfgPath));
29 29 
30- const char *dumpPath = acldumpGetPath(DATA_DUMP);30+ const char* dumpPath = acldumpGetPath(DATA_DUMP);
31 if (dumpPath != nullptr) {31 if (dumpPath != nullptr) {
32 INFO_LOG("Configured model dump path is: %s", dumpPath);32 INFO_LOG("Configured model dump path is: %s", dumpPath);
33 } else {33 } else {
@@ -40,14 +40,14 @@ int main()
40 std::vector<float> outHostData = {0, 0, 0, 0, 0, 0, 0, 0};40 std::vector<float> outHostData = {0, 0, 0, 0, 0, 0, 0, 0};
41 float alphaValue = 1.0f;41 float alphaValue = 1.0f;
42 42 
43- void *selfDeviceAddr = nullptr;43+ void* selfDeviceAddr = nullptr;
44- void *otherDeviceAddr = nullptr;44+ void* otherDeviceAddr = nullptr;
45- void *outDeviceAddr = nullptr;45+ void* outDeviceAddr = nullptr;
46- void *workspaceAddr = nullptr;46+ void* workspaceAddr = nullptr;
47- aclTensor *self = nullptr;47+ aclTensor* self = nullptr;
48- aclTensor *other = nullptr;48+ aclTensor* other = nullptr;
49- aclTensor *out = nullptr;49+ aclTensor* out = nullptr;
50- aclScalar *alpha = nullptr;50+ aclScalar* alpha = nullptr;
51 51 
52 CHECK_ERROR(adump::CreateAclTensor(selfHostData, shape, &selfDeviceAddr, aclDataType::ACL_FLOAT, &self));52 CHECK_ERROR(adump::CreateAclTensor(selfHostData, shape, &selfDeviceAddr, aclDataType::ACL_FLOAT, &self));
53 CHECK_ERROR(adump::CreateAclTensor(otherHostData, shape, &otherDeviceAddr, aclDataType::ACL_FLOAT, &other));53 CHECK_ERROR(adump::CreateAclTensor(otherHostData, shape, &otherDeviceAddr, aclDataType::ACL_FLOAT, &other));
@@ -60,7 +60,7 @@ int main()
60 }60 }
61 61 
62 uint64_t workspaceSize = 0;62 uint64_t workspaceSize = 0;
63- aclOpExecutor *executor = nullptr;63+ aclOpExecutor* executor = nullptr;
64 CHECK_ERROR(aclnnAddGetWorkspaceSize(self, other, alpha, out, &workspaceSize, &executor));64 CHECK_ERROR(aclnnAddGetWorkspaceSize(self, other, alpha, out, &workspaceSize, &executor));
65 if (workspaceSize > 0UL) {65 if (workspaceSize > 0UL) {
66 CHECK_ERROR(aclrtMalloc(&workspaceAddr, workspaceSize, ACL_MEM_MALLOC_HUGE_FIRST));66 CHECK_ERROR(aclrtMalloc(&workspaceAddr, workspaceSize, ACL_MEM_MALLOC_HUGE_FIRST));
@@ -69,8 +69,9 @@ int main()
69 CHECK_ERROR(aclrtSynchronizeStream(stream));69 CHECK_ERROR(aclrtSynchronizeStream(stream));
70 70 
71 std::vector<float> resultData(adump::GetShapeSize(shape), 0.0f);71 std::vector<float> resultData(adump::GetShapeSize(shape), 0.0f);
72- CHECK_ERROR(aclrtMemcpy(resultData.data(), resultData.size() * sizeof(float), outDeviceAddr,72+ CHECK_ERROR(aclrtMemcpy(
73- resultData.size() * sizeof(float), ACL_MEMCPY_DEVICE_TO_HOST));73+ resultData.data(), resultData.size() * sizeof(float), outDeviceAddr, resultData.size() * sizeof(float),
74+ ACL_MEMCPY_DEVICE_TO_HOST));
74 for (int64_t i = 0; i < static_cast<int64_t>(resultData.size()); ++i) {75 for (int64_t i = 0; i < static_cast<int64_t>(resultData.size()); ++i) {
75 INFO_LOG("result[%ld] is: %f", i, resultData[i]);76 INFO_LOG("result[%ld] is: %f", i, resultData[i]);
76 }77 }
Mexample/5_performance/adump/3_save_exception_info/main.cpp+28-23
@@ -20,15 +20,16 @@ using namespace std;
20 20 
21namespace {21namespace {
22// 将 shape 填充进 acldumpTensorInfo 的定长数组，并设置有效维度数22// 将 shape 填充进 acldumpTensorInfo 的定长数组，并设置有效维度数
23-void FillTensorInfo(acldumpTensorInfo &info, acldumpTensorType type, void *deviceAddr,23+void FillTensorInfo(
24- const std::vector<int64_t> &shape, int32_t dataType, size_t byteSize)24+ acldumpTensorInfo& info, acldumpTensorType type, void* deviceAddr, const std::vector<int64_t>& shape,
25+ int32_t dataType, size_t byteSize)
25{26{
26 (void)memset_s(&info, sizeof(info), 0, sizeof(info));27 (void)memset_s(&info, sizeof(info), 0, sizeof(info));
27 info.type = type;28 info.type = type;
28 info.tensorSize = byteSize;29 info.tensorSize = byteSize;
29 info.format = aclFormat::ACL_FORMAT_ND;30 info.format = aclFormat::ACL_FORMAT_ND;
30 info.dataType = dataType;31 info.dataType = dataType;
31- info.tensorAddr = static_cast<int64_t *>(deviceAddr);32+ info.tensorAddr = static_cast<int64_t*>(deviceAddr);
32 // 本接口直接以 tensorAddr 作为 Device 上的数据地址读取数据，因此地址类型为裸地址、数据位于 Device33 // 本接口直接以 tensorAddr 作为 Device 上的数据地址读取数据，因此地址类型为裸地址、数据位于 Device
33 info.addrType = ACL_DUMP_ADDR_RAW;34 info.addrType = ACL_DUMP_ADDR_RAW;
34 info.placement = ACL_DUMP_PLACEMENT_DEVICE;35 info.placement = ACL_DUMP_PLACEMENT_DEVICE;
@@ -40,14 +41,14 @@ void FillTensorInfo(acldumpTensorInfo &info, acldumpTensorType type, void *devic
40 info.originShape[i] = static_cast<uint64_t>(shape[i]);41 info.originShape[i] = static_cast<uint64_t>(shape[i]);
41 }42 }
42}43}
43-} // namespace44+} // namespace
44 45 
45int main()46int main()
46{47{
47 // The device id48 // The device id
48 int32_t deviceId = 0;49 int32_t deviceId = 0;
49 // The dump configuration path (enables aic_err_brief_dump exception dump)50 // The dump configuration path (enables aic_err_brief_dump exception dump)
50- const char *dumpCfgPath = "./acl.json";51+ const char* dumpCfgPath = "./acl.json";
51 aclrtStream stream = nullptr;52 aclrtStream stream = nullptr;
52 53 
53 // 1. AscendCL Init（加载开启 Exception Dump 的配置）54 // 1. AscendCL Init（加载开启 Exception Dump 的配置）
@@ -69,13 +70,13 @@ int main()
69 std::vector<int64_t> selfShape{4, 2};70 std::vector<int64_t> selfShape{4, 2};
70 std::vector<int64_t> otherShape{4, 2};71 std::vector<int64_t> otherShape{4, 2};
71 std::vector<int64_t> outShape{4, 2};72 std::vector<int64_t> outShape{4, 2};
72- void *selfDeviceAddr = nullptr;73+ void* selfDeviceAddr = nullptr;
73- void *otherDeviceAddr = nullptr;74+ void* otherDeviceAddr = nullptr;
74- void *outDeviceAddr = nullptr;75+ void* outDeviceAddr = nullptr;
75- aclTensor *self = nullptr;76+ aclTensor* self = nullptr;
76- aclTensor *other = nullptr;77+ aclTensor* other = nullptr;
77- aclScalar *alpha = nullptr;78+ aclScalar* alpha = nullptr;
78- aclTensor *out = nullptr;79+ aclTensor* out = nullptr;
79 std::vector<float> selfHostData = {0, 1, 2, 3, 4, 5, 6, 7};80 std::vector<float> selfHostData = {0, 1, 2, 3, 4, 5, 6, 7};
80 std::vector<float> otherHostData = {1, 1, 1, 2, 2, 2, 3, 3};81 std::vector<float> otherHostData = {1, 1, 1, 2, 2, 2, 3, 3};
81 std::vector<float> outHostData = {0, 0, 0, 0, 0, 0, 0, 0};82 std::vector<float> outHostData = {0, 0, 0, 0, 0, 0, 0, 0};
@@ -96,9 +97,9 @@ int main()
96 97 
97 // 4. Call the CANN operator library API(Custom Implementation) —— 跑通一个正常算子98 // 4. Call the CANN operator library API(Custom Implementation) —— 跑通一个正常算子
98 uint64_t workspaceSize = 0;99 uint64_t workspaceSize = 0;
99- aclOpExecutor *executor;100+ aclOpExecutor* executor;
100 CHECK_ERROR(aclnnAddGetWorkspaceSize(self, other, alpha, out, &workspaceSize, &executor));101 CHECK_ERROR(aclnnAddGetWorkspaceSize(self, other, alpha, out, &workspaceSize, &executor));
101- void *workspaceAddr = nullptr;102+ void* workspaceAddr = nullptr;
102 if (workspaceSize > 0lu) {103 if (workspaceSize > 0lu) {
103 CHECK_ERROR(aclrtMalloc(&workspaceAddr, workspaceSize, ACL_MEM_MALLOC_HUGE_FIRST));104 CHECK_ERROR(aclrtMalloc(&workspaceAddr, workspaceSize, ACL_MEM_MALLOC_HUGE_FIRST));
104 }105 }
@@ -108,14 +109,17 @@ int main()
108 // 5. 组合调用 acldumpSaveExceptionInfo，将算子的 tensor 主动落盘到 Exception Dump 路径下109 // 5. 组合调用 acldumpSaveExceptionInfo，将算子的 tensor 主动落盘到 Exception Dump 路径下
109 std::vector<acldumpTensorInfo> tensors(3);110 std::vector<acldumpTensorInfo> tensors(3);
110 const size_t elemBytes = static_cast<size_t>(adump::GetShapeSize(selfShape)) * sizeof(float);111 const size_t elemBytes = static_cast<size_t>(adump::GetShapeSize(selfShape)) * sizeof(float);
111- FillTensorInfo(tensors[0], ACL_DUMP_TENSOR_INPUT, selfDeviceAddr, selfShape,112+ FillTensorInfo(
112- static_cast<int32_t>(aclDataType::ACL_FLOAT), elemBytes);113+ tensors[0], ACL_DUMP_TENSOR_INPUT, selfDeviceAddr, selfShape, static_cast<int32_t>(aclDataType::ACL_FLOAT),
113- FillTensorInfo(tensors[1], ACL_DUMP_TENSOR_INPUT, otherDeviceAddr, otherShape,114+ elemBytes);
114- static_cast<int32_t>(aclDataType::ACL_FLOAT), elemBytes);115+ FillTensorInfo(
115- FillTensorInfo(tensors[2], ACL_DUMP_TENSOR_OUTPUT, outDeviceAddr, outShape,116+ tensors[1], ACL_DUMP_TENSOR_INPUT, otherDeviceAddr, otherShape, static_cast<int32_t>(aclDataType::ACL_FLOAT),
116- static_cast<int32_t>(aclDataType::ACL_FLOAT), elemBytes);117+ elemBytes);
118+ FillTensorInfo(
119+ tensors[2], ACL_DUMP_TENSOR_OUTPUT, outDeviceAddr, outShape, static_cast<int32_t>(aclDataType::ACL_FLOAT),
120+ elemBytes);
117 121 
118- const char *userTag = "component=demo;stage=forward;note=save_exception_info_example";122+ const char* userTag = "component=demo;stage=forward;note=save_exception_info_example";
119 aclError saveRet = acldumpSaveExceptionInfo("save_exception_info", userTag, tensors.data(), tensors.size());123 aclError saveRet = acldumpSaveExceptionInfo("save_exception_info", userTag, tensors.data(), tensors.size());
120 if (saveRet == ACL_SUCCESS) {124 if (saveRet == ACL_SUCCESS) {
121 INFO_LOG("acldumpSaveExceptionInfo success, data has been saved under exception dump path: %s", excDumpPath);125 INFO_LOG("acldumpSaveExceptionInfo success, data has been saved under exception dump path: %s", excDumpPath);
@@ -126,8 +130,9 @@ int main()
126 // 5. 校验算子结果130 // 5. 校验算子结果
127 auto size = adump::GetShapeSize(outShape);131 auto size = adump::GetShapeSize(outShape);
128 std::vector<float> resultData(size, 0);132 std::vector<float> resultData(size, 0);
129- CHECK_ERROR(aclrtMemcpy(resultData.data(), resultData.size() * sizeof(resultData[0]), outDeviceAddr,133+ CHECK_ERROR(aclrtMemcpy(
130- size * sizeof(float), ACL_MEMCPY_DEVICE_TO_HOST));134+ resultData.data(), resultData.size() * sizeof(resultData[0]), outDeviceAddr, size * sizeof(float),
135+ ACL_MEMCPY_DEVICE_TO_HOST));
131 for (int64_t i = 0; i < size; i++) {136 for (int64_t i = 0; i < size; i++) {
132 INFO_LOG("result[%ld] is: %f", i, resultData[i]);137 INFO_LOG("result[%ld] is: %f", i, resultData[i]);
133 }138 }
Mexample/5_performance/adump/adump_tensor_utils.h+11-9
@@ -16,12 +16,12 @@
16#include "runtime_init_utils.h"16#include "runtime_init_utils.h"
17 17 
18namespace adump {18namespace adump {
19-inline int InitRuntime(int32_t deviceId, aclrtStream *stream, const char *configPath = nullptr)19+inline int InitRuntime(int32_t deviceId, aclrtStream* stream, const char* configPath = nullptr)
20{20{
21 return static_cast<int>(runtime::InitRuntimeAndCreateStream(deviceId, stream, configPath));21 return static_cast<int>(runtime::InitRuntimeAndCreateStream(deviceId, stream, configPath));
22}22}
23 23 
24-inline int64_t GetShapeSize(const std::vector<int64_t> &shape)24+inline int64_t GetShapeSize(const std::vector<int64_t>& shape)
25{25{
26 int64_t shapeSize = 1;26 int64_t shapeSize = 1;
27 for (const int64_t dim : shape) {27 for (const int64_t dim : shape) {
@@ -31,8 +31,9 @@ inline int64_t GetShapeSize(const std::vector<int64_t> &shape)
31}31}
32 32 
33template <typename T>33template <typename T>
34-int CreateAclTensor(const std::vector<T> &hostData, const std::vector<int64_t> &shape, void **deviceAddr,34+int CreateAclTensor(
35- aclDataType dataType, aclTensor **tensor)35+ const std::vector<T>& hostData, const std::vector<int64_t>& shape, void** deviceAddr, aclDataType dataType,
36+ aclTensor** tensor)
36{37{
37 const size_t size = static_cast<size_t>(GetShapeSize(shape)) * sizeof(T);38 const size_t size = static_cast<size_t>(GetShapeSize(shape)) * sizeof(T);
38 aclError ret = aclrtMalloc(deviceAddr, size, ACL_MEM_MALLOC_HUGE_FIRST);39 aclError ret = aclrtMalloc(deviceAddr, size, ACL_MEM_MALLOC_HUGE_FIRST);
@@ -49,18 +50,19 @@ int CreateAclTensor(const std::vector<T> &hostData, const std::vector<int64_t> &
49 strides[static_cast<size_t>(i)] = shape[static_cast<size_t>(i + 1)] * strides[static_cast<size_t>(i + 1)];50 strides[static_cast<size_t>(i)] = shape[static_cast<size_t>(i + 1)] * strides[static_cast<size_t>(i + 1)];
50 }51 }
51 52 
52- *tensor = aclCreateTensor(shape.data(), shape.size(), dataType, strides.data(), 0, aclFormat::ACL_FORMAT_ND,53+ *tensor = aclCreateTensor(
53- shape.data(), shape.size(), *deviceAddr);54+ shape.data(), shape.size(), dataType, strides.data(), 0, aclFormat::ACL_FORMAT_ND, shape.data(), shape.size(),
55+ *deviceAddr);
54 if (*tensor == nullptr) {56 if (*tensor == nullptr) {
55 return -1;57 return -1;
56 }58 }
57 return ACL_SUCCESS;59 return ACL_SUCCESS;
58}60}
59 61 
60-inline void DestroyTensorResources(aclTensor *self, aclTensor *other, aclScalar *alpha, aclTensor *out)62+inline void DestroyTensorResources(aclTensor* self, aclTensor* other, aclScalar* alpha, aclTensor* out)
61{63{
62- aclTensor *tensorList[] = {self, other, out};64+ aclTensor* tensorList[] = {self, other, out};
63- for (aclTensor *tensor : tensorList) {65+ for (aclTensor* tensor : tensorList) {
64 if (tensor != nullptr) {66 if (tensor != nullptr) {
65 aclDestroyTensor(tensor);67 aclDestroyTensor(tensor);
66 }68 }
Mexample/5_performance/profiling/0_create_config/main.cpp+8-5
@@ -15,10 +15,11 @@
15#include "acl/acl.h"15#include "acl/acl.h"
16#include "acl/acl_prof.h"16#include "acl/acl_prof.h"
17 17 
18-int main(int argc, char *argv[]) {18+int main(int argc, char* argv[])
19+{
19 INFO_LOG("-------- Start --------");20 INFO_LOG("-------- Start --------");
20 // 初始化21 // 初始化
21- uint32_t deviceIdList[1] = {0}; // 根据实际环境的DeviceID配置22+ uint32_t deviceIdList[1] = {0}; // 根据实际环境的DeviceID配置
22 aclrtStream stream = nullptr;23 aclrtStream stream = nullptr;
23 // 申请运行时资源24 // 申请运行时资源
24 aclInit(nullptr);25 aclInit(nullptr);
@@ -32,14 +33,16 @@ int main(int argc, char *argv[]) {
32 INFO_LOG("profiling init done");33 INFO_LOG("profiling init done");
33 // 进行profiling配置34 // 进行profiling配置
34 // 创建配置结构体35 // 创建配置结构体
35- aclprofConfig *config = aclprofCreateConfig(deviceIdList, 1, ACL_AICORE_ARITHMETIC_UTILIZATION, 36+ aclprofConfig* config = aclprofCreateConfig(
36- nullptr, ACL_PROF_ACL_API | ACL_PROF_TASK_TIME | ACL_PROF_AICORE_METRICS | ACL_PROF_AICPU | ACL_PROF_L2CACHE | ACL_PROF_HCCL_TRACE | ACL_PROF_MSPROFTX | ACL_PROF_RUNTIME_API);37+ deviceIdList, 1, ACL_AICORE_ARITHMETIC_UTILIZATION, nullptr,
38+ ACL_PROF_ACL_API | ACL_PROF_TASK_TIME | ACL_PROF_AICORE_METRICS | ACL_PROF_AICPU | ACL_PROF_L2CACHE |
39+ ACL_PROF_HCCL_TRACE | ACL_PROF_MSPROFTX | ACL_PROF_RUNTIME_API);
37 std::string memFreq = "15";40 std::string memFreq = "15";
38 aclError ret = aclprofSetConfig(ACL_PROF_SYS_HARDWARE_MEM_FREQ, memFreq.c_str(), memFreq.length());41 aclError ret = aclprofSetConfig(ACL_PROF_SYS_HARDWARE_MEM_FREQ, memFreq.c_str(), memFreq.length());
39 if (ret != ACL_SUCCESS) {42 if (ret != ACL_SUCCESS) {
40 ERROR_LOG("profiling set config error: %d", ret);43 ERROR_LOG("profiling set config error: %d", ret);
41 return -1;44 return -1;
42- } 45+ }
43 INFO_LOG("profiling set config done");46 INFO_LOG("profiling set config done");
44 47 
45 aclprofStart(config);48 aclprofStart(config);
Mexample/5_performance/profiling/1_msproftx/main.cpp+16-19
@@ -16,9 +16,9 @@
16#include "acl/acl_prof.h"16#include "acl/acl_prof.h"
17 17 
18namespace {18namespace {
19-void *CreateStampWithMessage(const std::string &message)19+void* CreateStampWithMessage(const std::string& message)
20{20{
21- void *stamp = aclprofCreateStamp();21+ void* stamp = aclprofCreateStamp();
22 if (stamp == nullptr) {22 if (stamp == nullptr) {
23 ERROR_LOG("aclprofCreateStamp failed for message %s", message.c_str());23 ERROR_LOG("aclprofCreateStamp failed for message %s", message.c_str());
24 return nullptr;24 return nullptr;
@@ -26,15 +26,15 @@ void *CreateStampWithMessage(const std::string &message)
26 26 
27 aclError ret = aclprofSetStampTraceMessage(stamp, message.c_str(), static_cast<uint32_t>(message.length()));27 aclError ret = aclprofSetStampTraceMessage(stamp, message.c_str(), static_cast<uint32_t>(message.length()));
28 if (ret != ACL_SUCCESS) {28 if (ret != ACL_SUCCESS) {
29- ERROR_LOG("aclprofSetStampTraceMessage(%s) failed with error code %d",29+ ERROR_LOG(
30- message.c_str(), static_cast<int32_t>(ret));30+ "aclprofSetStampTraceMessage(%s) failed with error code %d", message.c_str(), static_cast<int32_t>(ret));
31 aclprofDestroyStamp(stamp);31 aclprofDestroyStamp(stamp);
32 return nullptr;32 return nullptr;
33 }33 }
34 return stamp;34 return stamp;
35}35}
36 36 
37-bool CheckPtr(const void *ptr, const char *name)37+bool CheckPtr(const void* ptr, const char* name)
38{38{
39 if (ptr == nullptr) {39 if (ptr == nullptr) {
40 ERROR_LOG("%s is nullptr", name);40 ERROR_LOG("%s is nullptr", name);
@@ -44,14 +44,14 @@ bool CheckPtr(const void *ptr, const char *name)
44}44}
45 45 
46struct StampSet {46struct StampSet {
47- void *loadMark = nullptr;47+ void* loadMark = nullptr;
48- void *outerRangePush = nullptr;48+ void* outerRangePush = nullptr;
49- void *innerRangePush = nullptr;49+ void* innerRangePush = nullptr;
50- void *preprocessRange = nullptr;50+ void* preprocessRange = nullptr;
51- void *execMark = nullptr;51+ void* execMark = nullptr;
52};52};
53 53 
54-int32_t CreateStampSet(StampSet *stampSet)54+int32_t CreateStampSet(StampSet* stampSet)
55{55{
56 stampSet->loadMark = CreateStampWithMessage("model_load_mark");56 stampSet->loadMark = CreateStampWithMessage("model_load_mark");
57 stampSet->outerRangePush = CreateStampWithMessage("forward_pass_push");57 stampSet->outerRangePush = CreateStampWithMessage("forward_pass_push");
@@ -77,7 +77,7 @@ int32_t CreateStampSet(StampSet *stampSet)
77 return 0;77 return 0;
78}78}
79 79 
80-void DestroyStampSet(const StampSet &stampSet)80+void DestroyStampSet(const StampSet& stampSet)
81{81{
82 aclprofDestroyStamp(stampSet.loadMark);82 aclprofDestroyStamp(stampSet.loadMark);
83 aclprofDestroyStamp(stampSet.outerRangePush);83 aclprofDestroyStamp(stampSet.outerRangePush);
@@ -88,7 +88,7 @@ void DestroyStampSet(const StampSet &stampSet)
88 88 
89int32_t RunTraceFlow(aclrtStream stream)89int32_t RunTraceFlow(aclrtStream stream)
90{90{
91- aclprofStepInfo *stepInfo = aclprofCreateStepInfo();91+ aclprofStepInfo* stepInfo = aclprofCreateStepInfo();
92 if (!CheckPtr(stepInfo, "aclprofCreateStepInfo")) {92 if (!CheckPtr(stepInfo, "aclprofCreateStepInfo")) {
93 return -1;93 return -1;
94 }94 }
@@ -141,11 +141,8 @@ int32_t RunMsproftxSample()
141 const std::string aclProfPath = "./output";141 const std::string aclProfPath = "./output";
142 CHECK_ERROR(aclprofInit(aclProfPath.c_str(), aclProfPath.length()));142 CHECK_ERROR(aclprofInit(aclProfPath.c_str(), aclProfPath.length()));
143 143 
144- aclprofConfig *config = aclprofCreateConfig(144+ aclprofConfig* config = aclprofCreateConfig(
145- deviceIdList,145+ deviceIdList, 1, ACL_AICORE_ARITHMETIC_UTILIZATION, nullptr,
146- 1,
147- ACL_AICORE_ARITHMETIC_UTILIZATION,
148- nullptr,
149 ACL_PROF_ACL_API | ACL_PROF_TASK_TIME | ACL_PROF_MSPROFTX);146 ACL_PROF_ACL_API | ACL_PROF_TASK_TIME | ACL_PROF_MSPROFTX);
150 if (!CheckPtr(config, "aclprofCreateConfig")) {147 if (!CheckPtr(config, "aclprofCreateConfig")) {
151 return -1;148 return -1;
@@ -170,7 +167,7 @@ int32_t RunMsproftxSample()
170}167}
171} // namespace168} // namespace
172 169 
173-int main(int argc, char *argv[])170+int main(int argc, char* argv[])
174{171{
175 (void)argc;172 (void)argc;
176 (void)argv;173 (void)argv;
Mexample/5_performance/profiling/2_subscribe_model/main.cpp+10-9
@@ -20,7 +20,7 @@
20 20 
21namespace {21namespace {
22// 自定义函数，实现从用户内存中读取订阅数据的函数22// 自定义函数，实现从用户内存中读取订阅数据的函数
23-void getModelInfo(void *data, uint32_t len)23+void getModelInfo(void* data, uint32_t len)
24{24{
25 uint32_t opNumber = 0;25 uint32_t opNumber = 0;
26 uint32_t dataLen = 0;26 uint32_t dataLen = 0;
@@ -51,7 +51,7 @@ void getModelInfo(void *data, uint32_t len)
51}51}
52 52 
53// 自定义函数，实现从管道中读取数据到用户内存的函数53// 自定义函数，实现从管道中读取数据到用户内存的函数
54-void *profDataRead(void *fd)54+void* profDataRead(void* fd)
55{55{
56 // 设置每次从管道中读取的算子信息个数56 // 设置每次从管道中读取的算子信息个数
57 uint64_t N = 10;57 uint64_t N = 10;
@@ -60,8 +60,9 @@ void *profDataRead(void *fd)
60 aclprofGetOpDescSize(&bufferSize);60 aclprofGetOpDescSize(&bufferSize);
61 // 计算存储算子信息的内存的大小，并且申请内存61 // 计算存储算子信息的内存的大小，并且申请内存
62 uint64_t readbufLen = bufferSize * N;62 uint64_t readbufLen = bufferSize * N;
63- char *readbuf = new char[readbufLen];63+ char* readbuf = new char[readbufLen];
64- // 从管道中读取数据到申请的内存中，读取到的实际数据大小dataLen可能小于bufferSize * N，如果管道中没有数据，默认会阻塞直到读取到数据为止64+ // 从管道中读取数据到申请的内存中，读取到的实际数据大小dataLen可能小于bufferSize *
65+ // N，如果管道中没有数据，默认会阻塞直到读取到数据为止
65 ssize_t dataLen = read(*(int*)fd, readbuf, readbufLen);66 ssize_t dataLen = read(*(int*)fd, readbuf, readbufLen);
66 // 读取数据到readbuf成功67 // 读取数据到readbuf成功
67 while (dataLen > 0) {68 while (dataLen > 0) {
@@ -73,18 +74,18 @@ void *profDataRead(void *fd)
73 memset_s(readbuf, readbufLen, 0, readbufLen);74 memset_s(readbuf, readbufLen, 0, readbufLen);
74 dataLen = read(*(int*)fd, readbuf, readbufLen);75 dataLen = read(*(int*)fd, readbuf, readbufLen);
75 }76 }
76- delete []readbuf;77+ delete[] readbuf;
77 return nullptr;78 return nullptr;
78}79}
79-}80+} // namespace
80 81 
81-int main(int argc, char *argv[])82+int main(int argc, char* argv[])
82{83{
83 INFO_LOG("-------- Start -------- \n");84 INFO_LOG("-------- Start -------- \n");
84 // 调用aclInit初始化85 // 调用aclInit初始化
85 aclInit(nullptr);86 aclInit(nullptr);
86 // 申请运行管理资源87 // 申请运行管理资源
87- uint32_t deviceIdList[1] = {0}; // 根据实际环境的DeviceID配置88+ uint32_t deviceIdList[1] = {0}; // 根据实际环境的DeviceID配置
88 aclrtSetDevice(deviceIdList[0]);89 aclrtSetDevice(deviceIdList[0]);
89 aclrtStream stream = nullptr;90 aclrtStream stream = nullptr;
90 aclrtCreateStream(&stream);91 aclrtCreateStream(&stream);
@@ -101,7 +102,7 @@ int main(int argc, char *argv[])
101 // 读管道指针指向subFd[0]，写管道指针指向subFd[1]102 // 读管道指针指向subFd[0]，写管道指针指向subFd[1]
102 CHECK_ERROR(pipe(subFd));103 CHECK_ERROR(pipe(subFd));
103 // 创建模型订阅的配置并且进行模型订阅104 // 创建模型订阅的配置并且进行模型订阅
104- aclprofSubscribeConfig *config = aclprofCreateSubscribeConfig(1, ACL_AICORE_NONE, &subFd[1]);105+ aclprofSubscribeConfig* config = aclprofCreateSubscribeConfig(1, ACL_AICORE_NONE, &subFd[1]);
105 // 模型订阅需要传入模型的modelId106 // 模型订阅需要传入模型的modelId
106 aclprofModelSubscribe(modelId, config);107 aclprofModelSubscribe(modelId, config);
107 108 
Mexample/5_performance/profiling/3_mstx_with_domain/common/util_acl.h+2-2
@@ -1,5 +1,5 @@
1/**1/**
2-* Copyright (c) 2026 Huawei Technologies Co., Ltd.2+ * Copyright (c) 2026 Huawei Technologies Co., Ltd.
3 * This program is free software, you can redistribute it and/or modify it under the terms and conditions of3 * This program is free software, you can redistribute it and/or modify it under the terms and conditions of
4 * CANN Open Software License Agreement Version 2.0 (the "License").4 * CANN Open Software License Agreement Version 2.0 (the "License").
5 * Please refer to the License for details. You may not use this file except in compliance with the License.5 * Please refer to the License for details. You may not use this file except in compliance with the License.
@@ -34,4 +34,4 @@ int DeInit(int32_t deviceId, aclrtContext* context, aclrtStream* stream)
34 return 0;34 return 0;
35}35}
36 36 
37-# endif // UTIL_ACL_H_37+#endif // UTIL_ACL_H_
Mexample/5_performance/profiling/3_mstx_with_domain/common/util_mstx.h+1-1
@@ -1,5 +1,5 @@
1/**1/**
2-* Copyright (c) 2026 Huawei Technologies Co., Ltd.2+ * Copyright (c) 2026 Huawei Technologies Co., Ltd.
3 * This program is free software, you can redistribute it and/or modify it under the terms and conditions of3 * This program is free software, you can redistribute it and/or modify it under the terms and conditions of
4 * CANN Open Software License Agreement Version 2.0 (the "License").4 * CANN Open Software License Agreement Version 2.0 (the "License").
5 * Please refer to the License for details. You may not use this file except in compliance with the License.5 * Please refer to the License for details. You may not use this file except in compliance with the License.
Mexample/5_performance/profiling/3_mstx_with_domain/mstx_with_domain.cpp+17-18
@@ -1,5 +1,5 @@
1/**1/**
2-* Copyright (c) 2026 Huawei Technologies Co., Ltd.2+ * Copyright (c) 2026 Huawei Technologies Co., Ltd.
3 * This program is free software, you can redistribute it and/or modify it under the terms and conditions of3 * This program is free software, you can redistribute it and/or modify it under the terms and conditions of
4 * CANN Open Software License Agreement Version 2.0 (the "License").4 * CANN Open Software License Agreement Version 2.0 (the "License").
5 * Please refer to the License for details. You may not use this file except in compliance with the License.5 * Please refer to the License for details. You may not use this file except in compliance with the License.
@@ -37,8 +37,9 @@ int64_t GetShapeSize(const std::vector<int64_t>& shape)
37}37}
38 38 
39template <typename T>39template <typename T>
40-int CreateAclTensor(const std::vector<T>& hostData, const std::vector<int64_t>& shape, void** deviceAddr,40+int CreateAclTensor(
41- aclDataType dataType, aclTensor** tensor)41+ const std::vector<T>& hostData, const std::vector<int64_t>& shape, void** deviceAddr, aclDataType dataType,
42+ aclTensor** tensor)
42{43{
43 auto size = GetShapeSize(shape) * sizeof(T);44 auto size = GetShapeSize(shape) * sizeof(T);
44 ACL_CALL(aclrtMalloc(deviceAddr, size, ACL_MEM_MALLOC_HUGE_FIRST));45 ACL_CALL(aclrtMalloc(deviceAddr, size, ACL_MEM_MALLOC_HUGE_FIRST));
@@ -50,8 +51,9 @@ int CreateAclTensor(const std::vector<T>& hostData, const std::vector<int64_t>&
50 }51 }
51 52 
52 // 调用aclCreateTensor接口创建aclTensor53 // 调用aclCreateTensor接口创建aclTensor
53- *tensor = aclCreateTensor(shape.data(), shape.size(), dataType, strides.data(), 0, aclFormat::ACL_FORMAT_ND,54+ *tensor = aclCreateTensor(
54- shape.data(), shape.size(), *deviceAddr);55+ shape.data(), shape.size(), dataType, strides.data(), 0, aclFormat::ACL_FORMAT_ND, shape.data(), shape.size(),
56+ *deviceAddr);
55 return 0;57 return 0;
56}58}
57 59 
@@ -96,9 +98,12 @@ int DoAclAdd(aclrtContext context, aclrtStream stream)
96 ACL_CALL(aclrtSynchronizeStream(stream));98 ACL_CALL(aclrtSynchronizeStream(stream));
97 auto size = GetShapeSize(outShape);99 auto size = GetShapeSize(outShape);
98 std::vector<float> resultData(size, 0);100 std::vector<float> resultData(size, 0);
99- ACL_CALL(aclrtMemcpy(resultData.data(), resultData.size() * sizeof(resultData[0]), outDeviceAddr,101+ ACL_CALL(aclrtMemcpy(
100- size * sizeof(float), ACL_MEMCPY_DEVICE_TO_HOST));102+ resultData.data(), resultData.size() * sizeof(resultData[0]), outDeviceAddr, size * sizeof(float),
101- for (int64_t i = 0; i < size; i++) { LOG_PRINT("result[%ld] is: %f\n", i, resultData[i]); }103+ ACL_MEMCPY_DEVICE_TO_HOST));
104+ for (int64_t i = 0; i < size; i++) {
105+ LOG_PRINT("result[%ld] is: %f\n", i, resultData[i]);
106+ }
102 aclDestroyTensor(self);107 aclDestroyTensor(self);
103 aclDestroyTensor(other);108 aclDestroyTensor(other);
104 aclDestroyScalar(alpha);109 aclDestroyScalar(alpha);
@@ -129,18 +134,12 @@ static int TestMstxWithDomain()
129 return ACL_SUCCESS;134 return ACL_SUCCESS;
130}135}
131 136 
132-void MstxDomainInit()137+void MstxDomainInit() { domainRange = mstxDomainCreateA(g_domainRangeName.c_str()); }
133-{
134- domainRange = mstxDomainCreateA(g_domainRangeName.c_str());
135-}
136 138 
137-void MstxDomainDeInit()139+void MstxDomainDeInit() { mstxDomainDestroy(domainRange); }
138-{140+} // namespace
139- mstxDomainDestroy(domainRange);
140-}
141-}
142 141 
143-int main(int argc, const char **argv)142+int main(int argc, const char** argv)
144{143{
145 int32_t deviceId = 0;144 int32_t deviceId = 0;
146 int ret = Init(deviceId, &context, &stream);145 int ret = Init(deviceId, &context, &stream);
Mexample/kernel_func/add_custom.cpp+10-10
@@ -15,22 +15,22 @@
15#include "kernel_operator.h"15#include "kernel_operator.h"
16 16 
17namespace {17namespace {
18- constexpr int32_t TOTAL_LENGTH = 8 * 2048; // total length of data18+constexpr int32_t TOTAL_LENGTH = 8 * 2048; // total length of data
19- constexpr int32_t USE_CORE_NUM = 8; // num of core used19+constexpr int32_t USE_CORE_NUM = 8; // num of core used
20- constexpr int32_t BLOCK_LENGTH = TOTAL_LENGTH / USE_CORE_NUM; // length computed of each core20+constexpr int32_t BLOCK_LENGTH = TOTAL_LENGTH / USE_CORE_NUM; // length computed of each core
21- constexpr int32_t TILE_NUM = 8; // split data into 8 tiles for each core21+constexpr int32_t TILE_NUM = 8; // split data into 8 tiles for each core
22- constexpr int32_t BUFFER_NUM = 2; // tensor num for each queue22+constexpr int32_t BUFFER_NUM = 2; // tensor num for each queue
23- constexpr int32_t TILE_LENGTH = BLOCK_LENGTH / TILE_NUM / BUFFER_NUM; // separate to 2 parts, due to double buffer23+constexpr int32_t TILE_LENGTH = BLOCK_LENGTH / TILE_NUM / BUFFER_NUM; // separate to 2 parts, due to double buffer
24-}24+} // namespace
25 25 
26class KernelAdd {26class KernelAdd {
27public:27public:
28 __aicore__ inline KernelAdd() {}28 __aicore__ inline KernelAdd() {}
29 __aicore__ inline void Init(GM_ADDR x, GM_ADDR y, GM_ADDR z)29 __aicore__ inline void Init(GM_ADDR x, GM_ADDR y, GM_ADDR z)
30 {30 {
31- xGm.SetGlobalBuffer((__gm__ half *)x + BLOCK_LENGTH * AscendC::GetBlockIdx(), BLOCK_LENGTH);31+ xGm.SetGlobalBuffer((__gm__ half*)x + BLOCK_LENGTH * AscendC::GetBlockIdx(), BLOCK_LENGTH);
32- yGm.SetGlobalBuffer((__gm__ half *)y + BLOCK_LENGTH * AscendC::GetBlockIdx(), BLOCK_LENGTH);32+ yGm.SetGlobalBuffer((__gm__ half*)y + BLOCK_LENGTH * AscendC::GetBlockIdx(), BLOCK_LENGTH);
33- zGm.SetGlobalBuffer((__gm__ half *)z + BLOCK_LENGTH * AscendC::GetBlockIdx(), BLOCK_LENGTH);33+ zGm.SetGlobalBuffer((__gm__ half*)z + BLOCK_LENGTH * AscendC::GetBlockIdx(), BLOCK_LENGTH);
34 pipe.InitBuffer(inQueueX, BUFFER_NUM, TILE_LENGTH * sizeof(half));34 pipe.InitBuffer(inQueueX, BUFFER_NUM, TILE_LENGTH * sizeof(half));
35 pipe.InitBuffer(inQueueY, BUFFER_NUM, TILE_LENGTH * sizeof(half));35 pipe.InitBuffer(inQueueY, BUFFER_NUM, TILE_LENGTH * sizeof(half));
36 pipe.InitBuffer(outQueueZ, BUFFER_NUM, TILE_LENGTH * sizeof(half));36 pipe.InitBuffer(outQueueZ, BUFFER_NUM, TILE_LENGTH * sizeof(half));
Mexample/kernel_func/add_custom_tiling.cpp+9-8
@@ -8,7 +8,7 @@
8 * See LICENSE in the root of the software repository for the full text of the License.8 * See LICENSE in the root of the software repository for the full text of the License.
9 */9 */
10 10 
11- /*11+/*
12 * Below is the operator used for computing the addition of two matrices.12 * Below is the operator used for computing the addition of two matrices.
13 * This operator allows users to configure the tiling information with placeholders.13 * This operator allows users to configure the tiling information with placeholders.
14 */14 */
@@ -16,9 +16,9 @@
16#include "kernel_operator.h"16#include "kernel_operator.h"
17 17 
18namespace {18namespace {
19- constexpr int32_t USE_CORE_NUM = 8; // num of core used 19+constexpr int32_t USE_CORE_NUM = 8; // num of core used
20- constexpr int32_t BUFFER_NUM = 2; // tensor num for each queue20+constexpr int32_t BUFFER_NUM = 2; // tensor num for each queue
21-}21+} // namespace
22 22 
23class KernelAdd {23class KernelAdd {
24public:24public:
@@ -31,9 +31,9 @@ public:
31 return;31 return;
32 }32 }
33 this->tileLength = this->blockLength / tileNum / BUFFER_NUM;33 this->tileLength = this->blockLength / tileNum / BUFFER_NUM;
34- xGm.SetGlobalBuffer((__gm__ half *)x + this->blockLength * AscendC::GetBlockIdx(), this->blockLength);34+ xGm.SetGlobalBuffer((__gm__ half*)x + this->blockLength * AscendC::GetBlockIdx(), this->blockLength);
35- yGm.SetGlobalBuffer((__gm__ half *)y + this->blockLength * AscendC::GetBlockIdx(), this->blockLength);35+ yGm.SetGlobalBuffer((__gm__ half*)y + this->blockLength * AscendC::GetBlockIdx(), this->blockLength);
36- zGm.SetGlobalBuffer((__gm__ half *)z + this->blockLength * AscendC::GetBlockIdx(), this->blockLength);36+ zGm.SetGlobalBuffer((__gm__ half*)z + this->blockLength * AscendC::GetBlockIdx(), this->blockLength);
37 pipe.InitBuffer(inQueueX, BUFFER_NUM, this->tileLength * sizeof(half));37 pipe.InitBuffer(inQueueX, BUFFER_NUM, this->tileLength * sizeof(half));
38 pipe.InitBuffer(inQueueY, BUFFER_NUM, this->tileLength * sizeof(half));38 pipe.InitBuffer(inQueueY, BUFFER_NUM, this->tileLength * sizeof(half));
39 pipe.InitBuffer(outQueueZ, BUFFER_NUM, this->tileLength * sizeof(half));39 pipe.InitBuffer(outQueueZ, BUFFER_NUM, this->tileLength * sizeof(half));
@@ -88,7 +88,8 @@ private:
88 uint32_t tileLength;88 uint32_t tileLength;
89};89};
90 90 
91-extern "C" __global__ __aicore__ void add_custom(GM_ADDR x, GM_ADDR y, GM_ADDR z, __gm__ int32_t* tilingLength, __gm__ int32_t* tilingNum)91+extern "C" __global__ __aicore__ void add_custom(
92+ GM_ADDR x, GM_ADDR y, GM_ADDR z, __gm__ int32_t* tilingLength, __gm__ int32_t* tilingNum)
92{93{
93 KernelAdd op;94 KernelAdd op;
94 uint32_t tileLength = *tilingLength;95 uint32_t tileLength = *tilingLength;
Mexample/kernel_func/easy_OP.cpp+21-26
@@ -1,29 +1,24 @@
1-/**1+/**
2- * Copyright (c) 2025 Huawei Technologies Co., Ltd.2+ * Copyright (c) 2025 Huawei Technologies Co., Ltd.
3- * This program is free software, you can redistribute it and/or modify it under the terms and conditions of3+ * This program is free software, you can redistribute it and/or modify it under the terms and conditions of
4- * CANN Open Software License Agreement Version 2.0 (the "License").4+ * CANN Open Software License Agreement Version 2.0 (the "License").
5- * Please refer to the License for details. You may not use this file except in compliance with the License.5+ * Please refer to the License for details. You may not use this file except in compliance with the License.
6- * THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED,6+ * THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED,
7- * INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE.7+ * INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE.
8- * See LICENSE in the root of the software repository for the full text of the License.8+ * See LICENSE in the root of the software repository for the full text of the License.
9- */9+ */
10- 10+ 
11#include "kernel_operator.h"11#include "kernel_operator.h"
12#include "kernel_ops.h"12#include "kernel_ops.h"
13- 13+ 
14-// 简单核函数，让x自增114+// 简单核函数，让x自增1
15-extern "C" __global__ __aicore__ void EasyOPf(__gm__ uint32_t* x)15+extern "C" __global__ __aicore__ void EasyOPf(__gm__ uint32_t* x)
16-{16+{
17- int32_t idx = block_idx;17+ int32_t idx = block_idx;
18- x[idx] += 1;18+ x[idx] += 1;
19-#if __NPU_ARCH__ == 351019+#if __NPU_ARCH__ == 3510
20- dcci(reinterpret_cast<__gm__ int64_t*>(x),20+ dcci(reinterpret_cast<__gm__ int64_t*>(x), cache_line_t::ENTIRE_DATA_CACHE, dcci_dst_t::CACHELINE_OUT);
21- cache_line_t::ENTIRE_DATA_CACHE,21+#endif
22- dcci_dst_t::CACHELINE_OUT);
23-#endif
24-}
25- 
26-void EasyOP(uint32_t blockDim, void *stream, uint32_t* x)
27-{
28- EasyOPf<<<blockDim, nullptr, stream>>>(x);
29}22}
23+ 
24+void EasyOP(uint32_t blockDim, void* stream, uint32_t* x) { EasyOPf<<<blockDim, nullptr, stream>>>(x); }
Mexample/kernel_func/error_OP.cpp+15-21
@@ -1,23 +1,17 @@
1-/**1+/**
2- * Copyright (c) 2025 Huawei Technologies Co., Ltd.2+ * Copyright (c) 2025 Huawei Technologies Co., Ltd.
3- * This program is free software, you can redistribute it and/or modify it under the terms and conditions of3+ * This program is free software, you can redistribute it and/or modify it under the terms and conditions of
4- * CANN Open Software License Agreement Version 2.0 (the "License").4+ * CANN Open Software License Agreement Version 2.0 (the "License").
5- * Please refer to the License for details. You may not use this file except in compliance with the License.5+ * Please refer to the License for details. You may not use this file except in compliance with the License.
6- * THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED,6+ * THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED,
7- * INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE.7+ * INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE.
8- * See LICENSE in the root of the software repository for the full text of the License.8+ * See LICENSE in the root of the software repository for the full text of the License.
9- */9+ */
10- 10+ 
11#include "kernel_operator.h"11#include "kernel_operator.h"
12#include "kernel_ops.h"12#include "kernel_ops.h"
13- 13+ 
14-// 核函数，会报错，用于遇错即停的情况14+// 核函数，会报错，用于遇错即停的情况
15-extern "C" __global__ __aicore__ void ErrorOPf()15+extern "C" __global__ __aicore__ void ErrorOPf() { asm("trap"); }
16-{16+ 
17- asm("trap");17+void ErrorOP(uint32_t blockDim, void* stream) { ErrorOPf<<<blockDim, nullptr, stream>>>(); }
18-}
19- 
20-void ErrorOP(uint32_t blockDim, void *stream)
21-{
22- ErrorOPf<<<blockDim, nullptr, stream>>>();
23-}
Mexample/kernel_func/kernel_add.cpp+111-111
@@ -1,114 +1,114 @@
1-/**1+/**
2- * Copyright (c) 2025 Huawei Technologies Co., Ltd.2+ * Copyright (c) 2025 Huawei Technologies Co., Ltd.
3- * This program is free software, you can redistribute it and/or modify it under the terms and conditions of3+ * This program is free software, you can redistribute it and/or modify it under the terms and conditions of
4- * CANN Open Software License Agreement Version 2.0 (the "License").4+ * CANN Open Software License Agreement Version 2.0 (the "License").
5- * Please refer to the License for details. You may not use this file except in compliance with the License.5+ * Please refer to the License for details. You may not use this file except in compliance with the License.
6- * THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED,6+ * THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED,
7- * INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE.7+ * INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE.
8- * See LICENSE in the root of the software repository for the full text of the License.8+ * See LICENSE in the root of the software repository for the full text of the License.
9- */9+ */
10- 10+ 
11#include "kernel_operator.h"11#include "kernel_operator.h"
12#include "kernel_add.h"12#include "kernel_add.h"
13- 13+ 
14-using namespace AscendC;14+using namespace AscendC;
15- 15+ 
16-namespace {16+namespace {
17- constexpr int32_t BUFFER_NUM = 2; // tensor num for each queue17+constexpr int32_t BUFFER_NUM = 2; // tensor num for each queue
18- 18+ 
19- class KernelAdd {19+class KernelAdd {
20- public:20+public:
21- __aicore__ inline KernelAdd() {}21+ __aicore__ inline KernelAdd() {}
22- 22+ 
23- __aicore__ inline void Init(__gm__ float* srcA, __gm__ float* srcB, __gm__ float* dst, uint32_t totalSize)23+ __aicore__ inline void Init(__gm__ float* srcA, __gm__ float* srcB, __gm__ float* dst, uint32_t totalSize)
24- {24+ {
25- blockIdx = GetBlockIdx();25+ blockIdx = GetBlockIdx();
26- blockDim = GetBlockNum();26+ blockDim = GetBlockNum();
27- elementsPerBlock = totalSize / blockDim;27+ elementsPerBlock = totalSize / blockDim;
28- tileLength = elementsPerBlock / TILE_NUM / BUFFER_NUM;28+ tileLength = elementsPerBlock / TILE_NUM / BUFFER_NUM;
29- startIdx = blockIdx * elementsPerBlock;29+ startIdx = blockIdx * elementsPerBlock;
30- 30+ 
31- // Set the global buffer31+ // Set the global buffer
32- srcAGlobal.SetGlobalBuffer(srcA + startIdx, elementsPerBlock);32+ srcAGlobal.SetGlobalBuffer(srcA + startIdx, elementsPerBlock);
33- srcBGlobal.SetGlobalBuffer(srcB + startIdx, elementsPerBlock);33+ srcBGlobal.SetGlobalBuffer(srcB + startIdx, elementsPerBlock);
34- dstGlobal.SetGlobalBuffer(dst + startIdx, elementsPerBlock);34+ dstGlobal.SetGlobalBuffer(dst + startIdx, elementsPerBlock);
35- 35+ 
36- // Initialize the pipe buffer36+ // Initialize the pipe buffer
37- pipe.InitBuffer(queueInA, BUFFER_NUM, tileLength * sizeof(float));37+ pipe.InitBuffer(queueInA, BUFFER_NUM, tileLength * sizeof(float));
38- pipe.InitBuffer(queueInB, BUFFER_NUM, tileLength * sizeof(float));38+ pipe.InitBuffer(queueInB, BUFFER_NUM, tileLength * sizeof(float));
39- pipe.InitBuffer(queueOut, BUFFER_NUM, tileLength * sizeof(float));39+ pipe.InitBuffer(queueOut, BUFFER_NUM, tileLength * sizeof(float));
40- }40+ }
41- 41+ 
42- __aicore__ inline void Process()42+ __aicore__ inline void Process()
43- {43+ {
44- int32_t loopCount = TILE_NUM * BUFFER_NUM;44+ int32_t loopCount = TILE_NUM * BUFFER_NUM;
45- for (int32_t i = 0; i < loopCount; i++) {45+ for (int32_t i = 0; i < loopCount; i++) {
46- CopyIn(i);46+ CopyIn(i);
47- Compute(i);47+ Compute(i);
48- CopyOut(i);48+ CopyOut(i);
49- }49+ }
50- }50+ }
51- 51+ 
52- private:52+private:
53- __aicore__ inline void CopyIn(int32_t progress)53+ __aicore__ inline void CopyIn(int32_t progress)
54- {54+ {
55- LocalTensor<float> localA = queueInA.AllocTensor<float>();55+ LocalTensor<float> localA = queueInA.AllocTensor<float>();
56- LocalTensor<float> localB = queueInB.AllocTensor<float>();56+ LocalTensor<float> localB = queueInB.AllocTensor<float>();
57- 57+ 
58- DataCopy(localA, srcAGlobal[progress * tileLength], tileLength);58+ DataCopy(localA, srcAGlobal[progress * tileLength], tileLength);
59- DataCopy(localB, srcBGlobal[progress * tileLength], tileLength);59+ DataCopy(localB, srcBGlobal[progress * tileLength], tileLength);
60- 60+ 
61- queueInA.EnQue(localA);61+ queueInA.EnQue(localA);
62- queueInB.EnQue(localB);62+ queueInB.EnQue(localB);
63- }63+ }
64- 64+ 
65- __aicore__ inline void Compute(int32_t progress)65+ __aicore__ inline void Compute(int32_t progress)
66- {66+ {
67- LocalTensor<float> localA = queueInA.DeQue<float>();67+ LocalTensor<float> localA = queueInA.DeQue<float>();
68- LocalTensor<float> localB = queueInB.DeQue<float>();68+ LocalTensor<float> localB = queueInB.DeQue<float>();
69- LocalTensor<float> localOut = queueOut.AllocTensor<float>();69+ LocalTensor<float> localOut = queueOut.AllocTensor<float>();
70- 70+ 
71- Add(localOut, localA, localB, tileLength);71+ Add(localOut, localA, localB, tileLength);
72- 72+ 
73- queueOut.EnQue<float>(localOut);73+ queueOut.EnQue<float>(localOut);
74- queueInA.FreeTensor(localA);74+ queueInA.FreeTensor(localA);
75- queueInB.FreeTensor(localB);75+ queueInB.FreeTensor(localB);
76- }76+ }
77- 77+ 
78- __aicore__ inline void CopyOut(int32_t progress)78+ __aicore__ inline void CopyOut(int32_t progress)
79- {79+ {
80- LocalTensor<float> localOut = queueOut.DeQue<float>();80+ LocalTensor<float> localOut = queueOut.DeQue<float>();
81- DataCopy(dstGlobal[progress * tileLength], localOut, tileLength);81+ DataCopy(dstGlobal[progress * tileLength], localOut, tileLength);
82- queueOut.FreeTensor(localOut);82+ queueOut.FreeTensor(localOut);
83- }83+ }
84- 84+ 
85- private:85+private:
86- TPipe pipe;86+ TPipe pipe;
87- TQue<QuePosition::VECIN, BUFFER_NUM> queueInA;87+ TQue<QuePosition::VECIN, BUFFER_NUM> queueInA;
88- TQue<QuePosition::VECIN, BUFFER_NUM> queueInB;88+ TQue<QuePosition::VECIN, BUFFER_NUM> queueInB;
89- TQue<QuePosition::VECOUT, BUFFER_NUM> queueOut;89+ TQue<QuePosition::VECOUT, BUFFER_NUM> queueOut;
90- GlobalTensor<float> srcAGlobal;90+ GlobalTensor<float> srcAGlobal;
91- GlobalTensor<float> srcBGlobal;91+ GlobalTensor<float> srcBGlobal;
92- GlobalTensor<float> dstGlobal;92+ GlobalTensor<float> dstGlobal;
93- 93+ 
94- uint32_t blockIdx;94+ uint32_t blockIdx;
95- uint32_t blockDim;95+ uint32_t blockDim;
96- uint32_t elementsPerBlock;96+ uint32_t elementsPerBlock;
97- uint32_t startIdx;97+ uint32_t startIdx;
98- uint32_t tileLength;98+ uint32_t tileLength;
99- static constexpr int32_t TILE_NUM = 8; // split data into 8 tiles for each core99+ static constexpr int32_t TILE_NUM = 8; // split data into 8 tiles for each core
100- };100+};
101-} // namespace101+} // namespace
102- 102+ 
103-extern "C" __global__ __aicore__ void AddKernel(__gm__ float* srcA, __gm__ float* srcB, __gm__ float* dst,103+extern "C" __global__ __aicore__ void AddKernel(
104- uint32_t totalSize)104+ __gm__ float* srcA, __gm__ float* srcB, __gm__ float* dst, uint32_t totalSize)
105-{105+{
106- KernelAdd op;106+ KernelAdd op;
107- op.Init(srcA, srcB, dst, totalSize);107+ op.Init(srcA, srcB, dst, totalSize);
108- op.Process();108+ op.Process();
109-}109+}
110- 110+ 
111-void AddDo(uint32_t blockDim, void* stream, float* srcA, float* srcB, float* dst, uint32_t totalSize)111+void AddDo(uint32_t blockDim, void* stream, float* srcA, float* srcB, float* dst, uint32_t totalSize)
112-{112+{
113- AddKernel<<<blockDim, nullptr, stream>>>(srcA, srcB, dst, totalSize);113+ AddKernel<<<blockDim, nullptr, stream>>>(srcA, srcB, dst, totalSize);
114}114}
Mexample/kernel_func/kernel_print.cpp+3-10
@@ -8,20 +8,13 @@
8 * See LICENSE in the root of the software repository for the full text of the License.8 * See LICENSE in the root of the software repository for the full text of the License.
9 */9 */
10 10 
11- /*11+/*
12 * Using the Ascend kernel to print the data12 * Using the Ascend kernel to print the data
13 */13 */
14 14 
15#include "kernel_operator.h"15#include "kernel_operator.h"
16#include "kernel_ops.h"16#include "kernel_ops.h"
17 17 
18-extern "C" __global__ __aicore__ void KernelPrint()18+extern "C" __global__ __aicore__ void KernelPrint() { AscendC::printf("Hello World\n"); }
19-{
20- AscendC::printf("Hello World\n");
21-}
22- 
23-void PrintDo(uint32_t blockDim, void *stream)
24-{
25- KernelPrint<<<blockDim, nullptr, stream>>>();
26-}
27 19 
20+void PrintDo(uint32_t blockDim, void* stream) { KernelPrint<<<blockDim, nullptr, stream>>>(); }
Mexample/kernel_func/long_OP.cpp+34-39
@@ -1,42 +1,37 @@
1-/**1+/**
2- * Copyright (c) 2025 Huawei Technologies Co., Ltd.2+ * Copyright (c) 2025 Huawei Technologies Co., Ltd.
3- * This program is free software, you can redistribute it and/or modify it under the terms and conditions of3+ * This program is free software, you can redistribute it and/or modify it under the terms and conditions of
4- * CANN Open Software License Agreement Version 2.0 (the "License").4+ * CANN Open Software License Agreement Version 2.0 (the "License").
5- * Please refer to the License for details. You may not use this file except in compliance with the License.5+ * Please refer to the License for details. You may not use this file except in compliance with the License.
6- * THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED,6+ * THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED,
7- * INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE.7+ * INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE.
8- * See LICENSE in the root of the software repository for the full text of the License.8+ * See LICENSE in the root of the software repository for the full text of the License.
9- */9+ */
10- 10+ 
11#include "kernel_operator.h"11#include "kernel_operator.h"
12#include "kernel_ops.h"12#include "kernel_ops.h"
13- 13+ 
14-// 核函数，长时间任务14+// 核函数，长时间任务
15-extern "C" __global__ __aicore__ void LongOPf(__gm__ uint32_t* x)15+extern "C" __global__ __aicore__ void LongOPf(__gm__ uint32_t* x)
16-{16+{
17- float temp = 0.0f;17+ float temp = 0.0f;
18- int idx = block_idx;18+ int idx = block_idx;
19- //复杂的浮点运算来模拟长时间任务19+ // 复杂的浮点运算来模拟长时间任务
20- for (int i = 0; i < 10000; i++) {20+ for (int i = 0; i < 10000; i++) {
21- temp += 0.00002f;21+ temp += 0.00002f;
22- temp *= 0.02f;22+ temp *= 0.02f;
23- for (int j = 0; j < 25000; j++){23+ for (int j = 0; j < 25000; j++) {
24- temp = temp * temp - 0.000002f;24+ temp = temp * temp - 0.000002f;
25- temp = temp + (temp * 0.05f);25+ temp = temp + (temp * 0.05f);
26- temp = temp * temp + 0.00001f;26+ temp = temp * temp + 0.00001f;
27- temp = temp > 1.0f ? 0.0f : temp;27+ temp = temp > 1.0f ? 0.0f : temp;
28- }28+ }
29- temp -= 0.00003f * temp;29+ temp -= 0.00003f * temp;
30- }30+ }
31- x[idx] += temp > 0.0f ? 1:0;31+ x[idx] += temp > 0.0f ? 1 : 0;
32-#if __NPU_ARCH__ == 351032+#if __NPU_ARCH__ == 3510
33- dcci(reinterpret_cast<__gm__ int64_t*>(x),33+ dcci(reinterpret_cast<__gm__ int64_t*>(x), cache_line_t::ENTIRE_DATA_CACHE, dcci_dst_t::CACHELINE_OUT);
34- cache_line_t::ENTIRE_DATA_CACHE,34+#endif
35- dcci_dst_t::CACHELINE_OUT);
36-#endif
37-}
38- 
39-void LongOP(uint32_t blockDim, void *stream, uint32_t* x)
40-{
41- LongOPf<<<blockDim, nullptr, stream>>>(x);
42}35}
36+ 
37+void LongOP(uint32_t blockDim, void* stream, uint32_t* x) { LongOPf<<<blockDim, nullptr, stream>>>(x); }
Mexample/kernel_func/short_OP.cpp+20-25
@@ -1,28 +1,23 @@
1-/**1+/**
2- * Copyright (c) 2025 Huawei Technologies Co., Ltd.2+ * Copyright (c) 2025 Huawei Technologies Co., Ltd.
3- * This program is free software, you can redistribute it and/or modify it under the terms and conditions of3+ * This program is free software, you can redistribute it and/or modify it under the terms and conditions of
4- * CANN Open Software License Agreement Version 2.0 (the "License").4+ * CANN Open Software License Agreement Version 2.0 (the "License").
5- * Please refer to the License for details. You may not use this file except in compliance with the License.5+ * Please refer to the License for details. You may not use this file except in compliance with the License.
6- * THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED,6+ * THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED,
7- * INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE.7+ * INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE.
8- * See LICENSE in the root of the software repository for the full text of the License.8+ * See LICENSE in the root of the software repository for the full text of the License.
9- */9+ */
10- 10+ 
11#include "kernel_operator.h"11#include "kernel_operator.h"
12#include "kernel_ops.h"12#include "kernel_ops.h"
13-// 核函数，自乘213+// 核函数，自乘2
14-extern "C" __global__ __aicore__ void ShortOPf(__gm__ uint32_t* x)14+extern "C" __global__ __aicore__ void ShortOPf(__gm__ uint32_t* x)
15-{15+{
16- int idx = block_idx;16+ int idx = block_idx;
17- x[idx] *= 2;17+ x[idx] *= 2;
18-#if __NPU_ARCH__ == 351018+#if __NPU_ARCH__ == 3510
19- dcci(reinterpret_cast<__gm__ int64_t*>(x),19+ dcci(reinterpret_cast<__gm__ int64_t*>(x), cache_line_t::ENTIRE_DATA_CACHE, dcci_dst_t::CACHELINE_OUT);
20- cache_line_t::ENTIRE_DATA_CACHE,20+#endif
21- dcci_dst_t::CACHELINE_OUT);
22-#endif
23-}
24- 
25-void ShortOP(uint32_t blockDim, void *stream, uint32_t* x)
26-{
27- ShortOPf<<<blockDim, nullptr, stream>>>(x);
28}21}
22+ 
23+void ShortOP(uint32_t blockDim, void* stream, uint32_t* x) { ShortOPf<<<blockDim, nullptr, stream>>>(x); }
Mexample/kernel_func/write_read_value.cpp+3-6
@@ -8,7 +8,7 @@
8 * See LICENSE in the root of the software repository for the full text of the License.8 * See LICENSE in the root of the software repository for the full text of the License.
9 */9 */
10 10 
11- /*11+/*
12 * Using the Ascend kernel to write the value at devPtrA and read the value at devPtrB12 * Using the Ascend kernel to write the value at devPtrA and read the value at devPtrB
13 */13 */
14 14 
@@ -22,7 +22,7 @@ extern "C" __global__ __aicore__ void DeviceWrite(__gm__ int* devPtr, int value)
22 AscendC::printf("Source data: %d\n", value);22 AscendC::printf("Source data: %d\n", value);
23}23}
24 24 
25-void WriteDo(uint32_t blockDim, void *stream, int* devPtr, int value)25+void WriteDo(uint32_t blockDim, void* stream, int* devPtr, int value)
26{26{
27 DeviceWrite<<<blockDim, nullptr, stream>>>(devPtr, value);27 DeviceWrite<<<blockDim, nullptr, stream>>>(devPtr, value);
28}28}
@@ -34,7 +34,4 @@ extern "C" __global__ __aicore__ void DeviceRead(__gm__ int* devPtr)
34 AscendC::printf("Destination data: %d\n", value);34 AscendC::printf("Destination data: %d\n", value);
35}35}
36 36 
37-void ReadDo(uint32_t blockDim, void *stream, int* devPtr)37+void ReadDo(uint32_t blockDim, void* stream, int* devPtr) { DeviceRead<<<blockDim, nullptr, stream>>>(devPtr); }
38-{
39- DeviceRead<<<blockDim, nullptr, stream>>>(devPtr);
40-}
Mexample/runtime_init_utils.h+1-1
@@ -14,7 +14,7 @@
14#include "acl/acl.h"14#include "acl/acl.h"
15 15 
16namespace runtime {16namespace runtime {
17-inline aclError InitRuntimeAndCreateStream(int32_t deviceId, aclrtStream *stream, const char *configPath = nullptr)17+inline aclError InitRuntimeAndCreateStream(int32_t deviceId, aclrtStream* stream, const char* configPath = nullptr)
18{18{
19 aclError ret = aclInit(configPath);19 aclError ret = aclInit(configPath);
20 if (ret != ACL_SUCCESS) {20 if (ret != ACL_SUCCESS) {
Mexample/tools/get_soc_version/get_soc_version.cpp+1-1
@@ -16,7 +16,7 @@ int main()
16{16{
17 // aclrtGetSocName returns the same SOC_VERSION string required by sample builds.17 // aclrtGetSocName returns the same SOC_VERSION string required by sample builds.
18 // It does not require aclInit, so this helper stays small and fast for shell use.18 // It does not require aclInit, so this helper stays small and fast for shell use.
19- const char *socName = aclrtGetSocName();19+ const char* socName = aclrtGetSocName();
20 if ((socName == nullptr) || (socName[0] == '\0')) {20 if ((socName == nullptr) || (socName[0] == '\0')) {
21 std::cerr << "[ERROR]: Failed to get SOC_VERSION by aclrtGetSocName." << std::endl;21 std::cerr << "[ERROR]: Failed to get SOC_VERSION by aclrtGetSocName." << std::endl;
22 return 1;22 return 1;
Mexample/utils.h+12-12
@@ -16,21 +16,21 @@
16#define WARN_LOG(fmt, args...) fprintf(stdout, "[WARN] " fmt "\n", ##args)16#define WARN_LOG(fmt, args...) fprintf(stdout, "[WARN] " fmt "\n", ##args)
17#define ERROR_LOG(fmt, args...) fprintf(stdout, "[ERROR] " fmt "\n", ##args)17#define ERROR_LOG(fmt, args...) fprintf(stdout, "[ERROR] " fmt "\n", ##args)
18 18 
19-#define CHECK_ERROR(call) \19+#define CHECK_ERROR(call) \
20- do { \20+ do { \
21- aclError __ret = (call); \21+ aclError __ret = (call); \
22- if (__ret != ACL_SUCCESS) { \22+ if (__ret != ACL_SUCCESS) { \
23 ERROR_LOG("Operation failed: %s returned error code %d", #call, static_cast<int32_t>(__ret)); \23 ERROR_LOG("Operation failed: %s returned error code %d", #call, static_cast<int32_t>(__ret)); \
24- return -1; \24+ return -1; \
25- } \25+ } \
26 } while (0)26 } while (0)
27- 27+ 
28-#define CHECK_ERROR_WITHOUT_RETURN(call) \28+#define CHECK_ERROR_WITHOUT_RETURN(call) \
29- do { \29+ do { \
30- aclError __ret = (call); \30+ aclError __ret = (call); \
31- if (__ret != ACL_SUCCESS) { \31+ if (__ret != ACL_SUCCESS) { \
32 ERROR_LOG("Operation failed: %s returned error code %d", #call, static_cast<int32_t>(__ret)); \32 ERROR_LOG("Operation failed: %s returned error code %d", #call, static_cast<int32_t>(__ret)); \
33- } \33+ } \
34 } while (0)34 } while (0)
35 35 
36#endif36#endif
Mscripts/package/runtime/pkg_inc/aicpu_engine_struct.h+1-1
@@ -18,4 +18,4 @@
18 */18 */
19#include "aicpu/aicpu_engine_struct.h"19#include "aicpu/aicpu_engine_struct.h"
20 20 
21-#endif // AICPU_ENGINE_STRUCT_COMPAT_H21+#endif // AICPU_ENGINE_STRUCT_COMPAT_H
Mtests/depends/ge/src/ge_stub.cpp+200-207
@@ -1,207 +1,200 @@
1-/**1+/**
2- * Copyright (c) 2025 Huawei Technologies Co., Ltd.2+ * Copyright (c) 2025 Huawei Technologies Co., Ltd.
3- * This program is free software, you can redistribute it and/or modify it under the terms and conditions of3+ * This program is free software, you can redistribute it and/or modify it under the terms and conditions of
4- * CANN Open Software License Agreement Version 2.0 (the "License").4+ * CANN Open Software License Agreement Version 2.0 (the "License").
5- * Please refer to the License for details. You may not use this file except in compliance with the License.5+ * Please refer to the License for details. You may not use this file except in compliance with the License.
6- * THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED,6+ * THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED,
7- * INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE.7+ * INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE.
8- * See LICENSE in the root of the software repository for the full text of the License.8+ * See LICENSE in the root of the software repository for the full text of the License.
9- */9+ */
10- 10+ 
11-#include <cstring>11+#include <cstring>
12-#include <vector>12+#include <vector>
13-#include <string>13+#include <string>
14-#include "acl_stub.h"14+#include "acl_stub.h"
15- 15+ 
16-namespace {16+namespace {
17- std::string STAGES_STR = "[TEST][TEST]";17+std::string STAGES_STR = "[TEST][TEST]";
18-}18+}
19- 19+ 
20-uint32_t aclStub::InitializePlatformInfo()20+uint32_t aclStub::InitializePlatformInfo() { return 0; }
21-{21+ 
22- return 0;22+uint32_t aclStub::GetPlatformInfos(
23-}23+ const std::string SoCVersion, fe::PlatFormInfos& platformInfo, fe::OptionalInfos& optionalInfo)
24- 24+{
25-uint32_t aclStub::GetPlatformInfos(25+ return 0;
26- const std::string SoCVersion, fe::PlatFormInfos &platformInfo, fe::OptionalInfos &optionalInfo)26+}
27-{27+ 
28- return 0;28+uint32_t aclStub::InitRuntimePlatformInfos(const std::string& SoCVersion) { return 0; }
29-}29+ 
30- 30+uint32_t aclStub::GetRuntimePlatformInfosByDevice(const uint32_t& device_id, fe::PlatFormInfos& platform_infos)
31-uint32_t aclStub::InitRuntimePlatformInfos(const std::string &SoCVersion)31+{
32-{32+ return 0;
33- return 0;33+}
34-}34+ 
35- 35+bool aclStub::GetPlatformResWithLock(const std::string& label, std::map<std::string, std::string>& res) { return true; }
36-uint32_t aclStub::GetRuntimePlatformInfosByDevice(const uint32_t &device_id, fe::PlatFormInfos &platform_infos) {36+ 
37- return 0;37+bool aclStub::GetPlatformResWithLock(const string& label, const string& key, string& val) { return true; }
38-}38+ 
39- 39+uint32_t aclStub::UpdateRuntimePlatformInfosByDevice(const uint32_t& device_id, fe::PlatFormInfos& platform_infos)
40-bool aclStub::GetPlatformResWithLock(const std::string &label, std::map<std::string, std::string> &res) {40+{
41- return true;41+ return 0;
42-}42+}
43- 43+ 
44-bool aclStub::GetPlatformResWithLock(const string &label, const string &key, string &val)44+std::unique_ptr<const char_t[]> aclStub::GetErrMgrErrorMessage()
45-{45+{
46- return true;46+ const char* str = "default";
47-}47+ std::unique_ptr<const char[]> errMsg(new char[std::strlen(str) + 1]);
48- 48+ std::strcpy(const_cast<char*>(errMsg.get()), str);
49-uint32_t aclStub::UpdateRuntimePlatformInfosByDevice(const uint32_t &device_id, fe::PlatFormInfos &platform_infos) {49+ return errMsg;
50- return 0;50+}
51-}51+ 
52- 52+int aclStub::Init() { return 0; }
53-std::unique_ptr<const char_t[]> aclStub::GetErrMgrErrorMessage()53+ 
54-{54+int error_message::ErrMgrInit(error_message::ErrorMessageMode error_mode)
55- const char *str = "default";55+{
56- std::unique_ptr<const char[]> errMsg(new char[std::strlen(str) + 1]);56+ (void)error_mode;
57- std::strcpy(const_cast<char*>(errMsg.get()), str);57+ return MockFunctionTest::aclStubInstance().Init();
58- return errMsg;58+}
59-}59+ 
60- 60+std::unique_ptr<const char_t[]> error_message::GetErrMgrErrorMessage()
61-int aclStub::Init()61+{
62-{62+ return MockFunctionTest::aclStubInstance().GetErrMgrErrorMessage();
63- return 0;63+}
64-}64+ 
65- 65+int32_t error_message::ReportPredefinedErrMsg(const char* error_code) { return 0; }
66-int error_message::ErrMgrInit(error_message::ErrorMessageMode error_mode)66+ 
67-{67+int32_t error_message::ReportPredefinedErrMsg(
68- (void)error_mode;68+ const char* error_code, const std::vector<const char*>& key, const std::vector<const char*>& value)
69- return MockFunctionTest::aclStubInstance().Init();69+{
70-}70+ return 0;
71- 71+}
72-std::unique_ptr<const char_t[]> error_message::GetErrMgrErrorMessage()72+ 
73-{73+int32_t error_message::ReportInnerErrMsg(
74- return MockFunctionTest::aclStubInstance().GetErrMgrErrorMessage();74+ const char* file_name, const char* func, uint32_t line, const char* error_code, const char* format, ...)
75-}75+{
76- 76+ return 0;
77-int32_t error_message::ReportPredefinedErrMsg(const char *error_code) {77+}
78- return 0;78+ 
79-}79+namespace fe {
80- 80+PlatformInfoManager::PlatformInfoManager() : init_flag_(false) {}
81-int32_t error_message::ReportPredefinedErrMsg(const char *error_code, const std::vector<const char *> &key,81+ 
82- const std::vector<const char *> &value)82+PlatformInfoManager::~PlatformInfoManager() {}
83-{83+ 
84- return 0;84+PlatformInfoManager& PlatformInfoManager::GeInstance()
85-}85+{
86- 86+ static PlatformInfoManager ge_platform_info;
87-int32_t error_message::ReportInnerErrMsg(const char *file_name, const char *func, uint32_t line, const char *error_code,87+ return ge_platform_info;
88- const char *format, ...)88+}
89-{89+ 
90- return 0;90+uint32_t PlatformInfoManager::InitRuntimePlatformInfos(const std::string& SoCVersion)
91-}91+{
92- 92+ return MockFunctionTest::aclStubInstance().InitRuntimePlatformInfos(SoCVersion);
93-namespace fe {93+}
94- PlatformInfoManager::PlatformInfoManager() : init_flag_(false) {}94+ 
95- 95+uint32_t PlatformInfoManager::GetRuntimePlatformInfosByDevice(
96- PlatformInfoManager::~PlatformInfoManager() {}96+ const uint32_t& device_id, PlatFormInfos& platform_infos, bool need_deep_copy)
97- 97+{
98- PlatformInfoManager &PlatformInfoManager::GeInstance() {98+ (void)need_deep_copy;
99- static PlatformInfoManager ge_platform_info;99+ return MockFunctionTest::aclStubInstance().GetRuntimePlatformInfosByDevice(device_id, platform_infos);
100- return ge_platform_info;100+}
101- }101+ 
102- 102+uint32_t PlatformInfoManager::UpdateRuntimePlatformInfosByDevice(
103- uint32_t PlatformInfoManager::InitRuntimePlatformInfos(const std::string &SoCVersion) {103+ const uint32_t& device_id, PlatFormInfos& platform_infos)
104- return MockFunctionTest::aclStubInstance().InitRuntimePlatformInfos(SoCVersion);104+{
105- }105+ return MockFunctionTest::aclStubInstance().UpdateRuntimePlatformInfosByDevice(device_id, platform_infos);
106- 106+}
107- uint32_t PlatformInfoManager::GetRuntimePlatformInfosByDevice(const uint32_t &device_id,107+ 
108- PlatFormInfos &platform_infos,108+uint32_t fe::PlatformInfoManager::InitializePlatformInfo()
109- bool need_deep_copy) {109+{
110- (void) need_deep_copy;110+ return MockFunctionTest::aclStubInstance().InitializePlatformInfo();
111- return MockFunctionTest::aclStubInstance().GetRuntimePlatformInfosByDevice(device_id, platform_infos);111+}
112- }112+ 
113- 113+uint32_t fe::PlatformInfoManager::GetPlatformInfos(
114- uint32_t PlatformInfoManager::UpdateRuntimePlatformInfosByDevice(const uint32_t &device_id,114+ const std::string SoCVersion, fe::PlatFormInfos& platformInfo, fe::OptionalInfos& optionalInfo)
115- PlatFormInfos &platform_infos) {115+{
116- return MockFunctionTest::aclStubInstance().UpdateRuntimePlatformInfosByDevice(device_id, platform_infos);116+ return MockFunctionTest::aclStubInstance().GetPlatformInfos(SoCVersion, platformInfo, optionalInfo);
117- }117+}
118- 118+ 
119- uint32_t fe::PlatformInfoManager::InitializePlatformInfo()119+bool PlatFormInfos::GetPlatformResWithLock(const std::string& label, std::map<std::string, std::string>& res)
120- {120+{
121- return MockFunctionTest::aclStubInstance().InitializePlatformInfo();121+ return MockFunctionTest::aclStubInstance().GetPlatformResWithLock(label, res);
122- }122+}
123- 123+ 
124- uint32_t fe::PlatformInfoManager::GetPlatformInfos(124+bool PlatFormInfos::GetPlatformResWithLock(const string& label, const string& key, string& val)
125- const std::string SoCVersion, fe::PlatFormInfos &platformInfo, fe::OptionalInfos &optionalInfo)125+{
126- {126+ return MockFunctionTest::aclStubInstance().GetPlatformResWithLock(label, key, val);
127- return MockFunctionTest::aclStubInstance().GetPlatformInfos(SoCVersion, platformInfo, optionalInfo);127+}
128- }128+ 
129- 129+void PlatFormInfos::SetPlatformResWithLock(const std::string& label, std::map<std::string, std::string>& res)
130- bool PlatFormInfos::GetPlatformResWithLock(const std::string &label, std::map<std::string, std::string> &res) {130+{
131- return MockFunctionTest::aclStubInstance().GetPlatformResWithLock(label, res);131+ return;
132- }132+}
133- 133+} // namespace fe
134- bool PlatFormInfos::GetPlatformResWithLock(const string &label, const string &key, string &val)134+ 
135- {135+MockFunctionTest::MockFunctionTest() { ResetToDefaultMock(); }
136- return MockFunctionTest::aclStubInstance().GetPlatformResWithLock(label, key, val);136+ 
137- }137+MockFunctionTest& MockFunctionTest::aclStubInstance()
138- 138+{
139- void PlatFormInfos::SetPlatformResWithLock(const std::string &label, std::map<std::string, std::string> &res) {139+ static MockFunctionTest stub;
140- return;140+ return stub;
141- }141+};
142-}142+ 
143- 143+void MockFunctionTest::ResetToDefaultMock()
144- 144+{
145-MockFunctionTest::MockFunctionTest()145+ // delegates the default actions of the RTS methods to aclStub
146-{146+ ON_CALL(*this, rtDvppMallocWithFlag)
147- ResetToDefaultMock();147+ .WillByDefault([this](void** devPtr, uint64_t size, uint32_t flag, uint16_t moduleId) {
148-}148+ return aclStub::rtDvppMallocWithFlag(devPtr, size, flag, moduleId);
149- 149+ });
150-MockFunctionTest& MockFunctionTest::aclStubInstance()150+ ON_CALL(*this, rtDvppMalloc).WillByDefault([this](void** devPtr, uint64_t size, uint16_t moduleId) {
151-{151+ return aclStub::rtDvppMalloc(devPtr, size, moduleId);
152- static MockFunctionTest stub;152+ });
153- return stub;153+ ON_CALL(*this, rtDvppFree).WillByDefault([this](void* devPtr) { return aclStub::rtDvppFree(devPtr); });
154-};154+ ON_CALL(*this, rtMalloc).WillByDefault([this](void** devPtr, uint64_t size, rtMemType_t type, uint16_t moduleId) {
155- 155+ return aclStub::rtMalloc(devPtr, size, type, moduleId);
156-void MockFunctionTest::ResetToDefaultMock() {156+ });
157- // delegates the default actions of the RTS methods to aclStub157+ ON_CALL(*this, rtMemAllocManaged)
158- ON_CALL(*this, rtDvppMallocWithFlag).WillByDefault([this](void **devPtr, uint64_t size, uint32_t flag, uint16_t moduleId) {158+ .WillByDefault([this](void** ptr, uint64_t size, uint32_t flag, const uint16_t moduleId) {
159- return aclStub::rtDvppMallocWithFlag(devPtr, size, flag, moduleId);});159+ return aclStub::rtMemAllocManaged(ptr, size, flag, moduleId);
160- ON_CALL(*this, rtDvppMalloc).WillByDefault([this](void **devPtr, uint64_t size, uint16_t moduleId) {160+ });
161- return aclStub::rtDvppMalloc(devPtr, size, moduleId);161+ ON_CALL(*this, rtMemManagedAdvise)
162- });162+ .WillByDefault([this](const void* const ptr, uint64_t size, uint16_t advise, rtMemManagedLocation location) {
163- ON_CALL(*this, rtDvppFree).WillByDefault([this](void *devPtr) {163+ return aclStub::rtMemManagedAdvise(ptr, size, advise, location);
164- return aclStub::rtDvppFree(devPtr);164+ });
165- });165+ ON_CALL(*this, rtMemManagedGetAttr)
166- ON_CALL(*this, rtMalloc).WillByDefault([this](void **devPtr, uint64_t size, rtMemType_t type, uint16_t moduleId) {166+ .WillByDefault(
167- return aclStub::rtMalloc(devPtr, size, type, moduleId);167+ [this](rtMemManagedRangeAttribute attribute, const void* ptr, size_t size, void* data, size_t dataSize) {
168- });168+ return aclStub::rtMemManagedGetAttr(attribute, ptr, size, data, dataSize);
169- ON_CALL(*this, rtMemAllocManaged).WillByDefault([this](void **ptr, uint64_t size, uint32_t flag, const uint16_t moduleId) {169+ });
170- return aclStub::rtMemAllocManaged(ptr, size, flag, moduleId);170+ ON_CALL(*this, rtMemManagedGetAttrs)
171- });171+ .WillByDefault([this](
172- ON_CALL(*this, rtMemManagedAdvise).WillByDefault([this](const void *const ptr, uint64_t size, uint16_t advise, rtMemManagedLocation location) {172+ rtMemManagedRangeAttribute* attributes, size_t numAttributes, const void* ptr, size_t size,
173- return aclStub::rtMemManagedAdvise(ptr, size, advise, location);173+ void** data, size_t* dataSizes) {
174- });174+ return aclStub::rtMemManagedGetAttrs(attributes, numAttributes, ptr, size, data, dataSizes);
175- ON_CALL(*this, rtMemManagedGetAttr).WillByDefault([this](rtMemManagedRangeAttribute attribute, const void *ptr, size_t size, void *data, size_t dataSize) {175+ });
176- return aclStub::rtMemManagedGetAttr(attribute, ptr, size, data, dataSize);176+ ON_CALL(*this, rtFree).WillByDefault([this](void* devPtr) { return aclStub::rtFree(devPtr); });
177- });177+ ON_CALL(*this, rtMallocHost).WillByDefault([this](void** hostPtr, uint64_t size, uint16_t moduleId) {
178- ON_CALL(*this, rtMemManagedGetAttrs).WillByDefault([this](rtMemManagedRangeAttribute *attributes, size_t numAttributes, const void *ptr, 178+ return aclStub::rtMallocHost(hostPtr, size, moduleId);
179- size_t size, void **data, size_t *dataSizes) {179+ });
180- return aclStub::rtMemManagedGetAttrs(attributes, numAttributes, ptr, size, data, dataSizes);180+ ON_CALL(*this, rtFreeHost).WillByDefault([this](void* devPtr) { return aclStub::rtFreeHost(devPtr); });
181- });181+ ON_CALL(*this, rtFreeWithDevSync).WillByDefault([this](void* devPtr) {
182- ON_CALL(*this, rtFree).WillByDefault([this](void *devPtr) {182+ return aclStub::rtFreeWithDevSync(devPtr);
183- return aclStub::rtFree(devPtr);183+ });
184- });184+ ON_CALL(*this, rtFreeHostWithDevSync).WillByDefault([this](void* hostPtr) {
185- ON_CALL(*this, rtMallocHost).WillByDefault([this](void **hostPtr, uint64_t size, uint16_t moduleId) {185+ return aclStub::rtFreeHostWithDevSync(hostPtr);
186- return aclStub::rtMallocHost(hostPtr, size, moduleId);186+ });
187- });187+ ON_CALL(*this, rtMallocCached)
188- ON_CALL(*this, rtFreeHost).WillByDefault([this](void *devPtr) {188+ .WillByDefault([this](void** devPtr, uint64_t size, rtMemType_t type, uint16_t moduleId) {
189- return aclStub::rtFreeHost(devPtr);189+ return aclStub::rtMallocCached(devPtr, size, type, moduleId);
190- });190+ });
191- ON_CALL(*this, rtFreeWithDevSync).WillByDefault([this](void *devPtr) {191+ ON_CALL(*this, rtMemcpy)
192- return aclStub::rtFreeWithDevSync(devPtr);192+ .WillByDefault([this](void* dst, uint64_t destMax, const void* src, uint64_t count, rtMemcpyKind_t kind) {
193- });193+ return aclStub::rtMemcpy(dst, destMax, src, count, kind);
194- ON_CALL(*this, rtFreeHostWithDevSync).WillByDefault([this](void *hostPtr) {194+ });
195- return aclStub::rtFreeHostWithDevSync(hostPtr);195+ ON_CALL(*this, GetErrMgrErrorMessage).WillByDefault([this]() { return aclStub::GetErrMgrErrorMessage(); });
196- });196+ ON_CALL(*this, rtHostMemMapCapabilities)
197- ON_CALL(*this, rtMallocCached).WillByDefault([this](void **devPtr, uint64_t size, rtMemType_t type, uint16_t moduleId) {197+ .WillByDefault([this](uint32_t deviceId, rtHacType hacType, rtHostMemMapCapability* capabilities) {
198- return aclStub::rtMallocCached(devPtr, size, type, moduleId);});198+ return aclStub::rtHostMemMapCapabilities(deviceId, hacType, capabilities);
199- ON_CALL(*this, rtMemcpy).WillByDefault([this](void *dst, uint64_t destMax, const void *src, uint64_t count, rtMemcpyKind_t kind) {199+ });
200- return aclStub::rtMemcpy(dst, destMax, src, count, kind);});200+}
201- ON_CALL(*this, GetErrMgrErrorMessage).WillByDefault([this]() {
202- return aclStub::GetErrMgrErrorMessage();});
203- ON_CALL(*this, rtHostMemMapCapabilities).WillByDefault([this](uint32_t deviceId, rtHacType hacType, rtHostMemMapCapability *capabilities) {
204- return aclStub::rtHostMemMapCapabilities(deviceId, hacType, capabilities);});
205-}
206- 
207-