Star725
1026
代码
代码
Issues60
Pull Requests202
Actions
项目讨论
Wiki
分析
Star725
1026
已合并
add tensor detail info for tools #219
好好说话95号创建于 1月31日
add tensor detail info for tools #219
已合并
好好说话95号创建于 1月31日
从huxj153/asc-devkit-tensormaster合入到cann/asc-devkitmaster
共 19 个文件变更+4844-1
Mimpl/basic_api/kernel_operator_data_copy_intf_impl.h+76-0
@@ -16,6 +16,7 @@
16#define ASCENDC_MODULE_OPERATOR_DATA_COPY_INTERFACE_IMPL_H16#define ASCENDC_MODULE_OPERATOR_DATA_COPY_INTERFACE_IMPL_H
17#include "kernel_tensor.h"17#include "kernel_tensor.h"
18#include "kernel_process_lock.h"18#include "kernel_process_lock.h"
19+#include "mstx_local_tensor_info.h"
19 20 
20#include "kernel_check.h"21#include "kernel_check.h"
21 22 
@@ -40,6 +41,9 @@ template <typename T>
40__aicore__ inline void __inout_pipe__(MTE2) DataCopy(const LocalTensor<T>& dst, const GlobalTensor<T>& src,41__aicore__ inline void __inout_pipe__(MTE2) DataCopy(const LocalTensor<T>& dst, const GlobalTensor<T>& src,
41 const DataCopyParams& repeatParams)42 const DataCopyParams& repeatParams)
42{43{
44+#ifdef __MSTX_DFX_REPORT__
45+ MstxTensor::GetMstxDataCopyInfo(dst, src, repeatParams, "DataCopy");
46+#endif
43 using PrimType = PrimT<T>;47 using PrimType = PrimT<T>;
44 const Hardware dstHWPos = GetPhyType((TPosition)dst.GetPosition());48 const Hardware dstHWPos = GetPhyType((TPosition)dst.GetPosition());
45#if ASCENDC_CPU_DEBUG49#if ASCENDC_CPU_DEBUG
@@ -240,6 +244,9 @@ template <typename T>
240__aicore__ inline __inout_pipe__(MTE3) void DataCopy(const GlobalTensor<T>& dst, const LocalTensor<T>& src,244__aicore__ inline __inout_pipe__(MTE3) void DataCopy(const GlobalTensor<T>& dst, const LocalTensor<T>& src,
241 const DataCopyParams& repeatParams)245 const DataCopyParams& repeatParams)
242{246{
247+#ifdef __MSTX_DFX_REPORT__
248+ MstxTensor::GetMstxDataCopyInfo(dst, src, repeatParams, "DataCopy");
249+#endif
243 using PrimType = PrimT<T>;250 using PrimType = PrimT<T>;
244 const Hardware srcHWPos = GetPhyType((TPosition)src.GetPosition());251 const Hardware srcHWPos = GetPhyType((TPosition)src.GetPosition());
245#ifdef ASCENDC_CPU_DEBUG252#ifdef ASCENDC_CPU_DEBUG
@@ -308,6 +315,9 @@ template <typename T>
308__aicore__ inline void DataCopy(const LocalTensor<T> &dst, const LocalTensor<T> &src,315__aicore__ inline void DataCopy(const LocalTensor<T> &dst, const LocalTensor<T> &src,
309 const DataCopyParams &repeatParams)316 const DataCopyParams &repeatParams)
310{317{
318+#ifdef __MSTX_DFX_REPORT__
319+ MstxTensor::GetMstxDataCopyInfo(dst, src, repeatParams, "DataCopy");
320+#endif
311 using PrimType = PrimT<T>;321 using PrimType = PrimT<T>;
312 322 
313 const Hardware dstHWPos = GetPhyType((TPosition)dst.GetPosition());323 const Hardware dstHWPos = GetPhyType((TPosition)dst.GetPosition());
@@ -380,6 +390,9 @@ template <typename T, typename U>
380__aicore__ inline void DataCopy(const LocalTensor<T> &dst, const LocalTensor<U> &src,390__aicore__ inline void DataCopy(const LocalTensor<T> &dst, const LocalTensor<U> &src,
381 const DataCopyParams &repeatParams)391 const DataCopyParams &repeatParams)
382{392{
393+#ifdef __MSTX_DFX_REPORT__
394+ MstxTensor::GetMstxDataCopyInfo(dst, src, repeatParams, "DataCopy");
395+#endif
383 using PrimDstType = PrimT<T>;396 using PrimDstType = PrimT<T>;
384 using PrimSrcType = PrimT<U>;397 using PrimSrcType = PrimT<U>;
385 const Hardware dstHWPos = GetPhyType((TPosition)dst.GetPosition());398 const Hardware dstHWPos = GetPhyType((TPosition)dst.GetPosition());
@@ -449,6 +462,9 @@ template <typename T, bool IsSetMask>
449__aicore__ inline __inout_pipe__(V) void Copy(const LocalTensor<T>& dst, const LocalTensor<T>& src,462__aicore__ inline __inout_pipe__(V) void Copy(const LocalTensor<T>& dst, const LocalTensor<T>& src,
450 const uint64_t mask[], const uint8_t repeatTime, const CopyRepeatParams& repeatParams)463 const uint64_t mask[], const uint8_t repeatTime, const CopyRepeatParams& repeatParams)
451{464{
465+#ifdef __MSTX_DFX_REPORT__
466+ MstxTensor::GetMstxVecCopyInfo(dst, src, mask[0], mask[1], repeatTime, repeatParams, IsSetMask, "Copy");
467+#endif
452 using PrimType = PrimT<T>;468 using PrimType = PrimT<T>;
453#if ASCENDC_CPU_DEBUG469#if ASCENDC_CPU_DEBUG
454 MaskSetter::Instance().SetMask(IsSetMask);470 MaskSetter::Instance().SetMask(IsSetMask);
@@ -465,6 +481,9 @@ template <typename T, bool IsSetMask>
465__aicore__ inline __inout_pipe__(V) void Copy(const LocalTensor<T>& dst, const LocalTensor<T>& src,481__aicore__ inline __inout_pipe__(V) void Copy(const LocalTensor<T>& dst, const LocalTensor<T>& src,
466 const uint64_t mask, const uint8_t repeatTime, const CopyRepeatParams& repeatParams)482 const uint64_t mask, const uint8_t repeatTime, const CopyRepeatParams& repeatParams)
467{483{
484+#ifdef __MSTX_DFX_REPORT__
485+ MstxTensor::GetMstxVecCopyInfo(dst, src, mask, repeatTime, repeatParams, IsSetMask, "Copy");
486+#endif
468 using PrimType = PrimT<T>;487 using PrimType = PrimT<T>;
469#if ASCENDC_CPU_DEBUG488#if ASCENDC_CPU_DEBUG
470 MaskSetter::Instance().SetMask(IsSetMask);489 MaskSetter::Instance().SetMask(IsSetMask);
@@ -871,6 +890,9 @@ template <typename T>
871__aicore__ inline __inout_pipe__(MTE2) void DataCopy(const LocalTensor<T>& dst, const GlobalTensor<T>& src,890__aicore__ inline __inout_pipe__(MTE2) void DataCopy(const LocalTensor<T>& dst, const GlobalTensor<T>& src,
872 const DataCopyParams& intriParams, const DataCopyEnhancedParams& enhancedParams)891 const DataCopyParams& intriParams, const DataCopyEnhancedParams& enhancedParams)
873{892{
893+#ifdef __MSTX_DFX_REPORT__
894+ MstxTensor::GetMstxDataCopyInfo(dst, src, intriParams, "DataCopy");
895+#endif
874 using PrimType = PrimT<T>;896 using PrimType = PrimT<T>;
875 const Hardware dstHWPos = GetPhyType((TPosition)dst.GetPosition());897 const Hardware dstHWPos = GetPhyType((TPosition)dst.GetPosition());
876 ASCENDC_REPORT_OVERFLOW_MEM((CheckDataCopyTensorSizeOverflow(dst, src, intriParams, enhancedParams)));898 ASCENDC_REPORT_OVERFLOW_MEM((CheckDataCopyTensorSizeOverflow(dst, src, intriParams, enhancedParams)));
@@ -906,6 +928,9 @@ template <typename T>
906__aicore__ inline __inout_pipe__(MTE3) void DataCopy(const GlobalTensor<T>& dst, const LocalTensor<T>& src,928__aicore__ inline __inout_pipe__(MTE3) void DataCopy(const GlobalTensor<T>& dst, const LocalTensor<T>& src,
907 const DataCopyParams& intriParams, const DataCopyEnhancedParams& enhancedParams)929 const DataCopyParams& intriParams, const DataCopyEnhancedParams& enhancedParams)
908{930{
931+#ifdef __MSTX_DFX_REPORT__
932+ MstxTensor::GetMstxDataCopyInfo(dst, src, intriParams, "DataCopy");
933+#endif
909 using PrimType = PrimT<T>;934 using PrimType = PrimT<T>;
910 const Hardware srcHWPos = GetPhyType((TPosition)src.GetPosition());935 const Hardware srcHWPos = GetPhyType((TPosition)src.GetPosition());
911 ASCENDC_REPORT_OVERFLOW_MEM((CheckDataCopyTensorSizeOverflow(dst, src, intriParams, enhancedParams)));936 ASCENDC_REPORT_OVERFLOW_MEM((CheckDataCopyTensorSizeOverflow(dst, src, intriParams, enhancedParams)));
@@ -943,6 +968,9 @@ template <typename T>
943__aicore__ inline void DataCopy(const LocalTensor<T> &dst, const LocalTensor<T> &src,968__aicore__ inline void DataCopy(const LocalTensor<T> &dst, const LocalTensor<T> &src,
944 const DataCopyParams &intriParams, const DataCopyEnhancedParams &enhancedParams)969 const DataCopyParams &intriParams, const DataCopyEnhancedParams &enhancedParams)
945{970{
971+#ifdef __MSTX_DFX_REPORT__
972+ MstxTensor::GetMstxDataCopyInfo(dst, src, intriParams, "DataCopy");
973+#endif
946 using PrimType = PrimT<T>;974 using PrimType = PrimT<T>;
947 const Hardware dstHWPos = GetPhyType((TPosition)dst.GetPosition());975 const Hardware dstHWPos = GetPhyType((TPosition)dst.GetPosition());
948 const Hardware srcHWPos = GetPhyType((TPosition)src.GetPosition());976 const Hardware srcHWPos = GetPhyType((TPosition)src.GetPosition());
@@ -1031,6 +1059,9 @@ template <typename T, typename U, typename Std::enable_if<Std::is_same<PrimT<T>,
1031__aicore__ inline void DataCopy(const LocalTensor<T>& dst, const LocalTensor<U>& src,1059__aicore__ inline void DataCopy(const LocalTensor<T>& dst, const LocalTensor<U>& src,
1032 const DataCopyParams& intriParams, const DataCopyEnhancedParams& enhancedParams)1060 const DataCopyParams& intriParams, const DataCopyEnhancedParams& enhancedParams)
1033{1061{
1062+#ifdef __MSTX_DFX_REPORT__
1063+ MstxTensor::GetMstxDataCopyInfo(dst, src, intriParams, "DataCopy");
1064+#endif
1034 const Hardware dstHWPos = GetPhyType((TPosition)dst.GetPosition());1065 const Hardware dstHWPos = GetPhyType((TPosition)dst.GetPosition());
1035 const Hardware srcHWPos = GetPhyType((TPosition)src.GetPosition());1066 const Hardware srcHWPos = GetPhyType((TPosition)src.GetPosition());
1036 ASCENDC_REPORT_OVERFLOW_MEM((CheckDataCopyTensorSizeOverflow(dst, src, intriParams, enhancedParams)));1067 ASCENDC_REPORT_OVERFLOW_MEM((CheckDataCopyTensorSizeOverflow(dst, src, intriParams, enhancedParams)));
@@ -1058,6 +1089,9 @@ template <typename T, typename U, typename Std::enable_if<Std::is_same<PrimT<T>,
1058__aicore__ inline void DataCopy(const LocalTensor<T>& dst, const LocalTensor<U>& src,1089__aicore__ inline void DataCopy(const LocalTensor<T>& dst, const LocalTensor<U>& src,
1059 const DataCopyParams& intriParams, const DataCopyEnhancedParams& enhancedParams)1090 const DataCopyParams& intriParams, const DataCopyEnhancedParams& enhancedParams)
1060{1091{
1092+#ifdef __MSTX_DFX_REPORT__
1093+ MstxTensor::GetMstxDataCopyInfo(dst, src, intriParams, "DataCopy");
1094+#endif
1061 const Hardware dstHWPos = GetPhyType((TPosition)dst.GetPosition());1095 const Hardware dstHWPos = GetPhyType((TPosition)dst.GetPosition());
1062 const Hardware srcHWPos = GetPhyType((TPosition)src.GetPosition());1096 const Hardware srcHWPos = GetPhyType((TPosition)src.GetPosition());
1063 ASCENDC_REPORT_OVERFLOW_MEM((CheckDataCopyTensorSizeOverflow(dst, src, intriParams, enhancedParams)));1097 ASCENDC_REPORT_OVERFLOW_MEM((CheckDataCopyTensorSizeOverflow(dst, src, intriParams, enhancedParams)));
@@ -1103,6 +1137,9 @@ template <typename T, typename U, typename Std::enable_if<Std::is_same<PrimT<T>,
1103__aicore__ inline __inout_pipe__(V) void DataCopy(const LocalTensor<T> &dst, const LocalTensor<U> &src,1137__aicore__ inline __inout_pipe__(V) void DataCopy(const LocalTensor<T> &dst, const LocalTensor<U> &src,
1104 const DataCopyParams &intriParams, const DataCopyEnhancedParams &enhancedParams)1138 const DataCopyParams &intriParams, const DataCopyEnhancedParams &enhancedParams)
1105{1139{
1140+#ifdef __MSTX_DFX_REPORT__
1141+ MstxTensor::GetMstxDataCopyInfo(dst, src, intriParams, "DataCopy");
1142+#endif
1106 CheckTensorL0C2UB(dst, src);1143 CheckTensorL0C2UB(dst, src);
1107 ASCENDC_REPORT_OVERFLOW_MEM((CheckDataCopyTensorSizeOverflow(dst, src, intriParams, enhancedParams)));1144 ASCENDC_REPORT_OVERFLOW_MEM((CheckDataCopyTensorSizeOverflow(dst, src, intriParams, enhancedParams)));
1108 DataCopyL0C2UBImpl((__ubuf__ half*)dst.GetPhyAddr(), (__cc__ int32_t*)src.GetPhyAddr(), intriParams,1145 DataCopyL0C2UBImpl((__ubuf__ half*)dst.GetPhyAddr(), (__cc__ int32_t*)src.GetPhyAddr(), intriParams,
@@ -1115,6 +1152,9 @@ template <typename T, typename U, typename Std::enable_if<Std::is_same<PrimT<T>,
1115__aicore__ inline __inout_pipe__(V) void DataCopy(const LocalTensor<T> &dst, const LocalTensor<U> &src,1152__aicore__ inline __inout_pipe__(V) void DataCopy(const LocalTensor<T> &dst, const LocalTensor<U> &src,
1116 const DataCopyParams &intriParams, const DataCopyEnhancedParams &enhancedParams)1153 const DataCopyParams &intriParams, const DataCopyEnhancedParams &enhancedParams)
1117{1154{
1155+#ifdef __MSTX_DFX_REPORT__
1156+ MstxTensor::GetMstxDataCopyInfo(dst, src, intriParams, "DataCopy");
1157+#endif
1118 CheckTensorL0C2UB(dst, src);1158 CheckTensorL0C2UB(dst, src);
1119 ASCENDC_REPORT_OVERFLOW_MEM((CheckDataCopyTensorSizeOverflow(dst, src, intriParams, enhancedParams)));1159 ASCENDC_REPORT_OVERFLOW_MEM((CheckDataCopyTensorSizeOverflow(dst, src, intriParams, enhancedParams)));
1120 DataCopyL0C2UBImpl((__ubuf__ int16_t*)dst.GetPhyAddr(), (__cc__ int32_t*)src.GetPhyAddr(), intriParams,1160 DataCopyL0C2UBImpl((__ubuf__ int16_t*)dst.GetPhyAddr(), (__cc__ int32_t*)src.GetPhyAddr(), intriParams,
@@ -1127,6 +1167,9 @@ template <typename T, typename U, typename Std::enable_if<Std::is_same<PrimT<T>,
1127__aicore__ inline __inout_pipe__(V) void DataCopy(const LocalTensor<T> &dst, const LocalTensor<U> &src,1167__aicore__ inline __inout_pipe__(V) void DataCopy(const LocalTensor<T> &dst, const LocalTensor<U> &src,
1128 const DataCopyParams &intriParams, const DataCopyEnhancedParams &enhancedParams)1168 const DataCopyParams &intriParams, const DataCopyEnhancedParams &enhancedParams)
1129{1169{
1170+#ifdef __MSTX_DFX_REPORT__
1171+ MstxTensor::GetMstxDataCopyInfo(dst, src, intriParams, "DataCopy");
1172+#endif
1130 CheckTensorL0C2UB(dst, src);1173 CheckTensorL0C2UB(dst, src);
1131 ASCENDC_REPORT_OVERFLOW_MEM((CheckDataCopyTensorSizeOverflow(dst, src, intriParams, enhancedParams)));1174 ASCENDC_REPORT_OVERFLOW_MEM((CheckDataCopyTensorSizeOverflow(dst, src, intriParams, enhancedParams)));
1132 DataCopyL0C2UBImpl((__ubuf__ int8_t*)dst.GetPhyAddr(), (__cc__ int32_t*)src.GetPhyAddr(), intriParams,1175 DataCopyL0C2UBImpl((__ubuf__ int8_t*)dst.GetPhyAddr(), (__cc__ int32_t*)src.GetPhyAddr(), intriParams,
@@ -1139,6 +1182,9 @@ template <typename T, typename U, typename Std::enable_if<Std::is_same<PrimT<T>,
1139__aicore__ inline __inout_pipe__(V) void DataCopy(const LocalTensor<T> &dst, const LocalTensor<U> &src,1182__aicore__ inline __inout_pipe__(V) void DataCopy(const LocalTensor<T> &dst, const LocalTensor<U> &src,
1140 const DataCopyParams &intriParams, const DataCopyEnhancedParams &enhancedParams)1183 const DataCopyParams &intriParams, const DataCopyEnhancedParams &enhancedParams)
1141{1184{
1185+#ifdef __MSTX_DFX_REPORT__
1186+ MstxTensor::GetMstxDataCopyInfo(dst, src, intriParams, "DataCopy");
1187+#endif
1142 CheckTensorL0C2UB(dst, src);1188 CheckTensorL0C2UB(dst, src);
1143 ASCENDC_REPORT_OVERFLOW_MEM((CheckDataCopyTensorSizeOverflow(dst, src, intriParams, enhancedParams)));1189 ASCENDC_REPORT_OVERFLOW_MEM((CheckDataCopyTensorSizeOverflow(dst, src, intriParams, enhancedParams)));
1144 DataCopyL0C2UBImpl((__ubuf__ uint8_t*)dst.GetPhyAddr(), (__cc__ int32_t*)src.GetPhyAddr(), intriParams,1190 DataCopyL0C2UBImpl((__ubuf__ uint8_t*)dst.GetPhyAddr(), (__cc__ int32_t*)src.GetPhyAddr(), intriParams,
@@ -1151,6 +1197,9 @@ template <typename T, typename U, typename Std::enable_if<Std::is_same<PrimT<T>,
1151__aicore__ inline __inout_pipe__(V) void DataCopy(const LocalTensor<T> &dst, const LocalTensor<U> &src,1197__aicore__ inline __inout_pipe__(V) void DataCopy(const LocalTensor<T> &dst, const LocalTensor<U> &src,
1152 const DataCopyParams &intriParams, const DataCopyEnhancedParams &enhancedParams)1198 const DataCopyParams &intriParams, const DataCopyEnhancedParams &enhancedParams)
1153{1199{
1200+#ifdef __MSTX_DFX_REPORT__
1201+ MstxTensor::GetMstxDataCopyInfo(dst, src, intriParams, "DataCopy");
1202+#endif
1154 CheckTensorPos<U>(src, Hardware::UB, "src", "CO2",1203 CheckTensorPos<U>(src, Hardware::UB, "src", "CO2",
1155 "DataCopy from LocalTensor(CO2) to LocalTensor(CO1) with DataCopyEnhancedParams");1204 "DataCopy from LocalTensor(CO2) to LocalTensor(CO1) with DataCopyEnhancedParams");
1156 CheckTensorPos<T>(dst, Hardware::L0C, "dst", "CO1",1205 CheckTensorPos<T>(dst, Hardware::L0C, "dst", "CO1",
@@ -1165,6 +1214,9 @@ template <typename T, PaddingMode mode>
1165__aicore__ inline __inout_pipe__(MTE2) void DataCopyPad(const LocalTensor<T> &dst,1214__aicore__ inline __inout_pipe__(MTE2) void DataCopyPad(const LocalTensor<T> &dst,
1166 const GlobalTensor<T> &src, const DataCopyParams &dataCopyParams, const DataCopyPadParams &padParams)1215 const GlobalTensor<T> &src, const DataCopyParams &dataCopyParams, const DataCopyPadParams &padParams)
1167{1216{
1217+#ifdef __MSTX_DFX_REPORT__
1218+ MstxTensor::GetMstxDataCopyPadInfo(dst, src, dataCopyParams, padParams, "DataCopyPad");
1219+#endif
1168 using PrimType = PrimT<T>;1220 using PrimType = PrimT<T>;
1169#if ASCENDC_CPU_DEBUG1221#if ASCENDC_CPU_DEBUG
1170 if (!CheckFuncDataCopyPad(dst, src, dataCopyParams, padParams, "DataCopyPad from GM to VECIN/VECOUT")) {1222 if (!CheckFuncDataCopyPad(dst, src, dataCopyParams, padParams, "DataCopyPad from GM to VECIN/VECOUT")) {
@@ -1193,6 +1245,9 @@ template <typename T, PaddingMode mode>
1193__aicore__ inline __inout_pipe__(MTE3) void DataCopyPad(const GlobalTensor<T> &dst,1245__aicore__ inline __inout_pipe__(MTE3) void DataCopyPad(const GlobalTensor<T> &dst,
1194 const LocalTensor<T> &src, const DataCopyParams &dataCopyParams)1246 const LocalTensor<T> &src, const DataCopyParams &dataCopyParams)
1195{1247{
1248+#ifdef __MSTX_DFX_REPORT__
1249+ MstxTensor::GetMstxDataCopyPadInfo(dst, src, dataCopyParams, "DataCopyPad");
1250+#endif
1196 using PrimType = PrimT<T>;1251 using PrimType = PrimT<T>;
1197#if (__NPU_ARCH__ != 5102)1252#if (__NPU_ARCH__ != 5102)
1198 if ASCEND_IS_AIC {1253 if ASCEND_IS_AIC {
@@ -1216,6 +1271,9 @@ template <typename T>
1216__aicore__ inline __inout_pipe__(MTE2) void DataCopyPad(const LocalTensor<T> &dst,1271__aicore__ inline __inout_pipe__(MTE2) void DataCopyPad(const LocalTensor<T> &dst,
1217 const GlobalTensor<T> &src, const DataCopyParams &dataCopyParams, const DataCopyPadParams &padParams)1272 const GlobalTensor<T> &src, const DataCopyParams &dataCopyParams, const DataCopyPadParams &padParams)
1218{1273{
1274+#ifdef __MSTX_DFX_REPORT__
1275+ MstxTensor::GetMstxDataCopyPadInfo(dst, src, dataCopyParams, padParams, "DataCopyPad");
1276+#endif
1219 using PrimType = PrimT<T>;1277 using PrimType = PrimT<T>;
1220 if ASCEND_IS_AIC {1278 if ASCEND_IS_AIC {
1221 return;1279 return;
@@ -1257,6 +1315,9 @@ template <typename T>
1257__aicore__ inline __inout_pipe__(MTE3) void DataCopyPad(const GlobalTensor<T> &dst,1315__aicore__ inline __inout_pipe__(MTE3) void DataCopyPad(const GlobalTensor<T> &dst,
1258 const LocalTensor<T> &src, const DataCopyParams &dataCopyParams)1316 const LocalTensor<T> &src, const DataCopyParams &dataCopyParams)
1259{1317{
1318+#ifdef __MSTX_DFX_REPORT__
1319+ MstxTensor::GetMstxDataCopyPadInfo(dst, src, dataCopyParams, "DataCopyPad");
1320+#endif
1260 using PrimType = PrimT<T>;1321 using PrimType = PrimT<T>;
1261 if ASCEND_IS_AIC {1322 if ASCEND_IS_AIC {
1262 return;1323 return;
@@ -1308,6 +1369,9 @@ template <typename T, PaddingMode mode>
1308__aicore__ inline __inout_pipe__(MTE2) void DataCopyPad(const LocalTensor<T> &dst,1369__aicore__ inline __inout_pipe__(MTE2) void DataCopyPad(const LocalTensor<T> &dst,
1309 const GlobalTensor<T> &src, const DataCopyExtParams &dataCopyParams, const DataCopyPadExtParams<T> &padParams)1370 const GlobalTensor<T> &src, const DataCopyExtParams &dataCopyParams, const DataCopyPadExtParams<T> &padParams)
1310{1371{
1372+#ifdef __MSTX_DFX_REPORT__
1373+ MstxTensor::GetMstxDataCopyPadInfo(dst, src, dataCopyParams, padParams, "DataCopyPad");
1374+#endif
1311#if ASCENDC_CPU_DEBUG1375#if ASCENDC_CPU_DEBUG
1312 if (!CheckFuncDataCopyPad(dst, src, dataCopyParams, padParams, "DataCopyPad from GM to VECIN/VECOUT")) {1376 if (!CheckFuncDataCopyPad(dst, src, dataCopyParams, padParams, "DataCopyPad from GM to VECIN/VECOUT")) {
1313 ASCENDC_REPORT_CHECK_ERROR("DataCopyPad from GM to VECIN / VECOUT", KernelFuncType::NONE_MODE);1377 ASCENDC_REPORT_CHECK_ERROR("DataCopyPad from GM to VECIN / VECOUT", KernelFuncType::NONE_MODE);
@@ -1333,6 +1397,9 @@ template <typename T>
1333__aicore__ inline __inout_pipe__(MTE2) void DataCopyPad(const LocalTensor<T> &dst,1397__aicore__ inline __inout_pipe__(MTE2) void DataCopyPad(const LocalTensor<T> &dst,
1334 const GlobalTensor<T> &src, const DataCopyExtParams &dataCopyParams, const DataCopyPadExtParams<T> &padParams)1398 const GlobalTensor<T> &src, const DataCopyExtParams &dataCopyParams, const DataCopyPadExtParams<T> &padParams)
1335{1399{
1400+#ifdef __MSTX_DFX_REPORT__
1401+ MstxTensor::GetMstxDataCopyPadInfo(dst, src, dataCopyParams, padParams, "DataCopyPad");
1402+#endif
1336 if ASCEND_IS_AIC {1403 if ASCEND_IS_AIC {
1337 return;1404 return;
1338 }1405 }
@@ -1377,6 +1444,9 @@ template <typename T, typename U, typename Std::enable_if<Std::is_same<PrimT<T>,
1377__aicore__ inline __inout_pipe__(MTE2) void DataCopyPad(const LocalTensor<T> &dst,1444__aicore__ inline __inout_pipe__(MTE2) void DataCopyPad(const LocalTensor<T> &dst,
1378 const GlobalTensor<T> &src, const DataCopyExtParams &dataCopyParams, const DataCopyPadExtParams<U> &padParams)1445 const GlobalTensor<T> &src, const DataCopyExtParams &dataCopyParams, const DataCopyPadExtParams<U> &padParams)
1379{1446{
1447+#ifdef __MSTX_DFX_REPORT__
1448+ MstxTensor::GetMstxDataCopyPadInfo(dst, src, dataCopyParams, padParams, "DataCopyPad");
1449+#endif
1380 using PrimType = PrimT<T>;1450 using PrimType = PrimT<T>;
1381 if ASCEND_IS_AIC {1451 if ASCEND_IS_AIC {
1382 return;1452 return;
@@ -1397,6 +1467,9 @@ template <typename T, PaddingMode mode>
1397__aicore__ inline __inout_pipe__(MTE3) void DataCopyPad(const GlobalTensor<T> &dst,1467__aicore__ inline __inout_pipe__(MTE3) void DataCopyPad(const GlobalTensor<T> &dst,
1398 const LocalTensor<T> &src, const DataCopyExtParams &dataCopyParams)1468 const LocalTensor<T> &src, const DataCopyExtParams &dataCopyParams)
1399{1469{
1470+#ifdef __MSTX_DFX_REPORT__
1471+ MstxTensor::GetMstxDataCopyPadInfo(dst, src, dataCopyParams, "DataCopyPad");
1472+#endif
1400 using PrimType = PrimT<T>;1473 using PrimType = PrimT<T>;
1401#if (__NPU_ARCH__ != 5102)1474#if (__NPU_ARCH__ != 5102)
1402 if ASCEND_IS_AIC {1475 if ASCEND_IS_AIC {
@@ -1420,6 +1493,9 @@ template <typename T>
1420__aicore__ inline __inout_pipe__(MTE3) void DataCopyPad(const GlobalTensor<T> &dst,1493__aicore__ inline __inout_pipe__(MTE3) void DataCopyPad(const GlobalTensor<T> &dst,
1421 const LocalTensor<T> &src, const DataCopyExtParams &dataCopyParams)1494 const LocalTensor<T> &src, const DataCopyExtParams &dataCopyParams)
1422{1495{
1496+#ifdef __MSTX_DFX_REPORT__
1497+ MstxTensor::GetMstxDataCopyPadInfo(dst, src, dataCopyParams, "DataCopyPad");
1498+#endif
1423 using PrimType = PrimT<T>;1499 using PrimType = PrimT<T>;
1424 if ASCEND_IS_AIC {1500 if ASCEND_IS_AIC {
1425 return;1501 return;
Mimpl/basic_api/kernel_operator_vec_bilinearinterpolation_intf_impl.h+9-0
@@ -16,6 +16,7 @@
16#define ASCENDC_MODULE_OPERATOR_VEC_BILINEARINTERPOLATION_INTERFACE_IMPL_H16#define ASCENDC_MODULE_OPERATOR_VEC_BILINEARINTERPOLATION_INTERFACE_IMPL_H
17#include "kernel_tensor.h"17#include "kernel_tensor.h"
18#include "kernel_check.h"18#include "kernel_check.h"
19+#include "mstx_local_tensor_info.h"
19 20 
20#if __NPU_ARCH__ == 100121#if __NPU_ARCH__ == 1001
21#include "dav_c100/kernel_operator_vec_bilinearinterpolation_impl.h"22#include "dav_c100/kernel_operator_vec_bilinearinterpolation_impl.h"
@@ -70,6 +71,10 @@ __aicore__ inline void BilinearInterpolation(const LocalTensor<T> &dst, const Lo
70 bool repeatMode, uint16_t dstBlkStride, uint16_t vROffset, uint8_t vRepeat,71 bool repeatMode, uint16_t dstBlkStride, uint16_t vROffset, uint8_t vRepeat,
71 const LocalTensor<uint8_t> &sharedTmpBuffer)72 const LocalTensor<uint8_t> &sharedTmpBuffer)
72{73{
74+#ifdef __MSTX_DFX_REPORT__
75+ MstxTensor::GetMstxVecBilinearInterpolationInfo(dst, src0, src0Offset, src1, mask, hRepeat,
76+ repeatMode, dstBlkStride, vROffset, vRepeat, sharedTmpBuffer, "BilinearInterpolation");
77+#endif
73#if ASCENDC_CPU_DEBUG78#if ASCENDC_CPU_DEBUG
74 if (!CheckFuncBilinearInterpolation(dst, src0, src0Offset, src1, mask, hRepeat, repeatMode,79 if (!CheckFuncBilinearInterpolation(dst, src0, src0Offset, src1, mask, hRepeat, repeatMode,
75 dstBlkStride, vROffset, vRepeat, "BilinearInterpolation")) {80 dstBlkStride, vROffset, vRepeat, "BilinearInterpolation")) {
@@ -87,6 +92,10 @@ __aicore__ inline void BilinearInterpolation(const LocalTensor<T> &dst, const Lo
87 bool repeatMode, uint16_t dstBlkStride, uint16_t vROffset, uint8_t vRepeat,92 bool repeatMode, uint16_t dstBlkStride, uint16_t vROffset, uint8_t vRepeat,
88 const LocalTensor<uint8_t> &sharedTmpBuffer)93 const LocalTensor<uint8_t> &sharedTmpBuffer)
89{94{
95+#ifdef __MSTX_DFX_REPORT__
96+ MstxTensor::GetMstxVecBilinearInterpolationInfo(dst, src0, src0Offset, src1, mask[0], mask[1], hRepeat,
97+ repeatMode, dstBlkStride, vROffset, vRepeat, sharedTmpBuffer, "BilinearInterpolation");
98+#endif
90#if ASCENDC_CPU_DEBUG99#if ASCENDC_CPU_DEBUG
91 if (!CheckFuncBilinearInterpolation(dst, src0, src0Offset, src1, mask, hRepeat, repeatMode,100 if (!CheckFuncBilinearInterpolation(dst, src0, src0Offset, src1, mask, hRepeat, repeatMode,
92 dstBlkStride, vROffset, vRepeat, "BilinearInterpolation")) {101 dstBlkStride, vROffset, vRepeat, "BilinearInterpolation")) {
Mimpl/basic_api/kernel_operator_vec_binary_intf_impl.h+139-0
@@ -17,6 +17,7 @@
17#include "kernel_tensor.h"17#include "kernel_tensor.h"
18#include "kernel_check.h"18#include "kernel_check.h"
19#include "kernel_struct_binary.h"19#include "kernel_struct_binary.h"
20+#include "mstx_local_tensor_info.h"
20 21 
21#if __NPU_ARCH__ == 100122#if __NPU_ARCH__ == 1001
22#include "dav_c100/kernel_operator_vec_binary_impl.h"23#include "dav_c100/kernel_operator_vec_binary_impl.h"
@@ -68,6 +69,9 @@ __aicore__ inline void Add(const LocalTensor<T>& dst, const LocalTensor<T>& src0
68 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "Add")) {69 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "Add")) {
69 ASCENDC_REPORT_CHECK_ERROR("Add", KernelFuncType::MASK_BIT_MODE);70 ASCENDC_REPORT_CHECK_ERROR("Add", KernelFuncType::MASK_BIT_MODE);
70 }71 }
72+#endif
73+#ifdef __MSTX_DFX_REPORT__
74+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, mask[0], mask[1], repeatTime, repeatParams, isSetMask, "Add");
71#endif75#endif
72 AddImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),76 AddImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),
73 (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);77 (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);
@@ -83,6 +87,9 @@ __aicore__ inline void Add(const LocalTensor<T>& dst, const LocalTensor<T>& src0
83 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "Add")) {87 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "Add")) {
84 ASCENDC_REPORT_CHECK_ERROR("Add", KernelFuncType::MASK_COUNT_MODE);88 ASCENDC_REPORT_CHECK_ERROR("Add", KernelFuncType::MASK_COUNT_MODE);
85 }89 }
90+#endif
91+#ifdef __MSTX_DFX_REPORT__
92+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, mask, repeatTime, repeatParams, isSetMask, "Add");
86#endif93#endif
87 AddImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),94 AddImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),
88 (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);95 (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);
@@ -105,6 +112,9 @@ __aicore__ inline void Add(const LocalTensor<T>& dst, const LocalTensor<T>& src0
105 if (!CheckFuncVecBinary(dst, src0, src1, count, "Add")) {112 if (!CheckFuncVecBinary(dst, src0, src1, count, "Add")) {
106 ASCENDC_REPORT_CHECK_ERROR("Add", KernelFuncType::CALCOUNT_MODE);113 ASCENDC_REPORT_CHECK_ERROR("Add", KernelFuncType::CALCOUNT_MODE);
107 }114 }
115+#endif
116+#ifdef __MSTX_DFX_REPORT__
117+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, "Add", count);
C
Cchenyiyuan2月6日

api name最好都统一放到尾巴

likedislike
108#endif118#endif
109 AddImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),119 AddImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),
110 (__ubuf__ PrimType*)src1.GetPhyAddr(), count);120 (__ubuf__ PrimType*)src1.GetPhyAddr(), count);
@@ -139,6 +149,9 @@ __aicore__ inline void Sub(const LocalTensor<T>& dst, const LocalTensor<T>& src0
139 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "Sub")) {149 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "Sub")) {
140 ASCENDC_REPORT_CHECK_ERROR("Sub", KernelFuncType::MASK_BIT_MODE);150 ASCENDC_REPORT_CHECK_ERROR("Sub", KernelFuncType::MASK_BIT_MODE);
141 }151 }
152+#endif
153+#ifdef __MSTX_DFX_REPORT__
154+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, mask[0], mask[1], repeatTime, repeatParams, isSetMask, "Sub");
142#endif155#endif
143 SubImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),156 SubImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),
144 (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);157 (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);
@@ -154,6 +167,9 @@ __aicore__ inline void Sub(const LocalTensor<T>& dst, const LocalTensor<T>& src0
154 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "Sub")) {167 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "Sub")) {
155 ASCENDC_REPORT_CHECK_ERROR("Sub", KernelFuncType::MASK_COUNT_MODE);168 ASCENDC_REPORT_CHECK_ERROR("Sub", KernelFuncType::MASK_COUNT_MODE);
156 }169 }
170+#endif
171+#ifdef __MSTX_DFX_REPORT__
172+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, mask, repeatTime, repeatParams, isSetMask, "Sub");
157#endif173#endif
158 SubImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),174 SubImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),
159 (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);175 (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);
@@ -176,6 +192,9 @@ __aicore__ inline void Sub(const LocalTensor<T>& dst, const LocalTensor<T>& src0
176 if (!CheckFuncVecBinary(dst, src0, src1, count, "Sub")) {192 if (!CheckFuncVecBinary(dst, src0, src1, count, "Sub")) {
177 ASCENDC_REPORT_CHECK_ERROR("Sub", KernelFuncType::CALCOUNT_MODE);193 ASCENDC_REPORT_CHECK_ERROR("Sub", KernelFuncType::CALCOUNT_MODE);
178 }194 }
195+#endif
196+#ifdef __MSTX_DFX_REPORT__
197+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, "Sub", count);
179#endif198#endif
180 SubImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),199 SubImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),
181 (__ubuf__ PrimType*)src1.GetPhyAddr(), count);200 (__ubuf__ PrimType*)src1.GetPhyAddr(), count);
@@ -210,6 +229,9 @@ __aicore__ inline void Mul(const LocalTensor<T>& dst, const LocalTensor<T>& src0
210 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "Mul")) {229 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "Mul")) {
211 ASCENDC_REPORT_CHECK_ERROR("Mul", KernelFuncType::MASK_BIT_MODE);230 ASCENDC_REPORT_CHECK_ERROR("Mul", KernelFuncType::MASK_BIT_MODE);
212 }231 }
232+#endif
233+#ifdef __MSTX_DFX_REPORT__
234+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, mask[0], mask[1], repeatTime, repeatParams, isSetMask, "Mul");
213#endif235#endif
214 MulImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),236 MulImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),
215 (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);237 (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);
@@ -225,6 +247,9 @@ __aicore__ inline void Mul(const LocalTensor<T>& dst, const LocalTensor<T>& src0
225 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "Mul")) {247 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "Mul")) {
226 ASCENDC_REPORT_CHECK_ERROR("Mul", KernelFuncType::MASK_COUNT_MODE);248 ASCENDC_REPORT_CHECK_ERROR("Mul", KernelFuncType::MASK_COUNT_MODE);
227 }249 }
250+#endif
251+#ifdef __MSTX_DFX_REPORT__
252+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, mask, repeatTime, repeatParams, isSetMask, "Mul");
228#endif253#endif
229 MulImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),254 MulImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),
230 (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);255 (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);
@@ -247,6 +272,9 @@ __aicore__ inline void Mul(const LocalTensor<T>& dst, const LocalTensor<T>& src0
247 if (!CheckFuncVecBinary(dst, src0, src1, count, "Mul")) {272 if (!CheckFuncVecBinary(dst, src0, src1, count, "Mul")) {
248 ASCENDC_REPORT_CHECK_ERROR("Mul", KernelFuncType::CALCOUNT_MODE);273 ASCENDC_REPORT_CHECK_ERROR("Mul", KernelFuncType::CALCOUNT_MODE);
249 }274 }
275+#endif
276+#ifdef __MSTX_DFX_REPORT__
277+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, "Mul", count);
250#endif278#endif
251 MulImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),279 MulImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),
252 (__ubuf__ PrimType*)src1.GetPhyAddr(), count);280 (__ubuf__ PrimType*)src1.GetPhyAddr(), count);
@@ -282,6 +310,9 @@ __aicore__ inline void Div(const LocalTensor<T> &dst, const LocalTensor<T> &src0
282 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "Div")) {310 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "Div")) {
283 ASCENDC_REPORT_CHECK_ERROR("Div", KernelFuncType::MASK_BIT_MODE);311 ASCENDC_REPORT_CHECK_ERROR("Div", KernelFuncType::MASK_BIT_MODE);
284 }312 }
313+#endif
314+#ifdef __MSTX_DFX_REPORT__
315+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, mask[0], mask[1], repeatTime, repeatParams, isSetMask, "Div");
285#endif316#endif
286 DivImpl<PrimType, isSetMask, config>((__ubuf__ PrimType *)dst.GetPhyAddr(),317 DivImpl<PrimType, isSetMask, config>((__ubuf__ PrimType *)dst.GetPhyAddr(),
287 (__ubuf__ PrimType *)src0.GetPhyAddr(),318 (__ubuf__ PrimType *)src0.GetPhyAddr(),
@@ -299,6 +330,9 @@ __aicore__ inline void Div(const LocalTensor<T> &dst, const LocalTensor<T> &src0
299 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "Div")) {330 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "Div")) {
300 ASCENDC_REPORT_CHECK_ERROR("Div", KernelFuncType::MASK_COUNT_MODE);331 ASCENDC_REPORT_CHECK_ERROR("Div", KernelFuncType::MASK_COUNT_MODE);
301 }332 }
333+#endif
334+#ifdef __MSTX_DFX_REPORT__
335+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, mask, repeatTime, repeatParams, isSetMask, "Div");
302#endif336#endif
303 DivImpl<PrimType, isSetMask, config>((__ubuf__ PrimType *)dst.GetPhyAddr(), 337 DivImpl<PrimType, isSetMask, config>((__ubuf__ PrimType *)dst.GetPhyAddr(),
304 (__ubuf__ PrimType *)src0.GetPhyAddr(),338 (__ubuf__ PrimType *)src0.GetPhyAddr(),
@@ -316,6 +350,9 @@ __aicore__ inline void Div(const LocalTensor<T>& dst, const LocalTensor<T>& src0
316 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "Div")) {350 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "Div")) {
317 ASCENDC_REPORT_CHECK_ERROR("Div", KernelFuncType::MASK_BIT_MODE);351 ASCENDC_REPORT_CHECK_ERROR("Div", KernelFuncType::MASK_BIT_MODE);
318 }352 }
353+#endif
354+#ifdef __MSTX_DFX_REPORT__
355+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, mask[0], mask[1], repeatTime, repeatParams, isSetMask, "Div");
319#endif356#endif
320 DivImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),357 DivImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),
321 (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);358 (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);
@@ -331,6 +368,9 @@ __aicore__ inline void Div(const LocalTensor<T>& dst, const LocalTensor<T>& src0
331 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "Div")) {368 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "Div")) {
332 ASCENDC_REPORT_CHECK_ERROR("Div", KernelFuncType::MASK_COUNT_MODE);369 ASCENDC_REPORT_CHECK_ERROR("Div", KernelFuncType::MASK_COUNT_MODE);
333 }370 }
371+#endif
372+#ifdef __MSTX_DFX_REPORT__
373+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, mask, repeatTime, repeatParams, isSetMask, "Div");
334#endif374#endif
335 DivImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),375 DivImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),
336 (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);376 (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);
@@ -355,6 +395,9 @@ __aicore__ inline void Div(const LocalTensor<T>& dst, const LocalTensor<T>& src0
355 if (!CheckFuncVecBinary(dst, src0, src1, count, "Div")) {395 if (!CheckFuncVecBinary(dst, src0, src1, count, "Div")) {
356 ASCENDC_REPORT_CHECK_ERROR("Div", KernelFuncType::CALCOUNT_MODE);396 ASCENDC_REPORT_CHECK_ERROR("Div", KernelFuncType::CALCOUNT_MODE);
357 }397 }
398+#endif
399+#ifdef __MSTX_DFX_REPORT__
400+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, "Div", count);
358#endif401#endif
359 DivImpl<PrimType, config>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),402 DivImpl<PrimType, config>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),
360 (__ubuf__ PrimType*)src1.GetPhyAddr(), count);403 (__ubuf__ PrimType*)src1.GetPhyAddr(), count);
@@ -369,6 +412,9 @@ __aicore__ inline void Div(const LocalTensor<T>& dst, const LocalTensor<T>& src0
369 if (!CheckFuncVecBinary(dst, src0, src1, count, "Div")) {412 if (!CheckFuncVecBinary(dst, src0, src1, count, "Div")) {
370 ASCENDC_REPORT_CHECK_ERROR("Div", KernelFuncType::CALCOUNT_MODE);413 ASCENDC_REPORT_CHECK_ERROR("Div", KernelFuncType::CALCOUNT_MODE);
371 }414 }
415+#endif
416+#ifdef __MSTX_DFX_REPORT__
417+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, "Div", count);
372#endif418#endif
373 DivImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),419 DivImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),
374 (__ubuf__ PrimType*)src1.GetPhyAddr(), count);420 (__ubuf__ PrimType*)src1.GetPhyAddr(), count);
@@ -405,6 +451,9 @@ __aicore__ inline void MulAddDst(const LocalTensor<T>& dst, const LocalTensor<U>
405 if (!CheckFuncVecBinaryDiffType(dst, src0, src1, mask, repeatTime, repeatParams, "MulAddDst")) {451 if (!CheckFuncVecBinaryDiffType(dst, src0, src1, mask, repeatTime, repeatParams, "MulAddDst")) {
406 ASCENDC_REPORT_CHECK_ERROR("MulAddDst", KernelFuncType::MASK_BIT_MODE);452 ASCENDC_REPORT_CHECK_ERROR("MulAddDst", KernelFuncType::MASK_BIT_MODE);
407 }453 }
454+#endif
455+#ifdef __MSTX_DFX_REPORT__
456+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, mask[0], mask[1], repeatTime, repeatParams, isSetMask, "MulAddDst");
408#endif457#endif
409 MulAddDstImpl<PrimDstType, PrimSrcType, isSetMask>((__ubuf__ PrimDstType*)dst.GetPhyAddr(),458 MulAddDstImpl<PrimDstType, PrimSrcType, isSetMask>((__ubuf__ PrimDstType*)dst.GetPhyAddr(),
410 (__ubuf__ PrimSrcType*)src0.GetPhyAddr(), (__ubuf__ PrimSrcType*)src1.GetPhyAddr(), mask, repeatTime,459 (__ubuf__ PrimSrcType*)src0.GetPhyAddr(), (__ubuf__ PrimSrcType*)src1.GetPhyAddr(), mask, repeatTime,
@@ -423,6 +472,9 @@ __aicore__ inline void MulAddDst(const LocalTensor<T>& dst, const LocalTensor<U>
423 if (!CheckFuncVecBinaryDiffType(dst, src0, src1, mask, repeatTime, repeatParams, "MulAddDst")) {472 if (!CheckFuncVecBinaryDiffType(dst, src0, src1, mask, repeatTime, repeatParams, "MulAddDst")) {
424 ASCENDC_REPORT_CHECK_ERROR("MulAddDst", KernelFuncType::MASK_COUNT_MODE);473 ASCENDC_REPORT_CHECK_ERROR("MulAddDst", KernelFuncType::MASK_COUNT_MODE);
425 }474 }
475+#endif
476+#ifdef __MSTX_DFX_REPORT__
477+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, mask, repeatTime, repeatParams, isSetMask, "MulAddDst");
426#endif478#endif
427 MulAddDstImpl<PrimDstType, PrimSrcType, isSetMask>((__ubuf__ PrimDstType*)dst.GetPhyAddr(),479 MulAddDstImpl<PrimDstType, PrimSrcType, isSetMask>((__ubuf__ PrimDstType*)dst.GetPhyAddr(),
428 (__ubuf__ PrimSrcType*)src0.GetPhyAddr(), (__ubuf__ PrimSrcType*)src1.GetPhyAddr(), mask, repeatTime,480 (__ubuf__ PrimSrcType*)src0.GetPhyAddr(), (__ubuf__ PrimSrcType*)src1.GetPhyAddr(), mask, repeatTime,
@@ -447,6 +499,9 @@ __aicore__ inline void MulAddDst(const LocalTensor<T>& dst, const LocalTensor<U>
447 if (!CheckFuncVecBinaryDiffType(dst, src0, src1, count, "MulAddDst")) {499 if (!CheckFuncVecBinaryDiffType(dst, src0, src1, count, "MulAddDst")) {
448 ASCENDC_REPORT_CHECK_ERROR("MulAddDst", KernelFuncType::CALCOUNT_MODE);500 ASCENDC_REPORT_CHECK_ERROR("MulAddDst", KernelFuncType::CALCOUNT_MODE);
449 }501 }
502+#endif
503+#ifdef __MSTX_DFX_REPORT__
504+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, "MulAddDst", count);
450#endif505#endif
451 MulAddDstImpl((__ubuf__ PrimDstType*)dst.GetPhyAddr(), (__ubuf__ PrimSrcType*)src0.GetPhyAddr(),506 MulAddDstImpl((__ubuf__ PrimDstType*)dst.GetPhyAddr(), (__ubuf__ PrimSrcType*)src0.GetPhyAddr(),
452 (__ubuf__ PrimSrcType*)src1.GetPhyAddr(), count);507 (__ubuf__ PrimSrcType*)src1.GetPhyAddr(), count);
@@ -481,6 +536,9 @@ __aicore__ inline void Max(const LocalTensor<T>& dst, const LocalTensor<T>& src0
481 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "Max")) {536 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "Max")) {
482 ASCENDC_REPORT_CHECK_ERROR("Max", KernelFuncType::MASK_BIT_MODE);537 ASCENDC_REPORT_CHECK_ERROR("Max", KernelFuncType::MASK_BIT_MODE);
483 }538 }
539+#endif
540+#ifdef __MSTX_DFX_REPORT__
541+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, mask[0], mask[1], repeatTime, repeatParams, isSetMask, "Max");
484#endif542#endif
485 MaxImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),543 MaxImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),
486 (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);544 (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);
@@ -496,6 +554,9 @@ __aicore__ inline void Max(const LocalTensor<T>& dst, const LocalTensor<T>& src0
496 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "Max")) {554 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "Max")) {
497 ASCENDC_REPORT_CHECK_ERROR("Max", KernelFuncType::MASK_COUNT_MODE);555 ASCENDC_REPORT_CHECK_ERROR("Max", KernelFuncType::MASK_COUNT_MODE);
498 }556 }
557+#endif
558+#ifdef __MSTX_DFX_REPORT__
559+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, mask, repeatTime, repeatParams, isSetMask, "Max");
499#endif560#endif
500 MaxImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),561 MaxImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),
501 (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);562 (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);
@@ -518,6 +579,9 @@ __aicore__ inline void Max(const LocalTensor<T>& dst, const LocalTensor<T>& src0
518 if (!CheckFuncVecBinary(dst, src0, src1, count, "Max")) {579 if (!CheckFuncVecBinary(dst, src0, src1, count, "Max")) {
519 ASCENDC_REPORT_CHECK_ERROR("Max", KernelFuncType::CALCOUNT_MODE);580 ASCENDC_REPORT_CHECK_ERROR("Max", KernelFuncType::CALCOUNT_MODE);
520 }581 }
582+#endif
583+#ifdef __MSTX_DFX_REPORT__
584+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, "Max", count);
521#endif585#endif
522 MaxImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),586 MaxImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),
523 (__ubuf__ PrimType*)src1.GetPhyAddr(), count);587 (__ubuf__ PrimType*)src1.GetPhyAddr(), count);
@@ -552,6 +616,9 @@ __aicore__ inline void Min(const LocalTensor<T>& dst, const LocalTensor<T>& src0
552 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "Min")) {616 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "Min")) {
553 ASCENDC_REPORT_CHECK_ERROR("Min", KernelFuncType::MASK_BIT_MODE);617 ASCENDC_REPORT_CHECK_ERROR("Min", KernelFuncType::MASK_BIT_MODE);
554 }618 }
619+#endif
620+#ifdef __MSTX_DFX_REPORT__
621+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, mask[0], mask[1], repeatTime, repeatParams, isSetMask, "Min");
555#endif622#endif
556 MinImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),623 MinImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),
557 (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);624 (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);
@@ -567,6 +634,9 @@ __aicore__ inline void Min(const LocalTensor<T>& dst, const LocalTensor<T>& src0
567 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "Min")) {634 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "Min")) {
568 ASCENDC_REPORT_CHECK_ERROR("Min", KernelFuncType::MASK_COUNT_MODE);635 ASCENDC_REPORT_CHECK_ERROR("Min", KernelFuncType::MASK_COUNT_MODE);
569 }636 }
637+#endif
638+#ifdef __MSTX_DFX_REPORT__
639+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, mask, repeatTime, repeatParams, isSetMask, "Min");
570#endif640#endif
571 MinImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),641 MinImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),
572 (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);642 (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);
@@ -589,6 +659,9 @@ __aicore__ inline void Min(const LocalTensor<T>& dst, const LocalTensor<T>& src0
589 if (!CheckFuncVecBinary(dst, src0, src1, count, "Min")) {659 if (!CheckFuncVecBinary(dst, src0, src1, count, "Min")) {
590 ASCENDC_REPORT_CHECK_ERROR("Min", KernelFuncType::CALCOUNT_MODE);660 ASCENDC_REPORT_CHECK_ERROR("Min", KernelFuncType::CALCOUNT_MODE);
591 }661 }
662+#endif
663+#ifdef __MSTX_DFX_REPORT__
664+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, "Min", count);
592#endif665#endif
593 MinImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),666 MinImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),
594 (__ubuf__ PrimType*)src1.GetPhyAddr(), count);667 (__ubuf__ PrimType*)src1.GetPhyAddr(), count);
@@ -623,6 +696,9 @@ __aicore__ inline void And(const LocalTensor<T>& dst, const LocalTensor<T>& src0
623 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "And")) {696 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "And")) {
624 ASCENDC_REPORT_CHECK_ERROR("And", KernelFuncType::MASK_BIT_MODE);697 ASCENDC_REPORT_CHECK_ERROR("And", KernelFuncType::MASK_BIT_MODE);
625 }698 }
699+#endif
700+#ifdef __MSTX_DFX_REPORT__
701+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, mask[0], mask[1], repeatTime, repeatParams, isSetMask, "And");
626#endif702#endif
627 AndImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),703 AndImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),
628 (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);704 (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);
@@ -638,6 +714,9 @@ __aicore__ inline void And(const LocalTensor<T>& dst, const LocalTensor<T>& src0
638 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "And")) {714 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "And")) {
639 ASCENDC_REPORT_CHECK_ERROR("And", KernelFuncType::MASK_COUNT_MODE);715 ASCENDC_REPORT_CHECK_ERROR("And", KernelFuncType::MASK_COUNT_MODE);
640 }716 }
717+#endif
718+#ifdef __MSTX_DFX_REPORT__
719+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, mask, repeatTime, repeatParams, isSetMask, "And");
641#endif720#endif
642 AndImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),721 AndImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),
643 (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);722 (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);
@@ -660,6 +739,9 @@ __aicore__ inline void And(const LocalTensor<T>& dst, const LocalTensor<T>& src0
660 if (!CheckFuncVecBinary(dst, src0, src1, count, "And")) {739 if (!CheckFuncVecBinary(dst, src0, src1, count, "And")) {
661 ASCENDC_REPORT_CHECK_ERROR("And", KernelFuncType::CALCOUNT_MODE);740 ASCENDC_REPORT_CHECK_ERROR("And", KernelFuncType::CALCOUNT_MODE);
662 }741 }
742+#endif
743+#ifdef __MSTX_DFX_REPORT__
744+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, "And", count);
663#endif745#endif
664 AndImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),746 AndImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),
665 (__ubuf__ PrimType*)src1.GetPhyAddr(), count);747 (__ubuf__ PrimType*)src1.GetPhyAddr(), count);
@@ -694,6 +776,9 @@ __aicore__ inline void Or(const LocalTensor<T>& dst, const LocalTensor<T>& src0,
694 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "Or")) {776 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "Or")) {
695 ASCENDC_REPORT_CHECK_ERROR("Or", KernelFuncType::MASK_BIT_MODE);777 ASCENDC_REPORT_CHECK_ERROR("Or", KernelFuncType::MASK_BIT_MODE);
696 }778 }
779+#endif
780+#ifdef __MSTX_DFX_REPORT__
781+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, mask[0], mask[1], repeatTime, repeatParams, isSetMask, "Or");
697#endif782#endif
698 OrImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),783 OrImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),
699 (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);784 (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);
@@ -709,6 +794,9 @@ __aicore__ inline void Or(const LocalTensor<T>& dst, const LocalTensor<T>& src0,
709 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "Or")) {794 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "Or")) {
710 ASCENDC_REPORT_CHECK_ERROR("Or", KernelFuncType::MASK_COUNT_MODE);795 ASCENDC_REPORT_CHECK_ERROR("Or", KernelFuncType::MASK_COUNT_MODE);
711 }796 }
797+#endif
798+#ifdef __MSTX_DFX_REPORT__
799+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, mask, repeatTime, repeatParams, isSetMask, "Or");
712#endif800#endif
713 OrImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),801 OrImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),
714 (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);802 (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);
@@ -731,6 +819,9 @@ __aicore__ inline void Or(const LocalTensor<T>& dst, const LocalTensor<T>& src0,
731 if (!CheckFuncVecBinary(dst, src0, src1, count, "Or")) {819 if (!CheckFuncVecBinary(dst, src0, src1, count, "Or")) {
732 ASCENDC_REPORT_CHECK_ERROR("Or", KernelFuncType::CALCOUNT_MODE);820 ASCENDC_REPORT_CHECK_ERROR("Or", KernelFuncType::CALCOUNT_MODE);
733 }821 }
822+#endif
823+#ifdef __MSTX_DFX_REPORT__
824+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, "Or", count);
734#endif825#endif
735 OrImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),826 OrImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),
736 (__ubuf__ PrimType*)src1.GetPhyAddr(), count);827 (__ubuf__ PrimType*)src1.GetPhyAddr(), count);
@@ -828,6 +919,9 @@ __aicore__ inline void AddRelu(const LocalTensor<T>& dst, const LocalTensor<T>&
828 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "AddRelu")) {919 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "AddRelu")) {
829 ASCENDC_REPORT_CHECK_ERROR("AddRelu", KernelFuncType::MASK_BIT_MODE);920 ASCENDC_REPORT_CHECK_ERROR("AddRelu", KernelFuncType::MASK_BIT_MODE);
830 }921 }
922+#endif
923+#ifdef __MSTX_DFX_REPORT__
924+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, mask[0], mask[1], repeatTime, repeatParams, isSetMask, "AddRelu");
831#endif925#endif
832 AddReluImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(),926 AddReluImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(),
833 (__ubuf__ PrimType*)src0.GetPhyAddr(), (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime,927 (__ubuf__ PrimType*)src0.GetPhyAddr(), (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime,
@@ -849,6 +943,9 @@ __aicore__ inline void AddRelu(const LocalTensor<T>& dst, const LocalTensor<T>&
849 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "AddRelu")) {943 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "AddRelu")) {
850 ASCENDC_REPORT_CHECK_ERROR("AddRelu", KernelFuncType::MASK_COUNT_MODE);944 ASCENDC_REPORT_CHECK_ERROR("AddRelu", KernelFuncType::MASK_COUNT_MODE);
851 }945 }
946+#endif
947+#ifdef __MSTX_DFX_REPORT__
948+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, mask, repeatTime, repeatParams, isSetMask, "AddRelu");
852#endif949#endif
853 AddReluImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(),950 AddReluImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(),
854 (__ubuf__ PrimType*)src0.GetPhyAddr(), (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime,951 (__ubuf__ PrimType*)src0.GetPhyAddr(), (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime,
@@ -869,6 +966,9 @@ __aicore__ inline void AddRelu(const LocalTensor<T>& dst, const LocalTensor<T>&
869 if (!CheckFuncVecBinary(dst, src0, src1, count, "AddRelu")) {966 if (!CheckFuncVecBinary(dst, src0, src1, count, "AddRelu")) {
870 ASCENDC_REPORT_CHECK_ERROR("AddRelu", KernelFuncType::CALCOUNT_MODE);967 ASCENDC_REPORT_CHECK_ERROR("AddRelu", KernelFuncType::CALCOUNT_MODE);
871 }968 }
969+#endif
970+#ifdef __MSTX_DFX_REPORT__
971+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, "AddRelu", count);
872#endif972#endif
873 AddReluImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),973 AddReluImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),
874 (__ubuf__ PrimType*)src1.GetPhyAddr(), count);974 (__ubuf__ PrimType*)src1.GetPhyAddr(), count);
@@ -902,6 +1002,9 @@ __aicore__ inline void AddDeqRelu(const LocalTensor<half>& dst, const LocalTenso
902 if (!CheckFuncVecBinaryDiffType(dst, src0, src1, mask, repeatTime, repeatParams, "AddDeqRelu")) {1002 if (!CheckFuncVecBinaryDiffType(dst, src0, src1, mask, repeatTime, repeatParams, "AddDeqRelu")) {
903 ASCENDC_REPORT_CHECK_ERROR("AddDeqRelu", KernelFuncType::MASK_BIT_MODE);1003 ASCENDC_REPORT_CHECK_ERROR("AddDeqRelu", KernelFuncType::MASK_BIT_MODE);
904 }1004 }
1005+#endif
1006+#ifdef __MSTX_DFX_REPORT__
1007+ MstxTensor::GetMstxVecBinaryAddReqReluInfo(dst, src0, src1, mask[0], mask[1], repeatTime, repeatParams, isSetMask, "AddDeqRelu");
905#endif1008#endif
906 AddDeqReluImpl<isSetMask>((__ubuf__ half*)dst.GetPhyAddr(), (__ubuf__ int32_t*)src0.GetPhyAddr(),1009 AddDeqReluImpl<isSetMask>((__ubuf__ half*)dst.GetPhyAddr(), (__ubuf__ int32_t*)src0.GetPhyAddr(),
907 (__ubuf__ int32_t*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);1010 (__ubuf__ int32_t*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);
@@ -921,6 +1024,9 @@ __aicore__ inline void AddDeqRelu(const LocalTensor<T>& dst, const LocalTensor<U
921 if (!CheckFuncVecBinaryDiffType(dst, src0, src1, mask, repeatTime, repeatParams, "AddDeqRelu")) {1024 if (!CheckFuncVecBinaryDiffType(dst, src0, src1, mask, repeatTime, repeatParams, "AddDeqRelu")) {
922 ASCENDC_REPORT_CHECK_ERROR("AddDeqRelu", KernelFuncType::MASK_BIT_MODE);1025 ASCENDC_REPORT_CHECK_ERROR("AddDeqRelu", KernelFuncType::MASK_BIT_MODE);
923 }1026 }
1027+#endif
1028+#ifdef __MSTX_DFX_REPORT__
1029+ MstxTensor::GetMstxVecBinaryAddReqReluInfo(dst, src0, src1, mask[0], mask[1], repeatTime, repeatParams, isSetMask, "AddDeqRelu");
924#endif1030#endif
925 AddDeqReluImpl<isSetMask>((__ubuf__ PrimDstType*)dst.GetPhyAddr(),1031 AddDeqReluImpl<isSetMask>((__ubuf__ PrimDstType*)dst.GetPhyAddr(),
926 (__ubuf__ PrimSrcType*)src0.GetPhyAddr(), (__ubuf__ PrimSrcType*)src1.GetPhyAddr(), mask, repeatTime,1032 (__ubuf__ PrimSrcType*)src0.GetPhyAddr(), (__ubuf__ PrimSrcType*)src1.GetPhyAddr(), mask, repeatTime,
@@ -937,6 +1043,9 @@ __aicore__ inline void AddDeqRelu(const LocalTensor<half> &dst, const LocalTenso
937 if (!CheckFuncVecBinaryDiffType(dst, src0, src1, mask, repeatTime, repeatParams, "AddDeqRelu")) {1043 if (!CheckFuncVecBinaryDiffType(dst, src0, src1, mask, repeatTime, repeatParams, "AddDeqRelu")) {
938 ASCENDC_REPORT_CHECK_ERROR("AddDeqRelu", KernelFuncType::MASK_COUNT_MODE);1044 ASCENDC_REPORT_CHECK_ERROR("AddDeqRelu", KernelFuncType::MASK_COUNT_MODE);
939 }1045 }
1046+#endif
1047+#ifdef __MSTX_DFX_REPORT__
1048+ MstxTensor::GetMstxVecBinaryAddReqReluInfo(dst, src0, src1, mask, repeatTime, repeatParams, isSetMask, "AddDeqRelu");
940#endif1049#endif
941 AddDeqReluImpl<isSetMask>((__ubuf__ half*)dst.GetPhyAddr(), (__ubuf__ int32_t*)src0.GetPhyAddr(),1050 AddDeqReluImpl<isSetMask>((__ubuf__ half*)dst.GetPhyAddr(), (__ubuf__ int32_t*)src0.GetPhyAddr(),
942 (__ubuf__ int32_t*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);1051 (__ubuf__ int32_t*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);
@@ -956,6 +1065,9 @@ __aicore__ inline void AddDeqRelu(const LocalTensor<T> &dst, const LocalTensor<U
956 if (!CheckFuncVecBinaryDiffType(dst, src0, src1, mask, repeatTime, repeatParams, "AddDeqRelu")) {1065 if (!CheckFuncVecBinaryDiffType(dst, src0, src1, mask, repeatTime, repeatParams, "AddDeqRelu")) {
957 ASCENDC_REPORT_CHECK_ERROR("AddDeqRelu", KernelFuncType::MASK_COUNT_MODE);1066 ASCENDC_REPORT_CHECK_ERROR("AddDeqRelu", KernelFuncType::MASK_COUNT_MODE);
958 }1067 }
1068+#endif
1069+#ifdef __MSTX_DFX_REPORT__
1070+ MstxTensor::GetMstxVecBinaryAddReqReluInfo(dst, src0, src1, mask, repeatTime, repeatParams, isSetMask, "AddDeqRelu");
959#endif1071#endif
960 AddDeqReluImpl<isSetMask>((__ubuf__ PrimDstType*)dst.GetPhyAddr(),1072 AddDeqReluImpl<isSetMask>((__ubuf__ PrimDstType*)dst.GetPhyAddr(),
961 (__ubuf__ PrimSrcType*)src0.GetPhyAddr(), (__ubuf__ PrimSrcType*)src1.GetPhyAddr(), mask, repeatTime,1073 (__ubuf__ PrimSrcType*)src0.GetPhyAddr(), (__ubuf__ PrimSrcType*)src1.GetPhyAddr(), mask, repeatTime,
@@ -977,6 +1089,9 @@ __aicore__ inline void AddDeqRelu(const LocalTensor<half>& dst, const LocalTenso
977 if (!CheckFuncVecBinaryDiffType(dst, src0, src1, count, "AddDeqRelu")) {1089 if (!CheckFuncVecBinaryDiffType(dst, src0, src1, count, "AddDeqRelu")) {
978 ASCENDC_REPORT_CHECK_ERROR("AddDeqRelu", KernelFuncType::CALCOUNT_MODE);1090 ASCENDC_REPORT_CHECK_ERROR("AddDeqRelu", KernelFuncType::CALCOUNT_MODE);
979 }1091 }
1092+#endif
1093+#ifdef __MSTX_DFX_REPORT__
1094+ MstxTensor::GetMstxVecBinaryAddReqReluInfo(dst, src0, src1, "AddDeqRelu", count);
980#endif1095#endif
981 AddDeqReluImpl((__ubuf__ half *)dst.GetPhyAddr(), (__ubuf__ int32_t *)src0.GetPhyAddr(),1096 AddDeqReluImpl((__ubuf__ half *)dst.GetPhyAddr(), (__ubuf__ int32_t *)src0.GetPhyAddr(),
982 (__ubuf__ int32_t *)src1.GetPhyAddr(), count);1097 (__ubuf__ int32_t *)src1.GetPhyAddr(), count);
@@ -994,6 +1109,9 @@ __aicore__ inline void AddDeqRelu(const LocalTensor<T>& dst, const LocalTensor<U
994 if (!CheckFuncVecBinaryDiffType(dst, src0, src1, count, "AddDeqRelu")) {1109 if (!CheckFuncVecBinaryDiffType(dst, src0, src1, count, "AddDeqRelu")) {
995 ASCENDC_REPORT_CHECK_ERROR("AddDeqRelu", KernelFuncType::CALCOUNT_MODE);1110 ASCENDC_REPORT_CHECK_ERROR("AddDeqRelu", KernelFuncType::CALCOUNT_MODE);
996 }1111 }
1112+#endif
1113+#ifdef __MSTX_DFX_REPORT__
1114+ MstxTensor::GetMstxVecBinaryAddReqReluInfo(dst, src0, src1, "AddDeqRelu", count);
997#endif1115#endif
998 AddDeqReluImpl((__ubuf__ PrimDstType*)dst.GetPhyAddr(), (__ubuf__ PrimSrcType*)src0.GetPhyAddr(),1116 AddDeqReluImpl((__ubuf__ PrimDstType*)dst.GetPhyAddr(), (__ubuf__ PrimSrcType*)src0.GetPhyAddr(),
999 (__ubuf__ PrimSrcType*)src1.GetPhyAddr(), count);1117 (__ubuf__ PrimSrcType*)src1.GetPhyAddr(), count);
@@ -1033,6 +1151,9 @@ __aicore__ inline void FusedMulAdd(const LocalTensor<T>& dst, const LocalTensor<
1033 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "FusedMulAdd")) {1151 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "FusedMulAdd")) {
1034 ASCENDC_REPORT_CHECK_ERROR("FusedMulAdd", KernelFuncType::MASK_BIT_MODE);1152 ASCENDC_REPORT_CHECK_ERROR("FusedMulAdd", KernelFuncType::MASK_BIT_MODE);
1035 }1153 }
1154+#endif
1155+#ifdef __MSTX_DFX_REPORT__
1156+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, mask[0], mask[1], repeatTime, repeatParams, isSetMask, "FusedMulAdd");
1036#endif1157#endif
1037 FusedMulAddImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(),1158 FusedMulAddImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(),
1038 (__ubuf__ PrimType*)src0.GetPhyAddr(), (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime,1159 (__ubuf__ PrimType*)src0.GetPhyAddr(), (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime,
@@ -1054,6 +1175,9 @@ __aicore__ inline void FusedMulAdd(const LocalTensor<T>& dst, const LocalTensor<
1054 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "FusedMulAdd")) {1175 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "FusedMulAdd")) {
1055 ASCENDC_REPORT_CHECK_ERROR("FusedMulAdd", KernelFuncType::MASK_COUNT_MODE);1176 ASCENDC_REPORT_CHECK_ERROR("FusedMulAdd", KernelFuncType::MASK_COUNT_MODE);
1056 }1177 }
1178+#endif
1179+#ifdef __MSTX_DFX_REPORT__
1180+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, mask, repeatTime, repeatParams, isSetMask, "FusedMulAdd");
1057#endif1181#endif
1058 FusedMulAddImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(),1182 FusedMulAddImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(),
1059 (__ubuf__ PrimType*)src0.GetPhyAddr(), (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime,1183 (__ubuf__ PrimType*)src0.GetPhyAddr(), (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime,
@@ -1082,6 +1206,9 @@ __aicore__ inline void FusedMulAdd(const LocalTensor<T>& dst, const LocalTensor<
1082 if (!CheckFuncVecBinary(dst, src0, src1, count, "FusedMulAdd")) {1206 if (!CheckFuncVecBinary(dst, src0, src1, count, "FusedMulAdd")) {
1083 ASCENDC_REPORT_CHECK_ERROR("FusedMulAdd", KernelFuncType::CALCOUNT_MODE);1207 ASCENDC_REPORT_CHECK_ERROR("FusedMulAdd", KernelFuncType::CALCOUNT_MODE);
1084 }1208 }
1209+#endif
1210+#ifdef __MSTX_DFX_REPORT__
1211+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, "FusedMulAdd", count);
1085#endif1212#endif
1086 FusedMulAddImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),1213 FusedMulAddImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),
1087 (__ubuf__ PrimType*)src1.GetPhyAddr(), count);1214 (__ubuf__ PrimType*)src1.GetPhyAddr(), count);
@@ -1234,6 +1361,9 @@ __aicore__ inline void FusedMulAddRelu(const LocalTensor<T>& dst, const LocalTen
1234 if (!CheckFuncVecBinary(dst, src0, src1, count, "FusedMulAddRelu")) {1361 if (!CheckFuncVecBinary(dst, src0, src1, count, "FusedMulAddRelu")) {
1235 ASCENDC_REPORT_CHECK_ERROR("FusedMulAddRelu", KernelFuncType::CALCOUNT_MODE);1362 ASCENDC_REPORT_CHECK_ERROR("FusedMulAddRelu", KernelFuncType::CALCOUNT_MODE);
1236 }1363 }
1364+#endif
1365+#ifdef __MSTX_DFX_REPORT__
1366+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, "FusedMulAddRelu", count);
1237#endif1367#endif
1238 FusedMulAddReluImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),1368 FusedMulAddReluImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),
1239 (__ubuf__ PrimType*)src1.GetPhyAddr(), count);1369 (__ubuf__ PrimType*)src1.GetPhyAddr(), count);
@@ -1273,6 +1403,9 @@ __aicore__ inline void SubRelu(const LocalTensor<T>& dst, const LocalTensor<T>&
1273 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "SubRelu")) {1403 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "SubRelu")) {
1274 ASCENDC_REPORT_CHECK_ERROR("SubRelu", KernelFuncType::MASK_BIT_MODE);1404 ASCENDC_REPORT_CHECK_ERROR("SubRelu", KernelFuncType::MASK_BIT_MODE);
1275 }1405 }
1406+#endif
1407+#ifdef __MSTX_DFX_REPORT__
1408+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, mask[0], mask[1], repeatTime, repeatParams, isSetMask, "SubRelu");
1276#endif1409#endif
1277 SubReluImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(),1410 SubReluImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(),
1278 (__ubuf__ PrimType*)src0.GetPhyAddr(), (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime,1411 (__ubuf__ PrimType*)src0.GetPhyAddr(), (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime,
@@ -1294,6 +1427,9 @@ __aicore__ inline void SubRelu(const LocalTensor<T>& dst, const LocalTensor<T>&
1294 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "SubRelu")) {1427 if (!CheckFuncVecBinary(dst, src0, src1, mask, repeatTime, repeatParams, "SubRelu")) {
1295 ASCENDC_REPORT_CHECK_ERROR("SubRelu", KernelFuncType::MASK_COUNT_MODE);1428 ASCENDC_REPORT_CHECK_ERROR("SubRelu", KernelFuncType::MASK_COUNT_MODE);
1296 }1429 }
1430+#endif
1431+#ifdef __MSTX_DFX_REPORT__
1432+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, mask, repeatTime, repeatParams, isSetMask, "SubRelu");
1297#endif1433#endif
1298 SubReluImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(),1434 SubReluImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(),
1299 (__ubuf__ PrimType*)src0.GetPhyAddr(), (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime,1435 (__ubuf__ PrimType*)src0.GetPhyAddr(), (__ubuf__ PrimType*)src1.GetPhyAddr(), mask, repeatTime,
@@ -1314,6 +1450,9 @@ __aicore__ inline void SubRelu(const LocalTensor<T>& dst, const LocalTensor<T>&
1314 if (!CheckFuncVecBinary(dst, src0, src1, count, "SubRelu")) {1450 if (!CheckFuncVecBinary(dst, src0, src1, count, "SubRelu")) {
1315 ASCENDC_REPORT_CHECK_ERROR("SubRelu", KernelFuncType::CALCOUNT_MODE);1451 ASCENDC_REPORT_CHECK_ERROR("SubRelu", KernelFuncType::CALCOUNT_MODE);
1316 }1452 }
1453+#endif
1454+#ifdef __MSTX_DFX_REPORT__
1455+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, "SubRelu", count);
1317#endif1456#endif
1318 SubReluImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),1457 SubReluImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),
1319 (__ubuf__ PrimType*)src1.GetPhyAddr(), count);1458 (__ubuf__ PrimType*)src1.GetPhyAddr(), count);
Mimpl/basic_api/kernel_operator_vec_binary_scalar_intf_impl.h+127-0
@@ -17,6 +17,7 @@
17#include "kernel_tensor.h"17#include "kernel_tensor.h"
18#include "kernel_check.h"18#include "kernel_check.h"
19#include "kernel_struct_unary.h"19#include "kernel_struct_unary.h"
20+#include "mstx_local_tensor_info.h"
20 21 
21#if __NPU_ARCH__ == 100122#if __NPU_ARCH__ == 1001
22#include "dav_c100/kernel_operator_vec_binary_scalar_impl.h"23#include "dav_c100/kernel_operator_vec_binary_scalar_impl.h"
@@ -64,6 +65,9 @@ __aicore__ inline void Adds(const LocalTensor<T>& dst, const LocalTensor<T>& src
64 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "Adds")) {65 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "Adds")) {
65 ASCENDC_REPORT_CHECK_ERROR("Adds", KernelFuncType::MASK_BIT_MODE);66 ASCENDC_REPORT_CHECK_ERROR("Adds", KernelFuncType::MASK_BIT_MODE);
66 }67 }
68+#endif
69+#ifdef __MSTX_DFX_REPORT__
70+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask[0], mask[1], repeatTime, repeatParams, "Adds");
67#endif71#endif
68 AddsImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue, mask,72 AddsImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue, mask,
69 repeatTime, repeatParams);73 repeatTime, repeatParams);
@@ -79,6 +83,9 @@ __aicore__ inline void Adds(const LocalTensor<T>& dst, const LocalTensor<T>& src
79 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "Adds")) {83 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "Adds")) {
80 ASCENDC_REPORT_CHECK_ERROR("Adds", KernelFuncType::MASK_BIT_MODE);84 ASCENDC_REPORT_CHECK_ERROR("Adds", KernelFuncType::MASK_BIT_MODE);
81 }85 }
86+#endif
87+#ifdef __MSTX_DFX_REPORT__
88+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask[0], mask[1], repeatTime, repeatParams, "Adds");
82#endif89#endif
83 AddsImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),90 AddsImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),
84 scalarValue, mask, repeatTime, repeatParams);91 scalarValue, mask, repeatTime, repeatParams);
@@ -93,6 +100,9 @@ __aicore__ inline void Adds(const LocalTensor<T>& dst, const LocalTensor<T>& src
93 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "Adds")) {100 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "Adds")) {
94 ASCENDC_REPORT_CHECK_ERROR("Adds", KernelFuncType::MASK_COUNT_MODE);101 ASCENDC_REPORT_CHECK_ERROR("Adds", KernelFuncType::MASK_COUNT_MODE);
95 }102 }
103+#endif
104+#ifdef __MSTX_DFX_REPORT__
105+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask, repeatTime, repeatParams, "Adds");
96#endif106#endif
97 AddsImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue, mask,107 AddsImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue, mask,
98 repeatTime, repeatParams);108 repeatTime, repeatParams);
@@ -108,6 +118,9 @@ __aicore__ inline void Adds(const LocalTensor<T>& dst, const LocalTensor<T>& src
108 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "Adds")) {118 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "Adds")) {
109 ASCENDC_REPORT_CHECK_ERROR("Adds", KernelFuncType::MASK_COUNT_MODE);119 ASCENDC_REPORT_CHECK_ERROR("Adds", KernelFuncType::MASK_COUNT_MODE);
110 }120 }
121+#endif
122+#ifdef __MSTX_DFX_REPORT__
123+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask, repeatTime, repeatParams, "Adds");
111#endif124#endif
112 AddsImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),125 AddsImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),
113 scalarValue, mask, repeatTime, repeatParams);126 scalarValue, mask, repeatTime, repeatParams);
@@ -130,6 +143,9 @@ __aicore__ inline void Adds(const LocalTensor<T>& dst, const LocalTensor<T>& src
130 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, count, "Adds")) {143 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, count, "Adds")) {
131 ASCENDC_REPORT_CHECK_ERROR("Adds", KernelFuncType::CALCOUNT_MODE);144 ASCENDC_REPORT_CHECK_ERROR("Adds", KernelFuncType::CALCOUNT_MODE);
132 }145 }
146+#endif
147+#ifdef __MSTX_DFX_REPORT__
148+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, "Adds", count);
133#endif149#endif
134 AddsImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue,150 AddsImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue,
135 count);151 count);
@@ -145,6 +161,9 @@ __aicore__ inline void Adds(const LocalTensor<T>& dst, const LocalTensor<T>& src
145 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, count, "Adds")) {161 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, count, "Adds")) {
146 ASCENDC_REPORT_CHECK_ERROR("Adds", KernelFuncType::CALCOUNT_MODE);162 ASCENDC_REPORT_CHECK_ERROR("Adds", KernelFuncType::CALCOUNT_MODE);
147 }163 }
164+#endif
165+#ifdef __MSTX_DFX_REPORT__
166+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, "Adds", count);
148#endif167#endif
149 AddsImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),168 AddsImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),
150 scalarValue, count);169 scalarValue, count);
@@ -175,6 +194,9 @@ __aicore__ inline void Muls(const LocalTensor<T>& dst, const LocalTensor<T>& src
175 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "Muls")) {194 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "Muls")) {
176 ASCENDC_REPORT_CHECK_ERROR("Muls", KernelFuncType::MASK_BIT_MODE);195 ASCENDC_REPORT_CHECK_ERROR("Muls", KernelFuncType::MASK_BIT_MODE);
177 }196 }
197+#endif
198+#ifdef __MSTX_DFX_REPORT__
199+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask[0], mask[1], repeatTime, repeatParams, "Muls");
178#endif200#endif
179 MulsImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue, mask,201 MulsImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue, mask,
180 repeatTime, repeatParams);202 repeatTime, repeatParams);
@@ -190,6 +212,9 @@ __aicore__ inline void Muls(const LocalTensor<T>& dst, const LocalTensor<T>& src
190 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "Muls")) {212 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "Muls")) {
191 ASCENDC_REPORT_CHECK_ERROR("Muls", KernelFuncType::MASK_BIT_MODE);213 ASCENDC_REPORT_CHECK_ERROR("Muls", KernelFuncType::MASK_BIT_MODE);
192 }214 }
215+#endif
216+#ifdef __MSTX_DFX_REPORT__
217+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask[0], mask[1], repeatTime, repeatParams, "Muls");
193#endif218#endif
194 MulsImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),219 MulsImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),
195 scalarValue, mask, repeatTime, repeatParams);220 scalarValue, mask, repeatTime, repeatParams);
@@ -204,6 +229,9 @@ __aicore__ inline void Muls(const LocalTensor<T>& dst, const LocalTensor<T>& src
204 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "Muls")) {229 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "Muls")) {
205 ASCENDC_REPORT_CHECK_ERROR("Muls", KernelFuncType::MASK_COUNT_MODE);230 ASCENDC_REPORT_CHECK_ERROR("Muls", KernelFuncType::MASK_COUNT_MODE);
206 }231 }
232+#endif
233+#ifdef __MSTX_DFX_REPORT__
234+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask, repeatTime, repeatParams, "Muls");
207#endif235#endif
208 MulsImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue, mask,236 MulsImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue, mask,
209 repeatTime, repeatParams);237 repeatTime, repeatParams);
@@ -219,6 +247,9 @@ __aicore__ inline void Muls(const LocalTensor<T>& dst, const LocalTensor<T>& src
219 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "Muls")) {247 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "Muls")) {
220 ASCENDC_REPORT_CHECK_ERROR("Muls", KernelFuncType::MASK_COUNT_MODE);248 ASCENDC_REPORT_CHECK_ERROR("Muls", KernelFuncType::MASK_COUNT_MODE);
221 }249 }
250+#endif
251+#ifdef __MSTX_DFX_REPORT__
252+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask, repeatTime, repeatParams, "Muls");
222#endif253#endif
223 MulsImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(), scalarValue, mask,254 MulsImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(), scalarValue, mask,
224 repeatTime, repeatParams);255 repeatTime, repeatParams);
@@ -241,6 +272,9 @@ __aicore__ inline void Muls(const LocalTensor<T>& dst, const LocalTensor<T>& src
241 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, count, "Muls")) {272 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, count, "Muls")) {
242 ASCENDC_REPORT_CHECK_ERROR("Muls", KernelFuncType::CALCOUNT_MODE);273 ASCENDC_REPORT_CHECK_ERROR("Muls", KernelFuncType::CALCOUNT_MODE);
243 }274 }
275+#endif
276+#ifdef __MSTX_DFX_REPORT__
277+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, "Muls", count);
244#endif278#endif
245 MulsImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue,279 MulsImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue,
246 count);280 count);
@@ -256,6 +290,9 @@ __aicore__ inline void Muls(const LocalTensor<T>& dst, const LocalTensor<T>& src
256 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, count, "Muls")) {290 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, count, "Muls")) {
257 ASCENDC_REPORT_CHECK_ERROR("Muls", KernelFuncType::CALCOUNT_MODE);291 ASCENDC_REPORT_CHECK_ERROR("Muls", KernelFuncType::CALCOUNT_MODE);
258 }292 }
293+#endif
294+#ifdef __MSTX_DFX_REPORT__
295+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, "Muls", count);
259#endif296#endif
260 MulsImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),297 MulsImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),
261 scalarValue, count);298 scalarValue, count);
@@ -286,6 +323,9 @@ __aicore__ inline void Maxs(const LocalTensor<T>& dst, const LocalTensor<T>& src
286 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "Maxs")) {323 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "Maxs")) {
287 ASCENDC_REPORT_CHECK_ERROR("Maxs", KernelFuncType::MASK_BIT_MODE);324 ASCENDC_REPORT_CHECK_ERROR("Maxs", KernelFuncType::MASK_BIT_MODE);
288 }325 }
326+#endif
327+#ifdef __MSTX_DFX_REPORT__
328+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask[0], mask[1], repeatTime, repeatParams, "Maxs");
289#endif329#endif
290 MaxsImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue, mask,330 MaxsImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue, mask,
291 repeatTime, repeatParams);331 repeatTime, repeatParams);
@@ -301,6 +341,9 @@ __aicore__ inline void Maxs(const LocalTensor<T>& dst, const LocalTensor<T>& src
301 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "Maxs")) {341 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "Maxs")) {
302 ASCENDC_REPORT_CHECK_ERROR("Maxs", KernelFuncType::MASK_BIT_MODE);342 ASCENDC_REPORT_CHECK_ERROR("Maxs", KernelFuncType::MASK_BIT_MODE);
303 }343 }
344+#endif
345+#ifdef __MSTX_DFX_REPORT__
346+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask[0], mask[1], repeatTime, repeatParams, "Maxs");
304#endif347#endif
305 MaxsImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),348 MaxsImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),
306 scalarValue, mask, repeatTime, repeatParams);349 scalarValue, mask, repeatTime, repeatParams);
@@ -315,6 +358,9 @@ __aicore__ inline void Maxs(const LocalTensor<T>& dst, const LocalTensor<T>& src
315 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "Maxs")) {358 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "Maxs")) {
316 ASCENDC_REPORT_CHECK_ERROR("Maxs", KernelFuncType::MASK_COUNT_MODE);359 ASCENDC_REPORT_CHECK_ERROR("Maxs", KernelFuncType::MASK_COUNT_MODE);
317 }360 }
361+#endif
362+#ifdef __MSTX_DFX_REPORT__
363+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask, repeatTime, repeatParams, "Maxs");
318#endif364#endif
319 MaxsImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue, mask,365 MaxsImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue, mask,
320 repeatTime, repeatParams);366 repeatTime, repeatParams);
@@ -330,6 +376,9 @@ __aicore__ inline void Maxs(const LocalTensor<T>& dst, const LocalTensor<T>& src
330 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "Maxs")) {376 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "Maxs")) {
331 ASCENDC_REPORT_CHECK_ERROR("Maxs", KernelFuncType::MASK_COUNT_MODE);377 ASCENDC_REPORT_CHECK_ERROR("Maxs", KernelFuncType::MASK_COUNT_MODE);
332 }378 }
379+#endif
380+#ifdef __MSTX_DFX_REPORT__
381+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask, repeatTime, repeatParams, "Maxs");
333#endif382#endif
334 MaxsImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),383 MaxsImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),
335 scalarValue, mask, repeatTime, repeatParams);384 scalarValue, mask, repeatTime, repeatParams);
@@ -352,6 +401,9 @@ __aicore__ inline void Maxs(const LocalTensor<T>& dst, const LocalTensor<T>& src
352 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, count, "Maxs")) {401 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, count, "Maxs")) {
353 ASCENDC_REPORT_CHECK_ERROR("Maxs", KernelFuncType::CALCOUNT_MODE);402 ASCENDC_REPORT_CHECK_ERROR("Maxs", KernelFuncType::CALCOUNT_MODE);
354 }403 }
404+#endif
405+#ifdef __MSTX_DFX_REPORT__
406+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, "Maxs", count);
355#endif407#endif
356 MaxsImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue,408 MaxsImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue,
357 count);409 count);
@@ -367,6 +419,9 @@ __aicore__ inline void Maxs(const LocalTensor<T>& dst, const LocalTensor<T>& src
367 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, count, "Maxs")) {419 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, count, "Maxs")) {
368 ASCENDC_REPORT_CHECK_ERROR("Maxs", KernelFuncType::CALCOUNT_MODE);420 ASCENDC_REPORT_CHECK_ERROR("Maxs", KernelFuncType::CALCOUNT_MODE);
369 }421 }
422+#endif
423+#ifdef __MSTX_DFX_REPORT__
424+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, "Maxs", count);
370#endif425#endif
371 MaxsImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),426 MaxsImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),
372 scalarValue, count);427 scalarValue, count);
@@ -397,6 +452,9 @@ __aicore__ inline void Mins(const LocalTensor<T>& dst, const LocalTensor<T>& src
397 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "Mins")) {452 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "Mins")) {
398 ASCENDC_REPORT_CHECK_ERROR("Mins", KernelFuncType::MASK_BIT_MODE);453 ASCENDC_REPORT_CHECK_ERROR("Mins", KernelFuncType::MASK_BIT_MODE);
399 }454 }
455+#endif
456+#ifdef __MSTX_DFX_REPORT__
457+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask[0], mask[1], repeatTime, repeatParams, "Mins");
400#endif458#endif
401 MinsImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue, mask,459 MinsImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue, mask,
402 repeatTime, repeatParams);460 repeatTime, repeatParams);
@@ -412,6 +470,9 @@ __aicore__ inline void Mins(const LocalTensor<T>& dst, const LocalTensor<T>& src
412 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "Mins")) {470 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "Mins")) {
413 ASCENDC_REPORT_CHECK_ERROR("Mins", KernelFuncType::MASK_BIT_MODE);471 ASCENDC_REPORT_CHECK_ERROR("Mins", KernelFuncType::MASK_BIT_MODE);
414 }472 }
473+#endif
474+#ifdef __MSTX_DFX_REPORT__
475+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask[0], mask[1], repeatTime, repeatParams, "Mins");
415#endif476#endif
416 MinsImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),477 MinsImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),
417 scalarValue, mask, repeatTime, repeatParams);478 scalarValue, mask, repeatTime, repeatParams);
@@ -426,6 +487,9 @@ __aicore__ inline void Mins(const LocalTensor<T>& dst, const LocalTensor<T>& src
426 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "Mins")) {487 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "Mins")) {
427 ASCENDC_REPORT_CHECK_ERROR("Mins", KernelFuncType::MASK_COUNT_MODE);488 ASCENDC_REPORT_CHECK_ERROR("Mins", KernelFuncType::MASK_COUNT_MODE);
428 }489 }
490+#endif
491+#ifdef __MSTX_DFX_REPORT__
492+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask, repeatTime, repeatParams, "Mins");
429#endif493#endif
430 MinsImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue, mask,494 MinsImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue, mask,
431 repeatTime, repeatParams);495 repeatTime, repeatParams);
@@ -441,6 +505,9 @@ __aicore__ inline void Mins(const LocalTensor<T>& dst, const LocalTensor<T>& src
441 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "Mins")) {505 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "Mins")) {
442 ASCENDC_REPORT_CHECK_ERROR("Mins", KernelFuncType::MASK_COUNT_MODE);506 ASCENDC_REPORT_CHECK_ERROR("Mins", KernelFuncType::MASK_COUNT_MODE);
443 }507 }
508+#endif
509+#ifdef __MSTX_DFX_REPORT__
510+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask, repeatTime, repeatParams, "Mins");
444#endif511#endif
445 MinsImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),512 MinsImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),
446 scalarValue, mask, repeatTime, repeatParams);513 scalarValue, mask, repeatTime, repeatParams);
@@ -463,6 +530,9 @@ __aicore__ inline void Mins(const LocalTensor<T>& dst, const LocalTensor<T>& src
463 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, count, "Mins")) {530 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, count, "Mins")) {
464 ASCENDC_REPORT_CHECK_ERROR("Mins", KernelFuncType::CALCOUNT_MODE);531 ASCENDC_REPORT_CHECK_ERROR("Mins", KernelFuncType::CALCOUNT_MODE);
465 }532 }
533+#endif
534+#ifdef __MSTX_DFX_REPORT__
535+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, "Mins", count);
466#endif536#endif
467 MinsImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue,537 MinsImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue,
468 count);538 count);
@@ -478,6 +548,9 @@ __aicore__ inline void Mins(const LocalTensor<T>& dst, const LocalTensor<T>& src
478 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, count, "Mins")) {548 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, count, "Mins")) {
479 ASCENDC_REPORT_CHECK_ERROR("Mins", KernelFuncType::CALCOUNT_MODE);549 ASCENDC_REPORT_CHECK_ERROR("Mins", KernelFuncType::CALCOUNT_MODE);
480 }550 }
551+#endif
552+#ifdef __MSTX_DFX_REPORT__
553+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, "Mins", count);
481#endif554#endif
482 MinsImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),555 MinsImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),
483 scalarValue, count);556 scalarValue, count);
@@ -508,6 +581,9 @@ __aicore__ inline void ShiftLeft(const LocalTensor<T>& dst, const LocalTensor<T>
508 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "ShiftLeft")) {581 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "ShiftLeft")) {
509 ASCENDC_REPORT_CHECK_ERROR("ShiftLeft", KernelFuncType::MASK_BIT_MODE);582 ASCENDC_REPORT_CHECK_ERROR("ShiftLeft", KernelFuncType::MASK_BIT_MODE);
510 }583 }
584+#endif
585+#ifdef __MSTX_DFX_REPORT__
586+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask[0], mask[1], repeatTime, repeatParams, "ShiftLeft");
511#endif587#endif
512 ShiftLeftImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue,588 ShiftLeftImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue,
513 mask, repeatTime, repeatParams);589 mask, repeatTime, repeatParams);
@@ -523,6 +599,9 @@ __aicore__ inline void ShiftLeft(const LocalTensor<T>& dst, const LocalTensor<T>
523 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "ShiftLeft")) {599 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "ShiftLeft")) {
524 ASCENDC_REPORT_CHECK_ERROR("ShiftLeft", KernelFuncType::MASK_BIT_MODE);600 ASCENDC_REPORT_CHECK_ERROR("ShiftLeft", KernelFuncType::MASK_BIT_MODE);
525 }601 }
602+#endif
603+#ifdef __MSTX_DFX_REPORT__
604+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask[0], mask[1], repeatTime, repeatParams, "ShiftLeft");
526#endif605#endif
527 ShiftLeftImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(),606 ShiftLeftImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(),
528 (__ubuf__ PrimType*)src.GetPhyAddr(), scalarValue, mask, repeatTime, repeatParams);607 (__ubuf__ PrimType*)src.GetPhyAddr(), scalarValue, mask, repeatTime, repeatParams);
@@ -537,6 +616,9 @@ __aicore__ inline void ShiftLeft(const LocalTensor<T>& dst, const LocalTensor<T>
537 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "ShiftLeft")) {616 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "ShiftLeft")) {
538 ASCENDC_REPORT_CHECK_ERROR("ShiftLeft", KernelFuncType::MASK_COUNT_MODE);617 ASCENDC_REPORT_CHECK_ERROR("ShiftLeft", KernelFuncType::MASK_COUNT_MODE);
539 }618 }
619+#endif
620+#ifdef __MSTX_DFX_REPORT__
621+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask, repeatTime, repeatParams, "ShiftLeft");
540#endif622#endif
541 ShiftLeftImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue,623 ShiftLeftImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue,
542 mask, repeatTime, repeatParams);624 mask, repeatTime, repeatParams);
@@ -552,6 +634,9 @@ __aicore__ inline void ShiftLeft(const LocalTensor<T>& dst, const LocalTensor<T>
552 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "ShiftLeft")) {634 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "ShiftLeft")) {
553 ASCENDC_REPORT_CHECK_ERROR("ShiftLeft", KernelFuncType::MASK_COUNT_MODE);635 ASCENDC_REPORT_CHECK_ERROR("ShiftLeft", KernelFuncType::MASK_COUNT_MODE);
554 }636 }
637+#endif
638+#ifdef __MSTX_DFX_REPORT__
639+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask, repeatTime, repeatParams, "ShiftLeft");
555#endif640#endif
556 ShiftLeftImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(),641 ShiftLeftImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(),
557 (__ubuf__ PrimType*)src.GetPhyAddr(), scalarValue, mask, repeatTime, repeatParams);642 (__ubuf__ PrimType*)src.GetPhyAddr(), scalarValue, mask, repeatTime, repeatParams);
@@ -574,6 +659,9 @@ __aicore__ inline void ShiftLeft(const LocalTensor<T>& dst, const LocalTensor<T>
574 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, count, "ShiftLeft")) {659 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, count, "ShiftLeft")) {
575 ASCENDC_REPORT_CHECK_ERROR("ShiftLeft", KernelFuncType::CALCOUNT_MODE);660 ASCENDC_REPORT_CHECK_ERROR("ShiftLeft", KernelFuncType::CALCOUNT_MODE);
576 }661 }
662+#endif
663+#ifdef __MSTX_DFX_REPORT__
664+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, "ShiftLeft", count);
577#endif665#endif
578 ShiftLeftImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue,666 ShiftLeftImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue,
579 count);667 count);
@@ -589,6 +677,9 @@ __aicore__ inline void ShiftLeft(const LocalTensor<T>& dst, const LocalTensor<T>
589 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, count, "ShiftLeft")) {677 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, count, "ShiftLeft")) {
590 ASCENDC_REPORT_CHECK_ERROR("ShiftLeft", KernelFuncType::CALCOUNT_MODE);678 ASCENDC_REPORT_CHECK_ERROR("ShiftLeft", KernelFuncType::CALCOUNT_MODE);
591 }679 }
680+#endif
681+#ifdef __MSTX_DFX_REPORT__
682+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, "ShiftLeft", count);
592#endif683#endif
593 ShiftLeftImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(),684 ShiftLeftImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(),
594 (__ubuf__ PrimType*)src.GetPhyAddr(), scalarValue, count);685 (__ubuf__ PrimType*)src.GetPhyAddr(), scalarValue, count);
@@ -619,6 +710,9 @@ __aicore__ inline void ShiftRight(const LocalTensor<T>& dst, const LocalTensor<T
619 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "ShiftRight")) {710 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "ShiftRight")) {
620 ASCENDC_REPORT_CHECK_ERROR("ShiftRight", KernelFuncType::MASK_BIT_MODE);711 ASCENDC_REPORT_CHECK_ERROR("ShiftRight", KernelFuncType::MASK_BIT_MODE);
621 }712 }
713+#endif
714+#ifdef __MSTX_DFX_REPORT__
715+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask[0], mask[1], repeatTime, repeatParams, "ShiftRight");
622#endif716#endif
623 ShiftRightImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue,717 ShiftRightImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue,
624 mask, repeatTime, repeatParams, roundEn);718 mask, repeatTime, repeatParams, roundEn);
@@ -634,6 +728,9 @@ __aicore__ inline void ShiftRight(const LocalTensor<T>& dst, const LocalTensor<T
634 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "ShiftRight")) {728 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "ShiftRight")) {
635 ASCENDC_REPORT_CHECK_ERROR("ShiftRight", KernelFuncType::MASK_BIT_MODE);729 ASCENDC_REPORT_CHECK_ERROR("ShiftRight", KernelFuncType::MASK_BIT_MODE);
636 }730 }
731+#endif
732+#ifdef __MSTX_DFX_REPORT__
733+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask[0], mask[1], repeatTime, repeatParams, "ShiftRight");
637#endif734#endif
638 ShiftRightImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(),735 ShiftRightImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(),
639 (__ubuf__ PrimType*)src.GetPhyAddr(), scalarValue, mask, repeatTime, repeatParams, roundEn);736 (__ubuf__ PrimType*)src.GetPhyAddr(), scalarValue, mask, repeatTime, repeatParams, roundEn);
@@ -648,6 +745,9 @@ __aicore__ inline void ShiftRight(const LocalTensor<T>& dst, const LocalTensor<T
648 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "ShiftRight")) {745 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "ShiftRight")) {
649 ASCENDC_REPORT_CHECK_ERROR("ShiftRight", KernelFuncType::MASK_COUNT_MODE);746 ASCENDC_REPORT_CHECK_ERROR("ShiftRight", KernelFuncType::MASK_COUNT_MODE);
650 }747 }
748+#endif
749+#ifdef __MSTX_DFX_REPORT__
750+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask, repeatTime, repeatParams, "ShiftRight");
651#endif751#endif
652 ShiftRightImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue,752 ShiftRightImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue,
653 mask, repeatTime, repeatParams, roundEn);753 mask, repeatTime, repeatParams, roundEn);
@@ -663,6 +763,9 @@ __aicore__ inline void ShiftRight(const LocalTensor<T>& dst, const LocalTensor<T
663 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "ShiftRight")) {763 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "ShiftRight")) {
664 ASCENDC_REPORT_CHECK_ERROR("ShiftRight", KernelFuncType::MASK_COUNT_MODE);764 ASCENDC_REPORT_CHECK_ERROR("ShiftRight", KernelFuncType::MASK_COUNT_MODE);
665 }765 }
766+#endif
767+#ifdef __MSTX_DFX_REPORT__
768+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask, repeatTime, repeatParams, "ShiftRight");
666#endif769#endif
667 ShiftRightImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(),770 ShiftRightImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(),
668 (__ubuf__ PrimType*)src.GetPhyAddr(), scalarValue, mask, repeatTime, repeatParams, roundEn);771 (__ubuf__ PrimType*)src.GetPhyAddr(), scalarValue, mask, repeatTime, repeatParams, roundEn);
@@ -685,6 +788,9 @@ __aicore__ inline void ShiftRight(const LocalTensor<T>& dst, const LocalTensor<T
685 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, count, "ShiftRight")) {788 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, count, "ShiftRight")) {
686 ASCENDC_REPORT_CHECK_ERROR("ShiftRight", KernelFuncType::CALCOUNT_MODE);789 ASCENDC_REPORT_CHECK_ERROR("ShiftRight", KernelFuncType::CALCOUNT_MODE);
687 }790 }
791+#endif
792+#ifdef __MSTX_DFX_REPORT__
793+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, "ShiftRight", count);
688#endif794#endif
689 ShiftRightImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue,795 ShiftRightImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue,
690 count);796 count);
@@ -700,6 +806,9 @@ __aicore__ inline void ShiftRight(const LocalTensor<T>& dst, const LocalTensor<T
700 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, count, "ShiftRight")) {806 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, count, "ShiftRight")) {
701 ASCENDC_REPORT_CHECK_ERROR("ShiftRight", KernelFuncType::CALCOUNT_MODE);807 ASCENDC_REPORT_CHECK_ERROR("ShiftRight", KernelFuncType::CALCOUNT_MODE);
702 }808 }
809+#endif
810+#ifdef __MSTX_DFX_REPORT__
811+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, "ShiftRight", count);
703#endif812#endif
704 ShiftRightImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(),813 ShiftRightImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(),
705 (__ubuf__ PrimType*)src.GetPhyAddr(), scalarValue, count);814 (__ubuf__ PrimType*)src.GetPhyAddr(), scalarValue, count);
@@ -730,6 +839,9 @@ __aicore__ inline void LeakyRelu(const LocalTensor<T>& dst, const LocalTensor<T>
730 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "LeakyRelu")) {839 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "LeakyRelu")) {
731 ASCENDC_REPORT_CHECK_ERROR("LeakyRelu", KernelFuncType::MASK_BIT_MODE);840 ASCENDC_REPORT_CHECK_ERROR("LeakyRelu", KernelFuncType::MASK_BIT_MODE);
732 }841 }
842+#endif
843+#ifdef __MSTX_DFX_REPORT__
844+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask[0], mask[1], repeatTime, repeatParams, "LeakyRelu");
733#endif845#endif
734 LeakyReluImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue,846 LeakyReluImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue,
735 mask, repeatTime, repeatParams);847 mask, repeatTime, repeatParams);
@@ -745,6 +857,9 @@ __aicore__ inline void LeakyRelu(const LocalTensor<T>& dst, const LocalTensor<T>
745 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "LeakyRelu")) {857 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "LeakyRelu")) {
746 ASCENDC_REPORT_CHECK_ERROR("LeakyRelu", KernelFuncType::MASK_BIT_MODE);858 ASCENDC_REPORT_CHECK_ERROR("LeakyRelu", KernelFuncType::MASK_BIT_MODE);
747 }859 }
860+#endif
861+#ifdef __MSTX_DFX_REPORT__
862+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask[0], mask[1], repeatTime, repeatParams, "LeakyRelu");
748#endif863#endif
749 LeakyReluImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(),864 LeakyReluImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(),
750 (__ubuf__ PrimType*)src.GetPhyAddr(), scalarValue, mask, repeatTime, repeatParams);865 (__ubuf__ PrimType*)src.GetPhyAddr(), scalarValue, mask, repeatTime, repeatParams);
@@ -759,6 +874,9 @@ __aicore__ inline void LeakyRelu(const LocalTensor<T>& dst, const LocalTensor<T>
759 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "LeakyRelu")) {874 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "LeakyRelu")) {
760 ASCENDC_REPORT_CHECK_ERROR("LeakyRelu", KernelFuncType::MASK_COUNT_MODE);875 ASCENDC_REPORT_CHECK_ERROR("LeakyRelu", KernelFuncType::MASK_COUNT_MODE);
761 }876 }
877+#endif
878+#ifdef __MSTX_DFX_REPORT__
879+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask, repeatTime, repeatParams, "LeakyRelu");
762#endif880#endif
763 LeakyReluImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue,881 LeakyReluImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue,
764 mask, repeatTime, repeatParams);882 mask, repeatTime, repeatParams);
@@ -774,6 +892,9 @@ __aicore__ inline void LeakyRelu(const LocalTensor<T>& dst, const LocalTensor<T>
774 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "LeakyRelu")) {892 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, mask, repeatTime, repeatParams, "LeakyRelu")) {
775 ASCENDC_REPORT_CHECK_ERROR("LeakyRelu", KernelFuncType::MASK_COUNT_MODE);893 ASCENDC_REPORT_CHECK_ERROR("LeakyRelu", KernelFuncType::MASK_COUNT_MODE);
776 }894 }
895+#endif
896+#ifdef __MSTX_DFX_REPORT__
897+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask, repeatTime, repeatParams, "LeakyRelu");
777#endif898#endif
778 LeakyReluImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(),899 LeakyReluImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(),
779 (__ubuf__ PrimType*)src.GetPhyAddr(), scalarValue, mask, repeatTime, repeatParams);900 (__ubuf__ PrimType*)src.GetPhyAddr(), scalarValue, mask, repeatTime, repeatParams);
@@ -796,6 +917,9 @@ __aicore__ inline void LeakyRelu(const LocalTensor<T>& dst, const LocalTensor<T>
796 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, count, "LeakyRelu")) {917 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, count, "LeakyRelu")) {
797 ASCENDC_REPORT_CHECK_ERROR("LeakyRelu", KernelFuncType::CALCOUNT_MODE);918 ASCENDC_REPORT_CHECK_ERROR("LeakyRelu", KernelFuncType::CALCOUNT_MODE);
798 }919 }
920+#endif
921+#ifdef __MSTX_DFX_REPORT__
922+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, "LeakyRelu", count);
799#endif923#endif
800 LeakyReluImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue,924 LeakyReluImpl<T, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ T*)src.GetPhyAddr(), scalarValue,
801 count);925 count);
@@ -811,6 +935,9 @@ __aicore__ inline void LeakyRelu(const LocalTensor<T>& dst, const LocalTensor<T>
811 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, count, "LeakyRelu")) {935 if (!CheckFunVecBinaryScalar(dst, src, scalarValue, count, "LeakyRelu")) {
812 ASCENDC_REPORT_CHECK_ERROR("LeakyRelu", KernelFuncType::CALCOUNT_MODE);936 ASCENDC_REPORT_CHECK_ERROR("LeakyRelu", KernelFuncType::CALCOUNT_MODE);
813 }937 }
938+#endif
939+#ifdef __MSTX_DFX_REPORT__
940+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, "LeakyRelu", count);
814#endif941#endif
815 LeakyReluImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(),942 LeakyReluImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(),
816 (__ubuf__ PrimType*)src.GetPhyAddr(), scalarValue, count);943 (__ubuf__ PrimType*)src.GetPhyAddr(), scalarValue, count);
Mimpl/basic_api/kernel_operator_vec_brcb_intf_impl.h+4-0
@@ -17,6 +17,7 @@
17#include "kernel_tensor.h"17#include "kernel_tensor.h"
18#include "kernel_check.h"18#include "kernel_check.h"
19#include "kernel_struct_brcb.h"19#include "kernel_struct_brcb.h"
20+#include "mstx_local_tensor_info.h"
20 21 
21#if __NPU_ARCH__ == 100122#if __NPU_ARCH__ == 1001
22#include "dav_c100/kernel_operator_vec_brcb_impl.h"23#include "dav_c100/kernel_operator_vec_brcb_impl.h"
@@ -52,6 +53,9 @@ template <typename T>
52__aicore__ inline void Brcb(const LocalTensor<T>& dst, const LocalTensor<T>& src0, const uint8_t repeatTime,53__aicore__ inline void Brcb(const LocalTensor<T>& dst, const LocalTensor<T>& src0, const uint8_t repeatTime,
53 const BrcbRepeatParams& repeatParams)54 const BrcbRepeatParams& repeatParams)
54{55{
56+#ifdef __MSTX_DFX_REPORT__
C
Cchenyiyuan2月6日

这个文件没包对应头文件

likedislike
57+ MstxTensor::GetMstxVecBrcbInfo(dst, src0, repeatTime, repeatParams, "Brcb");
58+#endif
55 using PrimType = PrimT<T>;59 using PrimType = PrimT<T>;
56#if ASCENDC_CPU_DEBUG60#if ASCENDC_CPU_DEBUG
57 if (!CheckFunBcB(dst, src0, repeatTime, repeatParams, "Brcb")) {61 if (!CheckFunBcB(dst, src0, repeatTime, repeatParams, "Brcb")) {
Mimpl/basic_api/kernel_operator_vec_cmpsel_intf_impl.h+55-1
@@ -18,6 +18,7 @@
18#include "kernel_check.h"18#include "kernel_check.h"
19#include "kernel_struct_binary.h"19#include "kernel_struct_binary.h"
20#include "kernel_struct_unary.h"20#include "kernel_struct_unary.h"
21+#include "mstx_local_tensor_info.h"
21 22 
22#if __NPU_ARCH__ == 100123#if __NPU_ARCH__ == 1001
23#include "dav_c100/kernel_operator_vec_cmpsel_impl.h"24#include "dav_c100/kernel_operator_vec_cmpsel_impl.h"
@@ -66,6 +67,9 @@ __aicore__ inline void Compare(const LocalTensor<U>& dst, const LocalTensor<T>&
66 const LocalTensor<T>& src1, CMPMODE cmpMode, const uint64_t mask[], uint8_t repeatTime,67 const LocalTensor<T>& src1, CMPMODE cmpMode, const uint64_t mask[], uint8_t repeatTime,
67 const BinaryRepeatParams& repeatParams)68 const BinaryRepeatParams& repeatParams)
68{69{
70+#ifdef __MSTX_DFX_REPORT__
71+ MstxTensor::GetMstxVecBinaryCmpInfo(dst, src0, src1, mask[0], mask[1], repeatTime, repeatParams, isSetMask, "Compare");
72+#endif
69 using SrcPrimType = PrimT<T>;73 using SrcPrimType = PrimT<T>;
70 using DstPrimType = PrimT<U>;74 using DstPrimType = PrimT<U>;
71#if __NPU_ARCH__ == 200275#if __NPU_ARCH__ == 2002
@@ -92,6 +96,9 @@ __aicore__ inline void Compare(const LocalTensor<U>& dst, const LocalTensor<T>&
92 const LocalTensor<T>& src1, CMPMODE cmpMode, const uint64_t mask, uint8_t repeatTime,96 const LocalTensor<T>& src1, CMPMODE cmpMode, const uint64_t mask, uint8_t repeatTime,
93 const BinaryRepeatParams& repeatParams)97 const BinaryRepeatParams& repeatParams)
94{98{
99+#ifdef __MSTX_DFX_REPORT__
100+ MstxTensor::GetMstxVecBinaryCmpInfo(dst, src0, src1, mask, repeatTime, repeatParams, isSetMask, "Compare");
101+#endif
95 using SrcPrimType = PrimT<T>;102 using SrcPrimType = PrimT<T>;
96 using DstPrimType = PrimT<U>;103 using DstPrimType = PrimT<U>;
97#if __NPU_ARCH__ == 2002104#if __NPU_ARCH__ == 2002
@@ -164,6 +171,9 @@ template <typename T, typename U>
164__aicore__ inline void Compare(const LocalTensor<U>& dst, const LocalTensor<T>& src0,171__aicore__ inline void Compare(const LocalTensor<U>& dst, const LocalTensor<T>& src0,
165 const LocalTensor<T>& src1, CMPMODE cmpMode, uint32_t count)172 const LocalTensor<T>& src1, CMPMODE cmpMode, uint32_t count)
166{173{
174+#ifdef __MSTX_DFX_REPORT__
175+ MstxTensor::GetMstxVecBinaryCmpInfo(dst, src0, src1, "Compare", count);
176+#endif
167 using SrcPrimType = PrimT<T>;177 using SrcPrimType = PrimT<T>;
168 using DstPrimType = PrimT<U>;178 using DstPrimType = PrimT<U>;
169#if __NPU_ARCH__ == 2002179#if __NPU_ARCH__ == 2002
@@ -275,6 +285,9 @@ __aicore__ inline void Compares(const LocalTensor<U>& dst, const LocalTensor<T>&
275 const T src1Scalar, CMPMODE cmpMode, const uint64_t mask[], uint8_t repeatTime,285 const T src1Scalar, CMPMODE cmpMode, const uint64_t mask[], uint8_t repeatTime,
276 const UnaryRepeatParams& repeatParams)286 const UnaryRepeatParams& repeatParams)
277{287{
288+#ifdef __MSTX_DFX_REPORT__
289+ MstxTensor::GetMstxVecUnaryCmpsInfo<U, T, isSetMask>(dst, src0, mask[0], mask[1], repeatTime, repeatParams, "Compares");
290+#endif
278#if __NPU_ARCH__ == 2002291#if __NPU_ARCH__ == 2002
279 ASCENDC_ASSERT((SupportType<T, half, float>() && SupportType<U, uint8_t>()),292 ASCENDC_ASSERT((SupportType<T, half, float>() && SupportType<U, uint8_t>()),
280 {KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in Compares, current api support dtype combination is "293 {KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in Compares, current api support dtype combination is "
@@ -299,6 +312,9 @@ __aicore__ inline void Compares(const LocalTensor<U>& dst, const LocalTensor<T>&
299 const T src1Scalar, CMPMODE cmpMode, const uint64_t mask, uint8_t repeatTime,312 const T src1Scalar, CMPMODE cmpMode, const uint64_t mask, uint8_t repeatTime,
300 const UnaryRepeatParams& repeatParams)313 const UnaryRepeatParams& repeatParams)
301{314{
315+#ifdef __MSTX_DFX_REPORT__
316+ MstxTensor::GetMstxVecUnaryCmpsInfo<U, T, isSetMask>(dst, src0, mask, repeatTime, repeatParams, "Compares");
317+#endif
302#if __NPU_ARCH__ == 2002318#if __NPU_ARCH__ == 2002
303 ASCENDC_ASSERT((SupportType<T, half, float>() && SupportType<U, uint8_t>()),319 ASCENDC_ASSERT((SupportType<T, half, float>() && SupportType<U, uint8_t>()),
304 {KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in Compares, current api support dtype combination is "320 {KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in Compares, current api support dtype combination is "
@@ -324,6 +340,9 @@ __aicore__ inline void CompareScalar(const LocalTensor<U>& dst, const LocalTenso
324 const T src1Scalar, CMPMODE cmpMode, const uint64_t mask[], uint8_t repeatTime,340 const T src1Scalar, CMPMODE cmpMode, const uint64_t mask[], uint8_t repeatTime,
325 const UnaryRepeatParams& repeatParams)341 const UnaryRepeatParams& repeatParams)
326{342{
343+#ifdef __MSTX_DFX_REPORT__
344+ MstxTensor::GetMstxVecUnaryCmpsInfo<U, T, isSetMask>(dst, src0, mask[0], mask[1], repeatTime, repeatParams, "CompareScalar");
345+#endif
327#if __NPU_ARCH__ == 2002346#if __NPU_ARCH__ == 2002
328 ASCENDC_ASSERT((SupportType<T, half, float>() && SupportType<U, uint8_t>()),347 ASCENDC_ASSERT((SupportType<T, half, float>() && SupportType<U, uint8_t>()),
329 {KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in CompareScalar, current api support dtype combination is "348 {KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in CompareScalar, current api support dtype combination is "
@@ -349,6 +368,9 @@ __aicore__ inline void CompareScalar(const LocalTensor<U>& dst, const LocalTenso
349 const T src1Scalar, CMPMODE cmpMode, const uint64_t mask, uint8_t repeatTime,368 const T src1Scalar, CMPMODE cmpMode, const uint64_t mask, uint8_t repeatTime,
350 const UnaryRepeatParams& repeatParams)369 const UnaryRepeatParams& repeatParams)
351{370{
371+#ifdef __MSTX_DFX_REPORT__
372+ MstxTensor::GetMstxVecUnaryCmpsInfo<U, T, isSetMask>(dst, src0, mask, repeatTime, repeatParams, "CompareScalar");
373+#endif
352#if __NPU_ARCH__ == 2002374#if __NPU_ARCH__ == 2002
353 ASCENDC_ASSERT((SupportType<T, half, float>() && SupportType<U, uint8_t>()),375 ASCENDC_ASSERT((SupportType<T, half, float>() && SupportType<U, uint8_t>()),
354 {KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in CompareScalar, current api support dtype combination is "376 {KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in CompareScalar, current api support dtype combination is "
@@ -381,6 +403,9 @@ template <typename T, typename U>
381__aicore__ inline void Compares(const LocalTensor<U>& dst, const LocalTensor<T>& src0,403__aicore__ inline void Compares(const LocalTensor<U>& dst, const LocalTensor<T>& src0,
382 const T src1Scalar, CMPMODE cmpMode, uint32_t count)404 const T src1Scalar, CMPMODE cmpMode, uint32_t count)
383{405{
406+#ifdef __MSTX_DFX_REPORT__
407+ MstxTensor::GetMstxVecUnaryCmpsInfo<U, T, true>(dst, src0, "Compares", count);
408+#endif
384#if __NPU_ARCH__ == 2002409#if __NPU_ARCH__ == 2002
385 ASCENDC_ASSERT((SupportType<T, half, float>() && SupportType<U, uint8_t>()),410 ASCENDC_ASSERT((SupportType<T, half, float>() && SupportType<U, uint8_t>()),
386 {KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in Compares, current api support dtype combination is "411 {KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in Compares, current api support dtype combination is "
@@ -406,6 +431,9 @@ template <typename T, typename U>
406__aicore__ inline void CompareScalar(const LocalTensor<U>& dst, const LocalTensor<T>& src0,431__aicore__ inline void CompareScalar(const LocalTensor<U>& dst, const LocalTensor<T>& src0,
407 const T src1Scalar, CMPMODE cmpMode, uint32_t count)432 const T src1Scalar, CMPMODE cmpMode, uint32_t count)
408{433{
434+#ifdef __MSTX_DFX_REPORT__
435+ MstxTensor::GetMstxVecUnaryCmpsInfo<U, T, true>(dst, src0, "CompareScalar", count);
436+#endif
409#if __NPU_ARCH__ == 2002437#if __NPU_ARCH__ == 2002
410 ASCENDC_ASSERT((SupportType<T, half, float>() && SupportType<U, uint8_t>()),438 ASCENDC_ASSERT((SupportType<T, half, float>() && SupportType<U, uint8_t>()),
411 {KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in CompareScalar, current api support dtype combination is "439 {KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in CompareScalar, current api support dtype combination is "
@@ -456,6 +484,9 @@ __aicore__ inline void Select(const LocalTensor<T>& dst, const LocalTensor<U>& s
456 const LocalTensor<T>& src0, const LocalTensor<T>& src1, SELMODE selMode, uint64_t mask[],484 const LocalTensor<T>& src0, const LocalTensor<T>& src1, SELMODE selMode, uint64_t mask[],
457 uint8_t repeatTime, const BinaryRepeatParams& repeatParams)485 uint8_t repeatTime, const BinaryRepeatParams& repeatParams)
458{486{
487+#ifdef __MSTX_DFX_REPORT__
488+ MstxTensor::GetMstxVecBinarySelInfo(dst, selMask, src0, src1, mask[0], mask[1], repeatTime, repeatParams, isSetMask, selMode, "Select");
489+#endif
459 using DataPrimType = PrimT<T>;490 using DataPrimType = PrimT<T>;
460 using MaskPrimType = PrimT<U>;491 using MaskPrimType = PrimT<U>;
461#if ASCENDC_CPU_DEBUG492#if ASCENDC_CPU_DEBUG
@@ -475,6 +506,9 @@ __aicore__ inline void Select(const LocalTensor<T>& dst, const LocalTensor<U>& s
475 const LocalTensor<T>& src0, const LocalTensor<T>& src1, SELMODE selMode, uint64_t mask,506 const LocalTensor<T>& src0, const LocalTensor<T>& src1, SELMODE selMode, uint64_t mask,
476 uint8_t repeatTime, const BinaryRepeatParams& repeatParams)507 uint8_t repeatTime, const BinaryRepeatParams& repeatParams)
477{508{
509+#ifdef __MSTX_DFX_REPORT__
510+ MstxTensor::GetMstxVecBinarySelInfo(dst, selMask, src0, src1, mask, repeatTime, repeatParams, isSetMask, selMode, "Select");
511+#endif
478 using DataPrimType = PrimT<T>;512 using DataPrimType = PrimT<T>;
479 using MaskPrimType = PrimT<U>;513 using MaskPrimType = PrimT<U>;
480#if ASCENDC_CPU_DEBUG514#if ASCENDC_CPU_DEBUG
@@ -492,6 +526,9 @@ template <typename T, SELMODE selMode>
492__aicore__ inline void Select(const LocalTensor<T>& dst, const LocalTensor<T>& src0,526__aicore__ inline void Select(const LocalTensor<T>& dst, const LocalTensor<T>& src0,
493 const LocalTensor<T>& src1, uint8_t repeatTime, const BinaryRepeatParams& repeatParams)527 const LocalTensor<T>& src1, uint8_t repeatTime, const BinaryRepeatParams& repeatParams)
494{528{
529+#ifdef __MSTX_DFX_REPORT__
530+ MstxTensor::GetMstxVecBinarySelInfo(dst, src0, src1, repeatTime, repeatParams, selMode, "Select");
531+#endif
495 using PrimType = PrimT<T>;532 using PrimType = PrimT<T>;
496 SelectCal<PrimType, selMode>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),533 SelectCal<PrimType, selMode>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src0.GetPhyAddr(),
497 (__ubuf__ PrimType*)src1.GetPhyAddr(), repeatTime, repeatParams);534 (__ubuf__ PrimType*)src1.GetPhyAddr(), repeatTime, repeatParams);
@@ -501,6 +538,9 @@ template <typename T, typename U>
501__aicore__ inline void Select(const LocalTensor<T>& dst, const LocalTensor<U>& selMask,538__aicore__ inline void Select(const LocalTensor<T>& dst, const LocalTensor<U>& selMask,
502 const LocalTensor<T>& src0, uint8_t repeatTime, const BinaryRepeatParams& repeatParams)539 const LocalTensor<T>& src0, uint8_t repeatTime, const BinaryRepeatParams& repeatParams)
503{540{
541+#ifdef __MSTX_DFX_REPORT__
542+ MstxTensor::GetMstxVecBinarySelInfo(dst, selMask, src0, repeatTime, repeatParams, SELMODE::VSEL_TENSOR_SCALAR_MODE, "Select");
543+#endif
504 using DataPrimType = PrimT<T>;544 using DataPrimType = PrimT<T>;
505 using MaskPrimType = PrimT<U>;545 using MaskPrimType = PrimT<U>;
506 SelectCal<DataPrimType, MaskPrimType>((__ubuf__ DataPrimType*)dst.GetPhyAddr(),546 SelectCal<DataPrimType, MaskPrimType>((__ubuf__ DataPrimType*)dst.GetPhyAddr(),
@@ -523,6 +563,9 @@ template <typename T, typename U>
523__aicore__ inline void Select(const LocalTensor<T>& dst, const LocalTensor<U>& selMask,563__aicore__ inline void Select(const LocalTensor<T>& dst, const LocalTensor<U>& selMask,
524 const LocalTensor<T>& src0, const LocalTensor<T>& src1, SELMODE selMode, uint32_t count)564 const LocalTensor<T>& src0, const LocalTensor<T>& src1, SELMODE selMode, uint32_t count)
525{565{
566+#ifdef __MSTX_DFX_REPORT__
567+ MstxTensor::GetMstxVecBinarySelInfo(dst, selMask, src0, src1, count, true, selMode, "Select");
568+#endif
526 using DataPrimType = PrimT<T>;569 using DataPrimType = PrimT<T>;
527 using MaskPrimType = PrimT<U>;570 using MaskPrimType = PrimT<U>;
528#if ASCENDC_CPU_DEBUG571#if ASCENDC_CPU_DEBUG
@@ -558,6 +601,10 @@ __aicore__ inline void Select(const LocalTensor<T>& dst, const LocalTensor<U>& s
558 const LocalTensor<T>& src0, T src1, SELMODE selMode, uint64_t mask[], uint8_t repeatTime,601 const LocalTensor<T>& src0, T src1, SELMODE selMode, uint64_t mask[], uint8_t repeatTime,
559 const BinaryRepeatParams& repeatParams)602 const BinaryRepeatParams& repeatParams)
560{603{
604+#ifdef __MSTX_DFX_REPORT__
605+ MstxTensor::GetMstxVecBinarySelInfo(dst, selMask, src0, mask[0], mask[1], repeatTime,
606+ repeatParams, isSetMask, selMode, "Select");
607+#endif
561#if ASCENDC_CPU_DEBUG608#if ASCENDC_CPU_DEBUG
562 MaskSetter::Instance().SetMask(isSetMask);609 MaskSetter::Instance().SetMask(isSetMask);
563 if (!CheckFunVecBinaryScalar(dst, src0, src1, mask, repeatTime, repeatParams, "Select")) {610 if (!CheckFunVecBinaryScalar(dst, src0, src1, mask, repeatTime, repeatParams, "Select")) {
@@ -575,6 +622,10 @@ __aicore__ inline void Select(const LocalTensor<T>& dst, const LocalTensor<U>& s
575 const LocalTensor<T>& src0, T src1, SELMODE selMode, uint64_t mask, uint8_t repeatTime,622 const LocalTensor<T>& src0, T src1, SELMODE selMode, uint64_t mask, uint8_t repeatTime,
576 const BinaryRepeatParams& repeatParams)623 const BinaryRepeatParams& repeatParams)
577{624{
625+#ifdef __MSTX_DFX_REPORT__
626+ MstxTensor::GetMstxVecBinarySelInfo(dst, selMask, src0, mask, repeatTime,
627+ repeatParams, isSetMask, selMode, "Select");
628+#endif
578#if ASCENDC_CPU_DEBUG629#if ASCENDC_CPU_DEBUG
579 MaskSetter::Instance().SetMask(isSetMask);630 MaskSetter::Instance().SetMask(isSetMask);
580 if (!CheckFunVecBinaryScalar(dst, src0, src1, mask, repeatTime, repeatParams, "Select")) {631 if (!CheckFunVecBinaryScalar(dst, src0, src1, mask, repeatTime, repeatParams, "Select")) {
@@ -601,6 +652,9 @@ template <typename T, typename U>
601__aicore__ inline void Select(const LocalTensor<T>& dst, const LocalTensor<U>& selMask,652__aicore__ inline void Select(const LocalTensor<T>& dst, const LocalTensor<U>& selMask,
602 const LocalTensor<T>& src0, T src1, SELMODE selMode, uint32_t count)653 const LocalTensor<T>& src0, T src1, SELMODE selMode, uint32_t count)
603{654{
655+#ifdef __MSTX_DFX_REPORT__
656+ MstxTensor::GetMstxVecBinarySelInfo(dst, selMask, src0, count, true, selMode, "Select");
657+#endif
604 using DataPrimType = PrimT<T>;658 using DataPrimType = PrimT<T>;
605 using MaskPrimType = PrimT<U>;659 using MaskPrimType = PrimT<U>;
606#if ASCENDC_CPU_DEBUG660#if ASCENDC_CPU_DEBUG
@@ -912,4 +966,4 @@ __aicore__ inline void Select(const T2& dst, const LocalTensor<T1>& selMask,
912#endif966#endif
913} // namespace AscendC967} // namespace AscendC
914#pragma end_pipe968#pragma end_pipe
915-#endif // ASCENDC_MODULE_OPERATOR_VEC_CMPSEL_INTERFACE_IMPL_H969+#endif // ASCENDC_MODULE_OPERATOR_VEC_CMPSEL_INTERFACE_IMPL_H
Mimpl/basic_api/kernel_operator_vec_createvecindex_intf_impl.h+10-0
@@ -16,6 +16,7 @@
16#define ASCENDC_MODULE_OPERATOR_VEC_CREATEVECINDEX_INTERFACE_IMPL_H16#define ASCENDC_MODULE_OPERATOR_VEC_CREATEVECINDEX_INTERFACE_IMPL_H
17#include "kernel_tensor.h"17#include "kernel_tensor.h"
18#include "kernel_check.h"18#include "kernel_check.h"
19+#include "mstx_local_tensor_info.h"
19 20 
20#if __NPU_ARCH__ == 100121#if __NPU_ARCH__ == 1001
21#include "dav_c100/kernel_operator_vec_createvecindex_impl.h"22#include "dav_c100/kernel_operator_vec_createvecindex_impl.h"
@@ -43,6 +44,9 @@ template <typename T>
43__aicore__ inline __in_pipe__(S) __out_pipe__(V) void CreateVecIndex(LocalTensor<T> &dst, const T &firstValue,44__aicore__ inline __in_pipe__(S) __out_pipe__(V) void CreateVecIndex(LocalTensor<T> &dst, const T &firstValue,
44 uint64_t mask, uint8_t repeatTime, uint16_t dstBlkStride, uint8_t dstRepStride)45 uint64_t mask, uint8_t repeatTime, uint16_t dstBlkStride, uint8_t dstRepStride)
45{46{
47+#ifdef __MSTX_DFX_REPORT__
48+ MstxTensor::GetMstxVecIndexInfo(dst, mask, repeatTime, dstBlkStride, dstRepStride, "CreateVecIndex");
49+#endif
46 ASCENDC_ASSERT((SupportType<T, half, int16_t, float, int32_t>()), {KERNEL_LOG(KERNEL_ERROR, "Failed to check "50 ASCENDC_ASSERT((SupportType<T, half, int16_t, float, int32_t>()), {KERNEL_LOG(KERNEL_ERROR, "Failed to check "
47 "dtype in CreateVecIndex, current api support dtype combination is dst: half / int16_t / float / int32_t");});51 "dtype in CreateVecIndex, current api support dtype combination is dst: half / int16_t / float / int32_t");});
48#if ASCENDC_CPU_DEBUG52#if ASCENDC_CPU_DEBUG
@@ -57,6 +61,9 @@ template <typename T>
57__aicore__ inline __in_pipe__(S) __out_pipe__(V) void CreateVecIndex(LocalTensor<T> &dst, const T &firstValue,61__aicore__ inline __in_pipe__(S) __out_pipe__(V) void CreateVecIndex(LocalTensor<T> &dst, const T &firstValue,
58 uint64_t mask[], uint8_t repeatTime, uint16_t dstBlkStride, uint8_t dstRepStride)62 uint64_t mask[], uint8_t repeatTime, uint16_t dstBlkStride, uint8_t dstRepStride)
59{63{
64+#ifdef __MSTX_DFX_REPORT__
65+ MstxTensor::GetMstxVecIndexInfo(dst, mask[0], mask[1], repeatTime, dstBlkStride, dstRepStride, "CreateVecIndex");
66+#endif
60 ASCENDC_ASSERT((SupportType<T, half, int16_t, float, int32_t>()), {KERNEL_LOG(KERNEL_ERROR, "Failed to check "67 ASCENDC_ASSERT((SupportType<T, half, int16_t, float, int32_t>()), {KERNEL_LOG(KERNEL_ERROR, "Failed to check "
61 "dtype in CreateVecIndex, current api support dtype combination is dst: half / int16_t / float / int32_t");});68 "dtype in CreateVecIndex, current api support dtype combination is dst: half / int16_t / float / int32_t");});
62#if ASCENDC_CPU_DEBUG69#if ASCENDC_CPU_DEBUG
@@ -71,6 +78,9 @@ template <typename T>
71__aicore__ inline __in_pipe__(S) __out_pipe__(V) void CreateVecIndex(LocalTensor<T> dst, const T &firstValue,78__aicore__ inline __in_pipe__(S) __out_pipe__(V) void CreateVecIndex(LocalTensor<T> dst, const T &firstValue,
72 uint32_t count)79 uint32_t count)
73{80{
81+#ifdef __MSTX_DFX_REPORT__
82+ MstxTensor::GetMstxVecIndexInfo(dst, count, "CreateVecIndex");
83+#endif
74#if ASCENDC_CPU_DEBUG84#if ASCENDC_CPU_DEBUG
75 if (!CheckFuncCreateVecIndex(dst, count, "CreateVecIndex")) {85 if (!CheckFuncCreateVecIndex(dst, count, "CreateVecIndex")) {
76 ASCENDC_REPORT_CHECK_ERROR("CreateVecIndex", KernelFuncType::CALCOUNT_MODE);86 ASCENDC_REPORT_CHECK_ERROR("CreateVecIndex", KernelFuncType::CALCOUNT_MODE);
Mimpl/basic_api/kernel_operator_vec_duplicate_intf_impl.h+13-0
@@ -16,6 +16,7 @@
16#define ASCENDC_MODULE_OPERATOR_VEC_DUPLICATE_INTERFACE_IMPL_H16#define ASCENDC_MODULE_OPERATOR_VEC_DUPLICATE_INTERFACE_IMPL_H
17#include "kernel_tensor.h"17#include "kernel_tensor.h"
18#include "kernel_check.h"18#include "kernel_check.h"
19+#include "mstx_local_tensor_info.h"
19 20 
20#if __NPU_ARCH__ == 100121#if __NPU_ARCH__ == 1001
21#include "dav_c100/kernel_operator_vec_duplicate_impl.h"22#include "dav_c100/kernel_operator_vec_duplicate_impl.h"
@@ -58,6 +59,9 @@ template <typename T, bool isSetMask>
58__aicore__ inline void Duplicate(const LocalTensor<T>& dst, const T& scalarValue, uint64_t mask,59__aicore__ inline void Duplicate(const LocalTensor<T>& dst, const T& scalarValue, uint64_t mask,
59 const uint8_t repeatTime, const uint16_t dstBlockStride, const uint8_t dstRepeatStride)60 const uint8_t repeatTime, const uint16_t dstBlockStride, const uint8_t dstRepeatStride)
60{61{
62+#ifdef __MSTX_DFX_REPORT__
63+ MstxTensor::GetMstxVecDupInfo(dst, mask, repeatTime, dstBlockStride, dstRepeatStride, isSetMask, "Duplicate");
64+#endif
61 CheckDuplicateSupportedType<T>();65 CheckDuplicateSupportedType<T>();
62#if ASCENDC_CPU_DEBUG66#if ASCENDC_CPU_DEBUG
63 MaskSetter::Instance().SetMask(isSetMask);67 MaskSetter::Instance().SetMask(isSetMask);
@@ -73,6 +77,9 @@ template <typename T, bool isSetMask>
73__aicore__ inline void Duplicate(const LocalTensor<T>& dst, const T& scalarValue, uint64_t mask[],77__aicore__ inline void Duplicate(const LocalTensor<T>& dst, const T& scalarValue, uint64_t mask[],
74 const uint8_t repeatTime, const uint16_t dstBlockStride, const uint8_t dstRepeatStride)78 const uint8_t repeatTime, const uint16_t dstBlockStride, const uint8_t dstRepeatStride)
75{79{
80+#ifdef __MSTX_DFX_REPORT__
81+ MstxTensor::GetMstxVecDupInfo(dst, mask[0], mask[1], repeatTime, dstBlockStride, dstRepeatStride, isSetMask, "Duplicate");
82+#endif
76 CheckDuplicateSupportedType<T>();83 CheckDuplicateSupportedType<T>();
77#if ASCENDC_CPU_DEBUG84#if ASCENDC_CPU_DEBUG
78 MaskSetter::Instance().SetMask(isSetMask);85 MaskSetter::Instance().SetMask(isSetMask);
@@ -94,6 +101,9 @@ __aicore__ inline void Duplicate(const LocalTensor<T>& dst, const T& scalarValue
94template <typename T>101template <typename T>
95__aicore__ inline void Duplicate(const LocalTensor<T>& dst, const T& scalarValue, const int32_t& count)102__aicore__ inline void Duplicate(const LocalTensor<T>& dst, const T& scalarValue, const int32_t& count)
96{103{
104+#ifdef __MSTX_DFX_REPORT__
105+ MstxTensor::GetMstxVecDupInfo(dst, count, "Duplicate");
106+#endif
97 CheckDuplicateSupportedType<T>();107 CheckDuplicateSupportedType<T>();
98#if ASCENDC_CPU_DEBUG108#if ASCENDC_CPU_DEBUG
99 if (!CheckFunDup(dst, count, "Duplicate")) {109 if (!CheckFunDup(dst, count, "Duplicate")) {
@@ -115,6 +125,9 @@ template <typename T>
115__aicore__ inline void Duplicate(const LocalTensor<T>& dst, const LocalTensor<T>& src,125__aicore__ inline void Duplicate(const LocalTensor<T>& dst, const LocalTensor<T>& src,
116 const int32_t& count)126 const int32_t& count)
117{127{
128+#ifdef __MSTX_DFX_REPORT__
129+ MstxTensor::GetMstxVecDupInfo(dst, count, "Duplicate");
130+#endif
118 using PrimType = PrimT<T>;131 using PrimType = PrimT<T>;
119 CheckDuplicateSupportedType<PrimType>();132 CheckDuplicateSupportedType<PrimType>();
120#if ASCENDC_CPU_DEBUG133#if ASCENDC_CPU_DEBUG
Mimpl/basic_api/kernel_operator_vec_gather_mask_intf_impl.h+7-0
@@ -16,6 +16,7 @@
16#define ASCENDC_MODULE_OPERATOR_VEC_REDUCEV2_INTERFACE_IMPL_H16#define ASCENDC_MODULE_OPERATOR_VEC_REDUCEV2_INTERFACE_IMPL_H
17#include "kernel_tensor.h"17#include "kernel_tensor.h"
18#include "kernel_struct_gather.h"18#include "kernel_struct_gather.h"
19+#include "mstx_local_tensor_info.h"
19#if __NPU_ARCH__ == 100120#if __NPU_ARCH__ == 1001
20#include "dav_c100/kernel_operator_vec_gather_mask_impl.h"21#include "dav_c100/kernel_operator_vec_gather_mask_impl.h"
21#elif __NPU_ARCH__ == 200222#elif __NPU_ARCH__ == 2002
@@ -44,6 +45,9 @@ __aicore__ inline void GatherMask(const LocalTensor<T>& dst, const LocalTensor<T
44 const LocalTensor<U>& src1Pattern, const bool reduceMode, const uint32_t mask,45 const LocalTensor<U>& src1Pattern, const bool reduceMode, const uint32_t mask,
45 const GatherMaskParams& gatherMaskParams, uint64_t& rsvdCnt)46 const GatherMaskParams& gatherMaskParams, uint64_t& rsvdCnt)
46{47{
48+#ifdef __MSTX_DFX_REPORT__
49+ MstxTensor::GetMstxVecGatherMaskInfo(dst, src0, mask, gatherMaskParams, mode, "GatherMask");
50+#endif
47 using DstPrimType = PrimT<T>;51 using DstPrimType = PrimT<T>;
48 using Src1PrimType = PrimT<U>;52 using Src1PrimType = PrimT<U>;
49#if ASCENDC_CPU_DEBUG53#if ASCENDC_CPU_DEBUG
@@ -69,6 +73,9 @@ __aicore__ inline void GatherMask(const LocalTensor<T>& dst, const LocalTensor<T
69 const uint8_t src1Pattern, const bool reduceMode, const uint32_t mask, const GatherMaskParams& gatherMaskParams,73 const uint8_t src1Pattern, const bool reduceMode, const uint32_t mask, const GatherMaskParams& gatherMaskParams,
70 uint64_t& rsvdCnt)74 uint64_t& rsvdCnt)
71{75{
76+#ifdef __MSTX_DFX_REPORT__
77+ MstxTensor::GetMstxVecGatherMaskInfo(dst, src0, mask, gatherMaskParams, mode, "GatherMask");
78+#endif
72 using PrimType = PrimT<T>;79 using PrimType = PrimT<T>;
73#if ASCENDC_CPU_DEBUG80#if ASCENDC_CPU_DEBUG
74 if (!CheckFuncVecGatherMask(dst, src0, src1Pattern, reduceMode, mask, gatherMaskParams, rsvdCnt,81 if (!CheckFuncVecGatherMask(dst, src0, src1Pattern, reduceMode, mask, gatherMaskParams, rsvdCnt,
Mimpl/basic_api/kernel_operator_vec_mulcast_intf_impl.h+10-0
@@ -17,6 +17,7 @@
17#include "kernel_tensor.h"17#include "kernel_tensor.h"
18#include "kernel_check.h"18#include "kernel_check.h"
19#include "kernel_struct_binary.h"19#include "kernel_struct_binary.h"
20+#include "mstx_local_tensor_info.h"
20 21 
21#if __NPU_ARCH__ == 100122#if __NPU_ARCH__ == 1001
22#include "dav_c100/kernel_operator_vec_mulcast_impl.h"23#include "dav_c100/kernel_operator_vec_mulcast_impl.h"
@@ -53,6 +54,9 @@ __aicore__ inline void MulCast(const LocalTensor<T> &dst, const LocalTensor<U> &
53 if (!CheckFuncVecBinaryDiffType(dst, src0, src1, mask, repeatTime, repeatParams, "MulCast")) {54 if (!CheckFuncVecBinaryDiffType(dst, src0, src1, mask, repeatTime, repeatParams, "MulCast")) {
54 ASCENDC_REPORT_CHECK_ERROR("MulCast", KernelFuncType::MASK_COUNT_MODE);55 ASCENDC_REPORT_CHECK_ERROR("MulCast", KernelFuncType::MASK_COUNT_MODE);
55 }56 }
57+#endif
58+#ifdef __MSTX_DFX_REPORT__
59+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, mask, repeatTime, repeatParams, isSetMask, "MulCast");
56#endif60#endif
57 MulCastCalc<DstPrimType, SrcPrimType, isSetMask>(dst, src0, src1, mask, repeatTime, repeatParams);61 MulCastCalc<DstPrimType, SrcPrimType, isSetMask>(dst, src0, src1, mask, repeatTime, repeatParams);
58}62}
@@ -71,6 +75,9 @@ __aicore__ inline void MulCast(const LocalTensor<T> &dst, const LocalTensor<U> &
71 if (!CheckFuncVecBinaryDiffType(dst, src0, src1, mask, repeatTime, repeatParams, "MulCast")) {75 if (!CheckFuncVecBinaryDiffType(dst, src0, src1, mask, repeatTime, repeatParams, "MulCast")) {
72 ASCENDC_REPORT_CHECK_ERROR("MulCast", KernelFuncType::MASK_BIT_MODE);76 ASCENDC_REPORT_CHECK_ERROR("MulCast", KernelFuncType::MASK_BIT_MODE);
73 }77 }
78+#endif
79+#ifdef __MSTX_DFX_REPORT__
80+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, mask[0], mask[1], repeatTime, repeatParams, isSetMask, "MulCast");
74#endif81#endif
75 MulCastCalc<DstPrimType, SrcPrimType, isSetMask>(dst, src0, src1, mask, repeatTime, repeatParams);82 MulCastCalc<DstPrimType, SrcPrimType, isSetMask>(dst, src0, src1, mask, repeatTime, repeatParams);
76}83}
@@ -83,6 +90,9 @@ __aicore__ inline void MulCast(const LocalTensor<T> &dst, const LocalTensor<U> &
83 if (!CheckFuncVecBinaryDiffType(dst, src0, src1, count, "MulCast")) {90 if (!CheckFuncVecBinaryDiffType(dst, src0, src1, count, "MulCast")) {
84 ASCENDC_REPORT_CHECK_ERROR("MulCast", KernelFuncType::CALCOUNT_MODE);91 ASCENDC_REPORT_CHECK_ERROR("MulCast", KernelFuncType::CALCOUNT_MODE);
85 }92 }
93+#endif
94+#ifdef __MSTX_DFX_REPORT__
95+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, "MulCast", count);
86#endif96#endif
87 MulCastCalc(dst, src0, src1, count);97 MulCastCalc(dst, src0, src1, count);
88}98}
Mimpl/basic_api/kernel_operator_vec_reduce_intf_impl.h+82-0
@@ -15,6 +15,7 @@
15#ifndef ASCENDC_MODULE_OPERATOR_VEC_REDUCE_INTERFACE_IMPL_H15#ifndef ASCENDC_MODULE_OPERATOR_VEC_REDUCE_INTERFACE_IMPL_H
16#define ASCENDC_MODULE_OPERATOR_VEC_REDUCE_INTERFACE_IMPL_H16#define ASCENDC_MODULE_OPERATOR_VEC_REDUCE_INTERFACE_IMPL_H
17#include "kernel_tensor.h"17#include "kernel_tensor.h"
18+#include "mstx_local_tensor_info.h"
18 19 
19#if __NPU_ARCH__ == 100120#if __NPU_ARCH__ == 1001
20#include "dav_c100/kernel_operator_vec_reduce_impl.h"21#include "dav_c100/kernel_operator_vec_reduce_impl.h"
@@ -57,6 +58,9 @@ __aicore__ inline void BlockReduceSum(const LocalTensor<T>& dst, const LocalTens
57 const int32_t repeatTime, const int32_t mask, const int32_t dstRepStride, const int32_t srcBlkStride,58 const int32_t repeatTime, const int32_t mask, const int32_t dstRepStride, const int32_t srcBlkStride,
58 const int32_t srcRepStride)59 const int32_t srcRepStride)
59{60{
61+#ifdef __MSTX_DFX_REPORT__
62+ MstxTensor::GetMstxVecReduceBlkInfo(dst, src, mask, repeatTime, dstRepStride, srcBlkStride, srcRepStride, isSetMask, "BlockReduceSum");
63+#endif
60 using PrimType = PrimT<T>;64 using PrimType = PrimT<T>;
61 ASCENDC_ASSERT((SupportType<PrimType, half, float>()), { KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in "65 ASCENDC_ASSERT((SupportType<PrimType, half, float>()), { KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in "
62 "BlockReduceSum, current api support dtype combination is src and dst both: half / float");});66 "BlockReduceSum, current api support dtype combination is src and dst both: half / float");});
@@ -88,6 +92,9 @@ __aicore__ inline void BlockReduceMax(const LocalTensor<T>& dst, const LocalTens
88 const int32_t repeatTime, const int32_t mask, const int32_t dstRepStride, const int32_t srcBlkStride,92 const int32_t repeatTime, const int32_t mask, const int32_t dstRepStride, const int32_t srcBlkStride,
89 const int32_t srcRepStride)93 const int32_t srcRepStride)
90{94{
95+#ifdef __MSTX_DFX_REPORT__
96+ MstxTensor::GetMstxVecReduceBlkInfo(dst, src, mask, repeatTime, dstRepStride, srcBlkStride, srcRepStride, isSetMask, "BlockReduceMax");
97+#endif
91 using PrimType = PrimT<T>;98 using PrimType = PrimT<T>;
92 ASCENDC_ASSERT((SupportType<PrimType, half, float>()), { KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in "99 ASCENDC_ASSERT((SupportType<PrimType, half, float>()), { KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in "
93 "BlockReduceMax, current api support dtype combination is src and dst both: half / float");});100 "BlockReduceMax, current api support dtype combination is src and dst both: half / float");});
@@ -119,6 +126,9 @@ __aicore__ inline void BlockReduceMin(const LocalTensor<T>& dst, const LocalTens
119 const int32_t repeatTime, const int32_t mask, const int32_t dstRepStride, const int32_t srcBlkStride,126 const int32_t repeatTime, const int32_t mask, const int32_t dstRepStride, const int32_t srcBlkStride,
120 const int32_t srcRepStride)127 const int32_t srcRepStride)
121{128{
129+#ifdef __MSTX_DFX_REPORT__
130+ MstxTensor::GetMstxVecReduceBlkInfo(dst, src, mask, repeatTime, dstRepStride, srcBlkStride, srcRepStride, isSetMask, "BlockReduceMin");
131+#endif
122 using PrimType = PrimT<T>;132 using PrimType = PrimT<T>;
123 ASCENDC_ASSERT((SupportType<PrimType, half, float>()), { KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in "133 ASCENDC_ASSERT((SupportType<PrimType, half, float>()), { KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in "
124 "BlockReduceMin, current api support dtype combination is src and dst both: half / float");});134 "BlockReduceMin, current api support dtype combination is src and dst both: half / float");});
@@ -150,6 +160,9 @@ __aicore__ inline void PairReduceSum(const LocalTensor<T>& dst, const LocalTenso
150 const int32_t repeatTime, const int32_t mask, const int32_t dstRepStride, const int32_t srcBlkStride,160 const int32_t repeatTime, const int32_t mask, const int32_t dstRepStride, const int32_t srcBlkStride,
151 const int32_t srcRepStride)161 const int32_t srcRepStride)
152{162{
163+#ifdef __MSTX_DFX_REPORT__
164+ MstxTensor::GetMstxVecReducePairInfo(dst, src, mask, repeatTime, dstRepStride, srcBlkStride, srcRepStride, isSetMask, "PairReduceSum");
165+#endif
153 using PrimType = PrimT<T>;166 using PrimType = PrimT<T>;
154 ASCENDC_ASSERT((SupportType<PrimType, half, float>()), { KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in "167 ASCENDC_ASSERT((SupportType<PrimType, half, float>()), { KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in "
155 "PairReduceSum, current api support dtype combination is src and dst both: half / float");});168 "PairReduceSum, current api support dtype combination is src and dst both: half / float");});
@@ -170,6 +183,9 @@ __aicore__ inline void BlockReduceSum(const LocalTensor<T>& dst, const LocalTens
170 const int32_t repeatTime, const uint64_t mask[], const int32_t dstRepStride, const int32_t srcBlkStride,183 const int32_t repeatTime, const uint64_t mask[], const int32_t dstRepStride, const int32_t srcBlkStride,
171 const int32_t srcRepStride)184 const int32_t srcRepStride)
172{185{
186+#ifdef __MSTX_DFX_REPORT__
187+ MstxTensor::GetMstxVecReduceBlkInfo(dst, src, mask[0], mask[1], repeatTime, dstRepStride, srcBlkStride, srcRepStride, isSetMask, "BlockReduceSum");
188+#endif
173 using PrimType = PrimT<T>;189 using PrimType = PrimT<T>;
174 ASCENDC_ASSERT((SupportType<PrimType, half, float>()), { KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in "190 ASCENDC_ASSERT((SupportType<PrimType, half, float>()), { KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in "
175 "BlockReduceSum, current api support dtype combination is src and dst both: half / float");});191 "BlockReduceSum, current api support dtype combination is src and dst both: half / float");});
@@ -190,6 +206,9 @@ __aicore__ inline void BlockReduceMax(const LocalTensor<T>& dst, const LocalTens
190 const int32_t repeatTime, const uint64_t mask[], const int32_t dstRepStride, const int32_t srcBlkStride,206 const int32_t repeatTime, const uint64_t mask[], const int32_t dstRepStride, const int32_t srcBlkStride,
191 const int32_t srcRepStride)207 const int32_t srcRepStride)
192{208{
209+#ifdef __MSTX_DFX_REPORT__
210+ MstxTensor::GetMstxVecReduceBlkInfo(dst, src, mask[0], mask[1], repeatTime, dstRepStride, srcBlkStride, srcRepStride, isSetMask, "BlockReduceMax");
211+#endif
193 using PrimType = PrimT<T>;212 using PrimType = PrimT<T>;
194 ASCENDC_ASSERT((SupportType<PrimType, half, float>()), { KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in "213 ASCENDC_ASSERT((SupportType<PrimType, half, float>()), { KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in "
195 "BlockReduceMax, current api support dtype combination is src and dst both: half / float");});214 "BlockReduceMax, current api support dtype combination is src and dst both: half / float");});
@@ -210,6 +229,9 @@ __aicore__ inline void BlockReduceMin(const LocalTensor<T>& dst, const LocalTens
210 const int32_t repeatTime, const uint64_t mask[], const int32_t dstRepStride, const int32_t srcBlkStride,229 const int32_t repeatTime, const uint64_t mask[], const int32_t dstRepStride, const int32_t srcBlkStride,
211 const int32_t srcRepStride)230 const int32_t srcRepStride)
212{231{
232+#ifdef __MSTX_DFX_REPORT__
233+ MstxTensor::GetMstxVecReduceBlkInfo(dst, src, mask[0], mask[1], repeatTime, dstRepStride, srcBlkStride, srcRepStride, isSetMask, "BlockReduceMin");
234+#endif
213 using PrimType = PrimT<T>;235 using PrimType = PrimT<T>;
214 ASCENDC_ASSERT((SupportType<PrimType, half, float>()), { KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in "236 ASCENDC_ASSERT((SupportType<PrimType, half, float>()), { KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in "
215 "BlockReduceMin, current api support dtype combination is src and dst both: half / float");});237 "BlockReduceMin, current api support dtype combination is src and dst both: half / float");});
@@ -230,6 +252,9 @@ __aicore__ inline void PairReduceSum(const LocalTensor<T>& dst, const LocalTenso
230 const int32_t repeatTime, const uint64_t mask[], const int32_t dstRepStride, const int32_t srcBlkStride,252 const int32_t repeatTime, const uint64_t mask[], const int32_t dstRepStride, const int32_t srcBlkStride,
231 const int32_t srcRepStride)253 const int32_t srcRepStride)
232{254{
255+#ifdef __MSTX_DFX_REPORT__
256+ MstxTensor::GetMstxVecReducePairInfo(dst, src, mask[0], mask[1], repeatTime, dstRepStride, srcBlkStride, srcRepStride, isSetMask, "PairReduceSum");
257+#endif
233 using PrimType = PrimT<T>;258 using PrimType = PrimT<T>;
234 ASCENDC_ASSERT((SupportType<PrimType, half, float>()), { KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in "259 ASCENDC_ASSERT((SupportType<PrimType, half, float>()), { KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in "
235 "PairReduceSum, current api support dtype combination is src and dst both: half / float");});260 "PairReduceSum, current api support dtype combination is src and dst both: half / float");});
@@ -251,6 +276,9 @@ __aicore__ inline void RepeatReduceSum(const LocalTensor<U>& dst, const LocalTen
251 const int32_t repeatTime, const int32_t mask, const int32_t dstBlkStride, const int32_t srcBlkStride,276 const int32_t repeatTime, const int32_t mask, const int32_t dstBlkStride, const int32_t srcBlkStride,
252 const int32_t dstRepStride, const int32_t srcRepStride)277 const int32_t dstRepStride, const int32_t srcRepStride)
253{278{
279+#ifdef __MSTX_DFX_REPORT__
280+ MstxTensor::GetMstxVecReduceRepeatInfo(dst, src, mask, repeatTime, dstRepStride, srcBlkStride, srcRepStride, isSetMask, "RepeatReduceSum");
281+#endif
254 using DstPrimType = PrimT<U>;282 using DstPrimType = PrimT<U>;
255 using SrcPrimType = PrimT<T>;283 using SrcPrimType = PrimT<T>;
256 ASCENDC_CHECK_VALUE_RANGE(repeatTime, 0, 255, "repeatTime", "RepeatReduceSum");284 ASCENDC_CHECK_VALUE_RANGE(repeatTime, 0, 255, "repeatTime", "RepeatReduceSum");
@@ -270,6 +298,9 @@ __aicore__ inline void RepeatReduceSum(const LocalTensor<T>& dst, const LocalTen
270 const int32_t repeatTime, const int32_t mask, const int32_t dstBlkStride, const int32_t srcBlkStride,298 const int32_t repeatTime, const int32_t mask, const int32_t dstBlkStride, const int32_t srcBlkStride,
271 const int32_t dstRepStride, const int32_t srcRepStride)299 const int32_t dstRepStride, const int32_t srcRepStride)
272{300{
301+#ifdef __MSTX_DFX_REPORT__
302+ MstxTensor::GetMstxVecReduceRepeatInfo(dst, src, mask, repeatTime, dstRepStride, srcBlkStride, srcRepStride, isSetMask, "RepeatReduceSum");
303+#endif
273 using PrimType = PrimT<T>;304 using PrimType = PrimT<T>;
274 ASCENDC_ASSERT((SupportType<PrimType, half, float>()), { KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in "305 ASCENDC_ASSERT((SupportType<PrimType, half, float>()), { KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in "
275 "RepeatReduceSum, current api support dtype combination is src and dst both: half / float");});306 "RepeatReduceSum, current api support dtype combination is src and dst both: half / float");});
@@ -304,6 +335,9 @@ __aicore__ inline void WholeReduceSum(const LocalTensor<U>& dst, const LocalTens
304 const uint64_t mask[], const int32_t repeatTime, const int32_t dstRepStride, const int32_t srcBlkStride,335 const uint64_t mask[], const int32_t repeatTime, const int32_t dstRepStride, const int32_t srcBlkStride,
305 const int32_t srcRepStride)336 const int32_t srcRepStride)
306{337{
338+#ifdef __MSTX_DFX_REPORT__
339+ MstxTensor::GetMstxVecReduceInfo(dst, src, mask[0], mask[1], repeatTime, dstRepStride, srcBlkStride, srcRepStride, isSetMask, "WholeReduceSum");
340+#endif
307 using DstPrimType = PrimT<U>;341 using DstPrimType = PrimT<U>;
308 using SrcPrmiType = PrimT<T>;342 using SrcPrmiType = PrimT<T>;
309 ASCENDC_CHECK_VALUE_RANGE(repeatTime, 0, 255, "repeatTime", "WholeReduceSum");343 ASCENDC_CHECK_VALUE_RANGE(repeatTime, 0, 255, "repeatTime", "WholeReduceSum");
@@ -323,6 +357,9 @@ __aicore__ inline void WholeReduceSum(const LocalTensor<T>& dst, const LocalTens
323 const uint64_t mask[], const int32_t repeatTime, const int32_t dstRepStride, const int32_t srcBlkStride,357 const uint64_t mask[], const int32_t repeatTime, const int32_t dstRepStride, const int32_t srcBlkStride,
324 const int32_t srcRepStride)358 const int32_t srcRepStride)
325{359{
360+#ifdef __MSTX_DFX_REPORT__
361+ MstxTensor::GetMstxVecReduceInfo(dst, src, mask[0], mask[1], repeatTime, dstRepStride, srcBlkStride, srcRepStride, isSetMask, "WholeReduceSum");
362+#endif
326 using PrimType = PrimT<T>;363 using PrimType = PrimT<T>;
327 ASCENDC_ASSERT((SupportType<PrimType, half, float>()), { KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in "364 ASCENDC_ASSERT((SupportType<PrimType, half, float>()), { KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in "
328 "WholeReduceSum, current api support dtype combination is src and dst both: half / float");});365 "WholeReduceSum, current api support dtype combination is src and dst both: half / float");});
@@ -355,6 +392,9 @@ __aicore__ inline void WholeReduceMax(const LocalTensor<T>& dst, const LocalTens
355 const uint64_t mask[], const int32_t repeatTime, const int32_t dstRepStride, const int32_t srcBlkStride,392 const uint64_t mask[], const int32_t repeatTime, const int32_t dstRepStride, const int32_t srcBlkStride,
356 const int32_t srcRepStride, ReduceOrder order)393 const int32_t srcRepStride, ReduceOrder order)
357{394{
395+#ifdef __MSTX_DFX_REPORT__
396+ MstxTensor::GetMstxVecReduceInfo(dst, src, mask[0], mask[1], repeatTime, dstRepStride, srcBlkStride, srcRepStride, isSetMask, "WholeReduceMax");
397+#endif
358 using PrimType = PrimT<T>;398 using PrimType = PrimT<T>;
359#if !((__NPU_ARCH__ == 3101) || (__NPU_ARCH__ == 5102) || (__NPU_ARCH__ == 3003) || (__NPU_ARCH__ == 3113))399#if !((__NPU_ARCH__ == 3101) || (__NPU_ARCH__ == 5102) || (__NPU_ARCH__ == 3003) || (__NPU_ARCH__ == 3113))
360 ASCENDC_ASSERT((SupportType<PrimType, half, float>()), { KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in "400 ASCENDC_ASSERT((SupportType<PrimType, half, float>()), { KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in "
@@ -405,6 +445,9 @@ __aicore__ inline void WholeReduceMin(const LocalTensor<T>& dst, const LocalTens
405 const uint64_t mask[], const int32_t repeatTime, const int32_t dstRepStride, const int32_t srcBlkStride,445 const uint64_t mask[], const int32_t repeatTime, const int32_t dstRepStride, const int32_t srcBlkStride,
406 const int32_t srcRepStride, ReduceOrder order)446 const int32_t srcRepStride, ReduceOrder order)
407{447{
448+#ifdef __MSTX_DFX_REPORT__
449+ MstxTensor::GetMstxVecReduceInfo(dst, src, mask[0], mask[1], repeatTime, dstRepStride, srcBlkStride, srcRepStride, isSetMask, "WholeReduceMin");
450+#endif
408 using PrimType = PrimT<T>;451 using PrimType = PrimT<T>;
409#if !((__NPU_ARCH__ == 3101) || (__NPU_ARCH__ == 5102))452#if !((__NPU_ARCH__ == 3101) || (__NPU_ARCH__ == 5102))
410 ASCENDC_ASSERT((SupportType<PrimType, half, float>()), { KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in "453 ASCENDC_ASSERT((SupportType<PrimType, half, float>()), { KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in "
@@ -445,6 +488,9 @@ __aicore__ inline void WholeReduceSum(const LocalTensor<U>& dst, const LocalTens
445 const int32_t mask, const int32_t repeatTime, const int32_t dstRepStride, const int32_t srcBlkStride,488 const int32_t mask, const int32_t repeatTime, const int32_t dstRepStride, const int32_t srcBlkStride,
446 const int32_t srcRepStride)489 const int32_t srcRepStride)
447{490{
491+#ifdef __MSTX_DFX_REPORT__
492+ MstxTensor::GetMstxVecReduceInfo(dst, src, mask, repeatTime, dstRepStride, srcBlkStride, srcRepStride, isSetMask, "WholeReduceSum");
493+#endif
448 using DstPrimType = PrimT<U>;494 using DstPrimType = PrimT<U>;
449 using SrcPrimType = PrimT<T>;495 using SrcPrimType = PrimT<T>;
450 ASCENDC_CHECK_VALUE_RANGE(repeatTime, 0, 255, "repeatTime", "WholeReduceSum");496 ASCENDC_CHECK_VALUE_RANGE(repeatTime, 0, 255, "repeatTime", "WholeReduceSum");
@@ -464,6 +510,9 @@ __aicore__ inline void WholeReduceSum(const LocalTensor<T>& dst, const LocalTens
464 const int32_t mask, const int32_t repeatTime, const int32_t dstRepStride, const int32_t srcBlkStride,510 const int32_t mask, const int32_t repeatTime, const int32_t dstRepStride, const int32_t srcBlkStride,
465 const int32_t srcRepStride)511 const int32_t srcRepStride)
466{512{
513+#ifdef __MSTX_DFX_REPORT__
514+ MstxTensor::GetMstxVecReduceInfo(dst, src, mask, repeatTime, dstRepStride, srcBlkStride, srcRepStride, isSetMask, "WholeReduceSum");
515+#endif
467 using PrimType = PrimT<T>;516 using PrimType = PrimT<T>;
468 ASCENDC_ASSERT((SupportType<PrimType, half, float>()), { KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in "517 ASCENDC_ASSERT((SupportType<PrimType, half, float>()), { KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in "
469 "WholeReduceSum, current api support dtype combination is src and dst both: half / float");});518 "WholeReduceSum, current api support dtype combination is src and dst both: half / float");});
@@ -485,6 +534,9 @@ __aicore__ inline void WholeReduceMax(const LocalTensor<T>& dst, const LocalTens
485 const int32_t mask, const int32_t repeatTime, const int32_t dstRepStride, const int32_t srcBlkStride,534 const int32_t mask, const int32_t repeatTime, const int32_t dstRepStride, const int32_t srcBlkStride,
486 const int32_t srcRepStride, ReduceOrder order)535 const int32_t srcRepStride, ReduceOrder order)
487{536{
537+#ifdef __MSTX_DFX_REPORT__
538+ MstxTensor::GetMstxVecReduceInfo(dst, src, mask, repeatTime, dstRepStride, srcBlkStride, srcRepStride, isSetMask, "WholeReduceMax");
539+#endif
488 using PrimType = PrimT<T>;540 using PrimType = PrimT<T>;
489#if !((__NPU_ARCH__ == 3101) || (__NPU_ARCH__ == 5102) || __NPU_ARCH__ == 3003 || __NPU_ARCH__ == 3113)541#if !((__NPU_ARCH__ == 3101) || (__NPU_ARCH__ == 5102) || __NPU_ARCH__ == 3003 || __NPU_ARCH__ == 3113)
490 ASCENDC_ASSERT((SupportType<PrimType, half, float>()), { KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in "542 ASCENDC_ASSERT((SupportType<PrimType, half, float>()), { KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in "
@@ -518,6 +570,9 @@ __aicore__ inline void WholeReduceMin(const LocalTensor<T>& dst, const LocalTens
518 const int32_t mask, const int32_t repeatTime, const int32_t dstRepStride, const int32_t srcBlkStride,570 const int32_t mask, const int32_t repeatTime, const int32_t dstRepStride, const int32_t srcBlkStride,
519 const int32_t srcRepStride, ReduceOrder order)571 const int32_t srcRepStride, ReduceOrder order)
520{572{
573+#ifdef __MSTX_DFX_REPORT__
574+ MstxTensor::GetMstxVecReduceInfo(dst, src, mask, repeatTime, dstRepStride, srcBlkStride, srcRepStride, isSetMask, "WholeReduceMin");
575+#endif
521 using PrimType = PrimT<T>;576 using PrimType = PrimT<T>;
522#if !((__NPU_ARCH__ == 3101) || (__NPU_ARCH__ == 5102))577#if !((__NPU_ARCH__ == 3101) || (__NPU_ARCH__ == 5102))
523 ASCENDC_ASSERT((SupportType<PrimType, half, float>()), { KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in "578 ASCENDC_ASSERT((SupportType<PrimType, half, float>()), { KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in "
@@ -569,6 +624,9 @@ __aicore__ inline void ReduceMax(const LocalTensor<T>& dst, const LocalTensor<T>
569 const LocalTensor<T>& sharedTmpBuffer, const int32_t mask, const int32_t repeatTime, const int32_t srcRepStride,624 const LocalTensor<T>& sharedTmpBuffer, const int32_t mask, const int32_t repeatTime, const int32_t srcRepStride,
570 bool calIndex)625 bool calIndex)
571{626{
627+#ifdef __MSTX_DFX_REPORT__
628+ MstxTensor::GetMstxVecReduceComplexInfo(dst, src, sharedTmpBuffer, mask, repeatTime, srcRepStride, "ReduceMax");
629+#endif
572 using PrimType = PrimT<T>;630 using PrimType = PrimT<T>;
573#if ASCENDC_CPU_DEBUG631#if ASCENDC_CPU_DEBUG
574 if (!CheckFunVecReduce(dst, src, sharedTmpBuffer, repeatTime, mask, calIndex, srcRepStride, "ReduceMax")) {632 if (!CheckFunVecReduce(dst, src, sharedTmpBuffer, repeatTime, mask, calIndex, srcRepStride, "ReduceMax")) {
@@ -607,6 +665,9 @@ __aicore__ inline void ReduceMin(const LocalTensor<T>& dst, const LocalTensor<T>
607 const LocalTensor<T>& sharedTmpBuffer, const int32_t mask, const int32_t repeatTime, const int32_t srcRepStride,665 const LocalTensor<T>& sharedTmpBuffer, const int32_t mask, const int32_t repeatTime, const int32_t srcRepStride,
608 bool calIndex)666 bool calIndex)
609{667{
668+#ifdef __MSTX_DFX_REPORT__
669+ MstxTensor::GetMstxVecReduceComplexInfo(dst, src, sharedTmpBuffer, mask, repeatTime, srcRepStride, "ReduceMin");
670+#endif
610 using PrimType = PrimT<T>;671 using PrimType = PrimT<T>;
611#if ASCENDC_CPU_DEBUG672#if ASCENDC_CPU_DEBUG
612 if (!CheckFunVecReduce(dst, src, sharedTmpBuffer, repeatTime, mask, calIndex, srcRepStride, "ReduceMin")) {673 if (!CheckFunVecReduce(dst, src, sharedTmpBuffer, repeatTime, mask, calIndex, srcRepStride, "ReduceMin")) {
@@ -643,6 +704,9 @@ template <typename T>
643__aicore__ inline void ReduceSum(const LocalTensor<T>& dst, const LocalTensor<T>& src,704__aicore__ inline void ReduceSum(const LocalTensor<T>& dst, const LocalTensor<T>& src,
644 const LocalTensor<T>& sharedTmpBuffer, const int32_t mask, const int32_t repeatTime, const int32_t srcRepStride)705 const LocalTensor<T>& sharedTmpBuffer, const int32_t mask, const int32_t repeatTime, const int32_t srcRepStride)
645{706{
707+#ifdef __MSTX_DFX_REPORT__
708+ MstxTensor::GetMstxVecReduceComplexInfo(dst, src, sharedTmpBuffer, mask, repeatTime, srcRepStride, "ReduceSum");
709+#endif
646 using PrimType = PrimT<T>;710 using PrimType = PrimT<T>;
647#if ASCENDC_CPU_DEBUG711#if ASCENDC_CPU_DEBUG
648 if (!CheckFunVecReduce(dst, src, sharedTmpBuffer, repeatTime, mask, srcRepStride, "ReduceSum")) {712 if (!CheckFunVecReduce(dst, src, sharedTmpBuffer, repeatTime, mask, srcRepStride, "ReduceSum")) {
@@ -670,6 +734,9 @@ __aicore__ inline void ReduceMax(const LocalTensor<T>& dst, const LocalTensor<T>
670 const LocalTensor<T>& sharedTmpBuffer, const uint64_t mask[], const int32_t repeatTime, const int32_t srcRepStride,734 const LocalTensor<T>& sharedTmpBuffer, const uint64_t mask[], const int32_t repeatTime, const int32_t srcRepStride,
671 bool calIndex)735 bool calIndex)
672{736{
737+#ifdef __MSTX_DFX_REPORT__
738+ MstxTensor::GetMstxVecReduceComplexInfo(dst, src, sharedTmpBuffer, mask[0], mask[1], repeatTime, srcRepStride, "ReduceMax");
739+#endif
673 using PrimType = PrimT<T>;740 using PrimType = PrimT<T>;
674#if ASCENDC_CPU_DEBUG741#if ASCENDC_CPU_DEBUG
675 if (!CheckFunVecReduce(dst, src, sharedTmpBuffer, repeatTime, mask, calIndex, srcRepStride, "ReduceMax")) {742 if (!CheckFunVecReduce(dst, src, sharedTmpBuffer, repeatTime, mask, calIndex, srcRepStride, "ReduceMax")) {
@@ -697,6 +764,9 @@ __aicore__ inline void ReduceMin(const LocalTensor<T>& dst, const LocalTensor<T>
697 const LocalTensor<T>& sharedTmpBuffer, const uint64_t mask[], const int32_t repeatTime, const int32_t srcRepStride,764 const LocalTensor<T>& sharedTmpBuffer, const uint64_t mask[], const int32_t repeatTime, const int32_t srcRepStride,
698 bool calIndex)765 bool calIndex)
699{766{
767+#ifdef __MSTX_DFX_REPORT__
768+ MstxTensor::GetMstxVecReduceComplexInfo(dst, src, sharedTmpBuffer, mask[0], mask[1], repeatTime, srcRepStride, "ReduceMin");
769+#endif
700 using PrimType = PrimT<T>;770 using PrimType = PrimT<T>;
701#if ASCENDC_CPU_DEBUG771#if ASCENDC_CPU_DEBUG
702 if (!CheckFunVecReduce(dst, src, sharedTmpBuffer, repeatTime, mask, calIndex, srcRepStride, "ReduceMin")) {772 if (!CheckFunVecReduce(dst, src, sharedTmpBuffer, repeatTime, mask, calIndex, srcRepStride, "ReduceMin")) {
@@ -723,6 +793,9 @@ template <typename T>
723__aicore__ inline void ReduceSum(const LocalTensor<T>& dst, const LocalTensor<T>& src,793__aicore__ inline void ReduceSum(const LocalTensor<T>& dst, const LocalTensor<T>& src,
724 const LocalTensor<T>& sharedTmpBuffer, const uint64_t mask[], const int32_t repeatTime, const int32_t srcRepStride)794 const LocalTensor<T>& sharedTmpBuffer, const uint64_t mask[], const int32_t repeatTime, const int32_t srcRepStride)
725{795{
796+#ifdef __MSTX_DFX_REPORT__
797+ MstxTensor::GetMstxVecReduceComplexInfo(dst, src, sharedTmpBuffer, mask[0], mask[1], repeatTime, srcRepStride, "ReduceSum");
798+#endif
726 using PrimType = PrimT<T>;799 using PrimType = PrimT<T>;
727#if ASCENDC_CPU_DEBUG800#if ASCENDC_CPU_DEBUG
728 if (!CheckFunVecReduce(dst, src, sharedTmpBuffer, repeatTime, mask, srcRepStride, "ReduceSum")) {801 if (!CheckFunVecReduce(dst, src, sharedTmpBuffer, repeatTime, mask, srcRepStride, "ReduceSum")) {
@@ -758,6 +831,9 @@ template <typename T>
758__aicore__ inline void ReduceMin(const LocalTensor<T>& dst, const LocalTensor<T>& src,831__aicore__ inline void ReduceMin(const LocalTensor<T>& dst, const LocalTensor<T>& src,
759 const LocalTensor<T>& sharedTmpBuffer, const int32_t count, bool calIndex)832 const LocalTensor<T>& sharedTmpBuffer, const int32_t count, bool calIndex)
760{833{
834+#ifdef __MSTX_DFX_REPORT__
835+ MstxTensor::GetMstxVecReduceComplexInfo(dst, src, sharedTmpBuffer, count, "ReduceMin");
836+#endif
761 using PrimType = PrimT<T>;837 using PrimType = PrimT<T>;
762#if defined(__NPU_ARCH__) && ((__NPU_ARCH__ == 3101) || (__NPU_ARCH__ == 5102) || (__NPU_ARCH__ == 3003) || (__NPU_ARCH__ == 3113))838#if defined(__NPU_ARCH__) && ((__NPU_ARCH__ == 3101) || (__NPU_ARCH__ == 5102) || (__NPU_ARCH__ == 3003) || (__NPU_ARCH__ == 3113))
763#if ASCENDC_CPU_DEBUG839#if ASCENDC_CPU_DEBUG
@@ -815,6 +891,9 @@ template <typename T>
815__aicore__ inline void ReduceMax(const LocalTensor<T>& dst, const LocalTensor<T>& src,891__aicore__ inline void ReduceMax(const LocalTensor<T>& dst, const LocalTensor<T>& src,
816 const LocalTensor<T>& sharedTmpBuffer, const int32_t count, bool calIndex)892 const LocalTensor<T>& sharedTmpBuffer, const int32_t count, bool calIndex)
817{893{
894+#ifdef __MSTX_DFX_REPORT__
895+ MstxTensor::GetMstxVecReduceComplexInfo(dst, src, sharedTmpBuffer, count, "ReduceMax");
896+#endif
818 using PrimType = PrimT<T>;897 using PrimType = PrimT<T>;
819#if defined(__NPU_ARCH__) && ((__NPU_ARCH__ == 3101) || (__NPU_ARCH__ == 5102) || (__NPU_ARCH__ == 3003) || (__NPU_ARCH__ == 3113))898#if defined(__NPU_ARCH__) && ((__NPU_ARCH__ == 3101) || (__NPU_ARCH__ == 5102) || (__NPU_ARCH__ == 3003) || (__NPU_ARCH__ == 3113))
820#if ASCENDC_CPU_DEBUG899#if ASCENDC_CPU_DEBUG
@@ -872,6 +951,9 @@ template <typename T, bool isSetMask>
872__aicore__ inline void ReduceSum(const LocalTensor<T>& dst, const LocalTensor<T>& src,951__aicore__ inline void ReduceSum(const LocalTensor<T>& dst, const LocalTensor<T>& src,
873 const LocalTensor<T>& sharedTmpBuffer, const int32_t count)952 const LocalTensor<T>& sharedTmpBuffer, const int32_t count)
874{953{
954+#ifdef __MSTX_DFX_REPORT__
955+ MstxTensor::GetMstxVecReduceComplexInfo(dst, src, sharedTmpBuffer, count, "ReduceSum");
956+#endif
875 using PrimType = PrimT<T>;957 using PrimType = PrimT<T>;
876 ASCENDC_CHECK_VALUE_RANGE(count, 1, TOTAL_UB_SIZE / sizeof(PrimType), "count", "ReduceSum");958 ASCENDC_CHECK_VALUE_RANGE(count, 1, TOTAL_UB_SIZE / sizeof(PrimType), "count", "ReduceSum");
877#if __NPU_ARCH__ == 2201959#if __NPU_ARCH__ == 2201
Mimpl/basic_api/kernel_operator_vec_ternary_scalar_intf_impl.h+10-0
@@ -16,6 +16,7 @@
16#define ASCENDC_MODULE_OPERATOR_VEC_TERNARY_SCALAR_INTERFACE_IMPL_H16#define ASCENDC_MODULE_OPERATOR_VEC_TERNARY_SCALAR_INTERFACE_IMPL_H
17#include "kernel_tensor.h"17#include "kernel_tensor.h"
18#include "kernel_struct_unary.h"18#include "kernel_struct_unary.h"
19+#include "mstx_local_tensor_info.h"
19 20 
20#if __NPU_ARCH__ == 100121#if __NPU_ARCH__ == 1001
21#include "dav_c100/kernel_operator_vec_ternary_scalar_impl.h"22#include "dav_c100/kernel_operator_vec_ternary_scalar_impl.h"
@@ -62,6 +63,9 @@ __aicore__ inline void Axpy(const LocalTensor<T>& dst, const LocalTensor<U>& src
62 if (!CheckFunVecBinaryScalarDiffType(dst, src, scalarValue, mask, repeatTime, repeatParams, "Axpy")) {63 if (!CheckFunVecBinaryScalarDiffType(dst, src, scalarValue, mask, repeatTime, repeatParams, "Axpy")) {
63 ASCENDC_REPORT_CHECK_ERROR("Axpy", KernelFuncType::MASK_COUNT_MODE);64 ASCENDC_REPORT_CHECK_ERROR("Axpy", KernelFuncType::MASK_COUNT_MODE);
64 }65 }
66+#endif
67+#ifdef __MSTX_DFX_REPORT__
68+ MstxTensor::GetMstxVecUnaryTenaryInfo<T, U, isSetMask>(dst, src, mask, repeatTime, repeatParams, "Axpy");
65#endif69#endif
66 AxpyImpl<T, U, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ U*)src.GetPhyAddr(), scalarValue, mask,70 AxpyImpl<T, U, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ U*)src.GetPhyAddr(), scalarValue, mask,
67 repeatTime, repeatParams);71 repeatTime, repeatParams);
@@ -76,6 +80,9 @@ __aicore__ inline void Axpy(const LocalTensor<T>& dst, const LocalTensor<U>& src
76 if (!CheckFunVecBinaryScalarDiffType(dst, src, scalarValue, mask, repeatTime, repeatParams, "Axpy")) {80 if (!CheckFunVecBinaryScalarDiffType(dst, src, scalarValue, mask, repeatTime, repeatParams, "Axpy")) {
77 ASCENDC_REPORT_CHECK_ERROR("Axpy", KernelFuncType::MASK_BIT_MODE);81 ASCENDC_REPORT_CHECK_ERROR("Axpy", KernelFuncType::MASK_BIT_MODE);
78 }82 }
83+#endif
84+#ifdef __MSTX_DFX_REPORT__
85+ MstxTensor::GetMstxVecUnaryTenaryInfo<T, U, isSetMask>(dst, src, mask[0], mask[1], repeatTime, repeatParams, "Axpy");
79#endif86#endif
80 AxpyImpl<T, U, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ U*)src.GetPhyAddr(), scalarValue, mask,87 AxpyImpl<T, U, isSetMask>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ U*)src.GetPhyAddr(), scalarValue, mask,
81 repeatTime, repeatParams);88 repeatTime, repeatParams);
@@ -97,6 +104,9 @@ __aicore__ inline void Axpy(const LocalTensor<T>& dst, const LocalTensor<U>& src
97 if (!CheckFunVecBinaryScalarDiffType(dst, src, scalarValue, count, "Axpy")) {104 if (!CheckFunVecBinaryScalarDiffType(dst, src, scalarValue, count, "Axpy")) {
98 ASCENDC_REPORT_CHECK_ERROR("Axpy", KernelFuncType::CALCOUNT_MODE);105 ASCENDC_REPORT_CHECK_ERROR("Axpy", KernelFuncType::CALCOUNT_MODE);
99 }106 }
107+#endif
108+#ifdef __MSTX_DFX_REPORT__
109+ MstxTensor::GetMstxVecUnaryTenaryInfo<T, U, true>(dst, src, "Axpy", count);
100#endif110#endif
101 AxpyImpl<T, U>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ U*)src.GetPhyAddr(), scalarValue, count);111 AxpyImpl<T, U>((__ubuf__ T*)dst.GetPhyAddr(), (__ubuf__ U*)src.GetPhyAddr(), scalarValue, count);
102}112}
Mimpl/basic_api/kernel_operator_vec_transpose_intf_impl.h+7-0
@@ -18,6 +18,7 @@
18#include "kernel_tpipe.h"18#include "kernel_tpipe.h"
19#include "kernel_check.h"19#include "kernel_check.h"
20#include "kernel_struct_transpose.h"20#include "kernel_struct_transpose.h"
21+#include "mstx_local_tensor_info.h"
21 22 
22#if __NPU_ARCH__ == 100123#if __NPU_ARCH__ == 1001
23#include "dav_c100/kernel_operator_vec_transpose_impl.h"24#include "dav_c100/kernel_operator_vec_transpose_impl.h"
@@ -52,6 +53,9 @@ namespace AscendC {
52 */53 */
53template <typename T> __aicore__ inline void Transpose(const LocalTensor<T>& dst, const LocalTensor<T>& src)54template <typename T> __aicore__ inline void Transpose(const LocalTensor<T>& dst, const LocalTensor<T>& src)
54{55{
56+#ifdef __MSTX_DFX_REPORT__
57+ MstxTensor::GetMstxVecTransposeInfo(dst, src, "Transpose");
58+#endif
55 ASCENDC_ASSERT((SupportType<PrimT<T>, int16_t, uint16_t, half>()),59 ASCENDC_ASSERT((SupportType<PrimT<T>, int16_t, uint16_t, half>()),
56 {KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in Transpose, current api support dtype combination is "60 {KERNEL_LOG(KERNEL_ERROR, "Failed to check dtype in Transpose, current api support dtype combination is "
57 "src and dst both: int16_t, uint16_t, half");});61 "src and dst both: int16_t, uint16_t, half");});
@@ -122,6 +126,9 @@ template <typename T>
122__aicore__ inline void Transpose(const LocalTensor<T> &dst, const LocalTensor<T> &src,126__aicore__ inline void Transpose(const LocalTensor<T> &dst, const LocalTensor<T> &src,
123 const LocalTensor<uint8_t> &sharedTmpBuffer, const TransposeParamsExt &transposeParams)127 const LocalTensor<uint8_t> &sharedTmpBuffer, const TransposeParamsExt &transposeParams)
124{128{
129+#ifdef __MSTX_DFX_REPORT__
130+ MstxTensor::GetMstxVecTransposeTempInfo(dst, src, sharedTmpBuffer, "Transpose");
131+#endif
125#if ASCENDC_CPU_DEBUG132#if ASCENDC_CPU_DEBUG
126 if (!CheckFunTranspose(dst, src, sharedTmpBuffer, transposeParams, "Transpose")) {133 if (!CheckFunTranspose(dst, src, sharedTmpBuffer, transposeParams, "Transpose")) {
127 ASCENDC_REPORT_CHECK_ERROR("Transpose", KernelFuncType::NONE_MODE);134 ASCENDC_REPORT_CHECK_ERROR("Transpose", KernelFuncType::NONE_MODE);
Mimpl/basic_api/kernel_operator_vec_unary_intf_impl.h+121-0
@@ -17,6 +17,7 @@
17#include "kernel_tensor.h"17#include "kernel_tensor.h"
18#include "kernel_check.h"18#include "kernel_check.h"
19#include "kernel_struct_unary.h"19#include "kernel_struct_unary.h"
20+#include "mstx_local_tensor_info.h"
20 21 
21#if __NPU_ARCH__ == 100122#if __NPU_ARCH__ == 1001
22#include "dav_c100/kernel_operator_vec_unary_impl.h"23#include "dav_c100/kernel_operator_vec_unary_impl.h"
@@ -68,6 +69,9 @@ __aicore__ inline void Relu(const LocalTensor<T>& dst, const LocalTensor<T>& src
68 "Relu")) {69 "Relu")) {
69 ASCENDC_REPORT_CHECK_ERROR("Relu", KernelFuncType::MASK_BIT_MODE);70 ASCENDC_REPORT_CHECK_ERROR("Relu", KernelFuncType::MASK_BIT_MODE);
70 }71 }
72+#endif
73+#ifdef __MSTX_DFX_REPORT__
74+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask[0], mask[1], repeatTime, repeatParams, "Relu");
71#endif75#endif
72 ReluImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),76 ReluImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),
73 mask, repeatTime, repeatParams);77 mask, repeatTime, repeatParams);
@@ -83,6 +87,9 @@ __aicore__ inline void Relu(const LocalTensor<T>& dst, const LocalTensor<T>& src
83 "Relu")) {87 "Relu")) {
84 ASCENDC_REPORT_CHECK_ERROR("Relu", KernelFuncType::MASK_COUNT_MODE);88 ASCENDC_REPORT_CHECK_ERROR("Relu", KernelFuncType::MASK_COUNT_MODE);
85 }89 }
90+#endif
91+#ifdef __MSTX_DFX_REPORT__
92+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask, repeatTime, repeatParams, "Relu");
86#endif93#endif
87 ReluImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),94 ReluImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),
88 mask, repeatTime, repeatParams);95 mask, repeatTime, repeatParams);
@@ -103,6 +110,9 @@ __aicore__ inline void Relu(const LocalTensor<T>& dst, const LocalTensor<T>& src
103 if (!CheckFunVecBinaryScalar(dst, src, static_cast<PrimType>(0), count, "Relu")) {110 if (!CheckFunVecBinaryScalar(dst, src, static_cast<PrimType>(0), count, "Relu")) {
104 ASCENDC_REPORT_CHECK_ERROR("Relu", KernelFuncType::CALCOUNT_MODE);111 ASCENDC_REPORT_CHECK_ERROR("Relu", KernelFuncType::CALCOUNT_MODE);
105 }112 }
113+#endif
114+#ifdef __MSTX_DFX_REPORT__
115+ MstxTensor::GetMstxVecUnaryInfo<T, T, true>(dst, src, "Relu", count);
C
Cchenyiyuan2月6日

调整下api name顺序

likedislike
106#endif116#endif
107 ReluImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(), count);117 ReluImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(), count);
108}118}
@@ -132,6 +142,9 @@ __aicore__ inline void Exp(const LocalTensor<T>& dst, const LocalTensor<T>& src,
132 "Exp")) {142 "Exp")) {
133 ASCENDC_REPORT_CHECK_ERROR("Exp", KernelFuncType::MASK_BIT_MODE);143 ASCENDC_REPORT_CHECK_ERROR("Exp", KernelFuncType::MASK_BIT_MODE);
134 }144 }
145+#endif
146+#ifdef __MSTX_DFX_REPORT__
147+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask[0], mask[1], repeatTime, repeatParams, "Exp");
135#endif148#endif
136 ExpImpl<PrimType, isSetMask, config>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),149 ExpImpl<PrimType, isSetMask, config>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),
137 mask, repeatTime, repeatParams);150 mask, repeatTime, repeatParams);
@@ -147,6 +160,9 @@ __aicore__ inline void Exp(const LocalTensor<T>& dst, const LocalTensor<T>& src,
147 "Exp")) {160 "Exp")) {
148 ASCENDC_REPORT_CHECK_ERROR("Exp", KernelFuncType::MASK_COUNT_MODE);161 ASCENDC_REPORT_CHECK_ERROR("Exp", KernelFuncType::MASK_COUNT_MODE);
149 }162 }
163+#endif
164+#ifdef __MSTX_DFX_REPORT__
165+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask, repeatTime, repeatParams, "Exp");
150#endif166#endif
151 ExpImpl<PrimType, isSetMask, config>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),167 ExpImpl<PrimType, isSetMask, config>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),
152 mask, repeatTime, repeatParams);168 mask, repeatTime, repeatParams);
@@ -163,6 +179,9 @@ __aicore__ inline void Exp(const LocalTensor<T>& dst, const LocalTensor<T>& src,
163 "Exp")) {179 "Exp")) {
164 ASCENDC_REPORT_CHECK_ERROR("Exp", KernelFuncType::MASK_BIT_MODE);180 ASCENDC_REPORT_CHECK_ERROR("Exp", KernelFuncType::MASK_BIT_MODE);
165 }181 }
182+#endif
183+#ifdef __MSTX_DFX_REPORT__
184+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask[0], mask[1], repeatTime, repeatParams, "Exp");
166#endif185#endif
167 ExpImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),186 ExpImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),
168 mask, repeatTime, repeatParams);187 mask, repeatTime, repeatParams);
@@ -178,6 +197,9 @@ __aicore__ inline void Exp(const LocalTensor<T>& dst, const LocalTensor<T>& src,
178 "Exp")) {197 "Exp")) {
179 ASCENDC_REPORT_CHECK_ERROR("Exp", KernelFuncType::MASK_COUNT_MODE);198 ASCENDC_REPORT_CHECK_ERROR("Exp", KernelFuncType::MASK_COUNT_MODE);
180 }199 }
200+#endif
201+#ifdef __MSTX_DFX_REPORT__
202+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask, repeatTime, repeatParams, "Exp");
181#endif203#endif
182 ExpImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),204 ExpImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),
183 mask, repeatTime, repeatParams);205 mask, repeatTime, repeatParams);
@@ -200,6 +222,9 @@ __aicore__ inline void Exp(const LocalTensor<T>& dst, const LocalTensor<T>& src,
200 if (!CheckFunVecBinaryScalar(dst, src, static_cast<PrimType>(0), count, "Exp")) {222 if (!CheckFunVecBinaryScalar(dst, src, static_cast<PrimType>(0), count, "Exp")) {
201 ASCENDC_REPORT_CHECK_ERROR("Exp", KernelFuncType::CALCOUNT_MODE);223 ASCENDC_REPORT_CHECK_ERROR("Exp", KernelFuncType::CALCOUNT_MODE);
202 }224 }
225+#endif
226+#ifdef __MSTX_DFX_REPORT__
227+ MstxTensor::GetMstxVecUnaryInfo<T, T, true>(dst, src, "Exp", count);
203#endif228#endif
204 ExpImpl<PrimType, config>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(), count);229 ExpImpl<PrimType, config>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(), count);
205}230}
@@ -212,6 +237,9 @@ __aicore__ inline void Exp(const LocalTensor<T>& dst, const LocalTensor<T>& src,
212 if (!CheckFunVecBinaryScalar(dst, src, static_cast<PrimType>(0), count, "Exp")) {237 if (!CheckFunVecBinaryScalar(dst, src, static_cast<PrimType>(0), count, "Exp")) {
213 ASCENDC_REPORT_CHECK_ERROR("Exp", KernelFuncType::CALCOUNT_MODE);238 ASCENDC_REPORT_CHECK_ERROR("Exp", KernelFuncType::CALCOUNT_MODE);
214 }239 }
240+#endif
241+#ifdef __MSTX_DFX_REPORT__
242+ MstxTensor::GetMstxVecUnaryInfo<T, T, true>(dst, src, "Exp", count);
215#endif243#endif
216 ExpImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(), count);244 ExpImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(), count);
217}245}
@@ -241,6 +269,9 @@ __aicore__ inline void Ln(const LocalTensor<T>& dst, const LocalTensor<T>& src,
241 if (!CheckFunVecBinaryScalar(dst, src, static_cast<PrimType>(0), mask, repeatTime, repeatParams, "Ln")) {269 if (!CheckFunVecBinaryScalar(dst, src, static_cast<PrimType>(0), mask, repeatTime, repeatParams, "Ln")) {
242 ASCENDC_REPORT_CHECK_ERROR("Ln", KernelFuncType::MASK_BIT_MODE);270 ASCENDC_REPORT_CHECK_ERROR("Ln", KernelFuncType::MASK_BIT_MODE);
243 }271 }
272+#endif
273+#ifdef __MSTX_DFX_REPORT__
274+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask[0], mask[1], repeatTime, repeatParams, "Ln");
244#endif275#endif
245 LnImpl<PrimType, isSetMask, config>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),276 LnImpl<PrimType, isSetMask, config>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),
246 mask, repeatTime, repeatParams);277 mask, repeatTime, repeatParams);
@@ -255,6 +286,9 @@ __aicore__ inline void Ln(const LocalTensor<T>& dst, const LocalTensor<T>& src,
255 if (!CheckFunVecBinaryScalar(dst, src, static_cast<PrimType>(0), mask, repeatTime, repeatParams, "Ln")) {286 if (!CheckFunVecBinaryScalar(dst, src, static_cast<PrimType>(0), mask, repeatTime, repeatParams, "Ln")) {
256 ASCENDC_REPORT_CHECK_ERROR("Ln", KernelFuncType::MASK_COUNT_MODE);287 ASCENDC_REPORT_CHECK_ERROR("Ln", KernelFuncType::MASK_COUNT_MODE);
257 }288 }
289+#endif
290+#ifdef __MSTX_DFX_REPORT__
291+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask, repeatTime, repeatParams, "Ln");
258#endif292#endif
259 LnImpl<PrimType, isSetMask, config>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),293 LnImpl<PrimType, isSetMask, config>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),
260 mask, repeatTime, repeatParams);294 mask, repeatTime, repeatParams);
@@ -270,6 +304,9 @@ __aicore__ inline void Ln(const LocalTensor<T>& dst, const LocalTensor<T>& src,
270 if (!CheckFunVecBinaryScalar(dst, src, static_cast<PrimType>(0), mask, repeatTime, repeatParams, "Ln")) {304 if (!CheckFunVecBinaryScalar(dst, src, static_cast<PrimType>(0), mask, repeatTime, repeatParams, "Ln")) {
271 ASCENDC_REPORT_CHECK_ERROR("Ln", KernelFuncType::MASK_BIT_MODE);305 ASCENDC_REPORT_CHECK_ERROR("Ln", KernelFuncType::MASK_BIT_MODE);
272 }306 }
307+#endif
308+#ifdef __MSTX_DFX_REPORT__
309+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask[0], mask[1], repeatTime, repeatParams, "Ln");
273#endif310#endif
274 LnImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),311 LnImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),
275 mask, repeatTime, repeatParams);312 mask, repeatTime, repeatParams);
@@ -284,6 +321,9 @@ __aicore__ inline void Ln(const LocalTensor<T>& dst, const LocalTensor<T>& src,
284 if (!CheckFunVecBinaryScalar(dst, src, static_cast<PrimType>(0), mask, repeatTime, repeatParams, "Ln")) {321 if (!CheckFunVecBinaryScalar(dst, src, static_cast<PrimType>(0), mask, repeatTime, repeatParams, "Ln")) {
285 ASCENDC_REPORT_CHECK_ERROR("Ln", KernelFuncType::MASK_COUNT_MODE);322 ASCENDC_REPORT_CHECK_ERROR("Ln", KernelFuncType::MASK_COUNT_MODE);
286 }323 }
324+#endif
325+#ifdef __MSTX_DFX_REPORT__
326+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask, repeatTime, repeatParams, "Ln");
287#endif327#endif
288 LnImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),328 LnImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),
289 mask, repeatTime, repeatParams);329 mask, repeatTime, repeatParams);
@@ -306,6 +346,9 @@ __aicore__ inline void Ln(const LocalTensor<T>& dst, const LocalTensor<T>& src,
306 if (!CheckFunVecBinaryScalar(dst, src, static_cast<PrimType>(0), count, "Ln")) {346 if (!CheckFunVecBinaryScalar(dst, src, static_cast<PrimType>(0), count, "Ln")) {
307 ASCENDC_REPORT_CHECK_ERROR("Ln", KernelFuncType::CALCOUNT_MODE);347 ASCENDC_REPORT_CHECK_ERROR("Ln", KernelFuncType::CALCOUNT_MODE);
308 }348 }
349+#endif
350+#ifdef __MSTX_DFX_REPORT__
351+ MstxTensor::GetMstxVecUnaryInfo<T, T, true>(dst, src, "Ln", count);
309#endif352#endif
310 LnImpl<PrimType, config>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(), count);353 LnImpl<PrimType, config>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(), count);
311}354}
@@ -318,6 +361,9 @@ __aicore__ inline void Ln(const LocalTensor<T>& dst, const LocalTensor<T>& src,
318 if (!CheckFunVecBinaryScalar(dst, src, static_cast<PrimType>(0), count, "Ln")) {361 if (!CheckFunVecBinaryScalar(dst, src, static_cast<PrimType>(0), count, "Ln")) {
319 ASCENDC_REPORT_CHECK_ERROR("Ln", KernelFuncType::CALCOUNT_MODE);362 ASCENDC_REPORT_CHECK_ERROR("Ln", KernelFuncType::CALCOUNT_MODE);
320 }363 }
364+#endif
365+#ifdef __MSTX_DFX_REPORT__
366+ MstxTensor::GetMstxVecUnaryInfo<T, T, true>(dst, src, "Ln", count);
321#endif367#endif
322 LnImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(), count);368 LnImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(), count);
323}369}
@@ -347,6 +393,9 @@ __aicore__ inline void Abs(const LocalTensor<T>& dst, const LocalTensor<T>& src,
347 "Abs")) {393 "Abs")) {
348 ASCENDC_REPORT_CHECK_ERROR("Abs", KernelFuncType::MASK_BIT_MODE);394 ASCENDC_REPORT_CHECK_ERROR("Abs", KernelFuncType::MASK_BIT_MODE);
349 }395 }
396+#endif
397+#ifdef __MSTX_DFX_REPORT__
398+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask[0], mask[1], repeatTime, repeatParams, "Abs");
350#endif399#endif
351 AbsImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),400 AbsImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),
352 mask, repeatTime, repeatParams);401 mask, repeatTime, repeatParams);
@@ -362,6 +411,9 @@ __aicore__ inline void Abs(const LocalTensor<T>& dst, const LocalTensor<T>& src,
362 "Abs")) {411 "Abs")) {
363 ASCENDC_REPORT_CHECK_ERROR("Abs", KernelFuncType::MASK_COUNT_MODE);412 ASCENDC_REPORT_CHECK_ERROR("Abs", KernelFuncType::MASK_COUNT_MODE);
364 }413 }
414+#endif
415+#ifdef __MSTX_DFX_REPORT__
416+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask, repeatTime, repeatParams, "Abs");
365#endif417#endif
366 AbsImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),418 AbsImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),
367 mask, repeatTime, repeatParams);419 mask, repeatTime, repeatParams);
@@ -382,6 +434,9 @@ __aicore__ inline void Abs(const LocalTensor<T>& dst, const LocalTensor<T>& src,
382 if (!CheckFunVecBinaryScalar(dst, src, static_cast<PrimType>(0), count, "Abs")) {434 if (!CheckFunVecBinaryScalar(dst, src, static_cast<PrimType>(0), count, "Abs")) {
383 ASCENDC_REPORT_CHECK_ERROR("Abs", KernelFuncType::CALCOUNT_MODE);435 ASCENDC_REPORT_CHECK_ERROR("Abs", KernelFuncType::CALCOUNT_MODE);
384 }436 }
437+#endif
438+#ifdef __MSTX_DFX_REPORT__
439+ MstxTensor::GetMstxVecUnaryInfo<T, T, true>(dst, src, "Abs", count);
385#endif440#endif
386 AbsImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(), count);441 AbsImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(), count);
387}442}
@@ -401,6 +456,9 @@ __aicore__ inline void Abs(const LocalTensor<T>& dst, const LocalTensor<U>& src,
401 if (!CheckFunVecBinaryScalarDiffType(dst, src, static_cast<PrimT<U>>(0), count, "vabs")) {456 if (!CheckFunVecBinaryScalarDiffType(dst, src, static_cast<PrimT<U>>(0), count, "vabs")) {
402 ASSERT(false && "check vabs instr failed");457 ASSERT(false && "check vabs instr failed");
403 }458 }
459+#endif
460+#ifdef __MSTX_DFX_REPORT__
461+ MstxTensor::GetMstxVecUnaryInfo<T, T, true>(dst, src, "Abs", count);
404#endif462#endif
405 AbsImpl((__ubuf__ PrimT<T>*)dst.GetPhyAddr(), (__ubuf__ PrimT<U>*)src.GetPhyAddr(), count);463 AbsImpl((__ubuf__ PrimT<T>*)dst.GetPhyAddr(), (__ubuf__ PrimT<U>*)src.GetPhyAddr(), count);
406}464}
@@ -431,6 +489,9 @@ __aicore__ inline void Reciprocal(const LocalTensor<T>& dst, const LocalTensor<T
431 "Reciprocal")) {489 "Reciprocal")) {
432 ASCENDC_REPORT_CHECK_ERROR("Reciprocal", KernelFuncType::MASK_BIT_MODE);490 ASCENDC_REPORT_CHECK_ERROR("Reciprocal", KernelFuncType::MASK_BIT_MODE);
433 }491 }
492+#endif
493+#ifdef __MSTX_DFX_REPORT__
494+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask[0], mask[1], repeatTime, repeatParams, "Reciprocal");
434#endif495#endif
435 ReciprocalImpl<PrimType, isSetMask, config>((__ubuf__ PrimType*)dst.GetPhyAddr(),496 ReciprocalImpl<PrimType, isSetMask, config>((__ubuf__ PrimType*)dst.GetPhyAddr(),
436 (__ubuf__ PrimType*)src.GetPhyAddr(), mask, repeatTime, repeatParams);497 (__ubuf__ PrimType*)src.GetPhyAddr(), mask, repeatTime, repeatParams);
@@ -446,6 +507,9 @@ __aicore__ inline void Reciprocal(const LocalTensor<T>& dst, const LocalTensor<T
446 "Reciprocal")) {507 "Reciprocal")) {
447 ASCENDC_REPORT_CHECK_ERROR("Reciprocal", KernelFuncType::MASK_COUNT_MODE);508 ASCENDC_REPORT_CHECK_ERROR("Reciprocal", KernelFuncType::MASK_COUNT_MODE);
448 }509 }
510+#endif
511+#ifdef __MSTX_DFX_REPORT__
512+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask, repeatTime, repeatParams, "Reciprocal");
449#endif513#endif
450 ReciprocalImpl<PrimType, isSetMask, config>((__ubuf__ PrimType*)dst.GetPhyAddr(),514 ReciprocalImpl<PrimType, isSetMask, config>((__ubuf__ PrimType*)dst.GetPhyAddr(),
451 (__ubuf__ PrimType*)src.GetPhyAddr(), mask, repeatTime, repeatParams);515 (__ubuf__ PrimType*)src.GetPhyAddr(), mask, repeatTime, repeatParams);
@@ -462,6 +526,9 @@ __aicore__ inline void Reciprocal(const LocalTensor<T>& dst, const LocalTensor<T
462 "Reciprocal")) {526 "Reciprocal")) {
463 ASCENDC_REPORT_CHECK_ERROR("Reciprocal", KernelFuncType::MASK_BIT_MODE);527 ASCENDC_REPORT_CHECK_ERROR("Reciprocal", KernelFuncType::MASK_BIT_MODE);
464 }528 }
529+#endif
530+#ifdef __MSTX_DFX_REPORT__
531+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask[0], mask[1], repeatTime, repeatParams, "Reciprocal");
465#endif532#endif
466 ReciprocalImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(),533 ReciprocalImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(),
467 (__ubuf__ PrimType*)src.GetPhyAddr(), mask, repeatTime, repeatParams);534 (__ubuf__ PrimType*)src.GetPhyAddr(), mask, repeatTime, repeatParams);
@@ -477,6 +544,9 @@ __aicore__ inline void Reciprocal(const LocalTensor<T>& dst, const LocalTensor<T
477 "Reciprocal")) {544 "Reciprocal")) {
478 ASCENDC_REPORT_CHECK_ERROR("Reciprocal", KernelFuncType::MASK_COUNT_MODE);545 ASCENDC_REPORT_CHECK_ERROR("Reciprocal", KernelFuncType::MASK_COUNT_MODE);
479 }546 }
547+#endif
548+#ifdef __MSTX_DFX_REPORT__
549+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask, repeatTime, repeatParams, "Reciprocal");
480#endif550#endif
481 ReciprocalImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(),551 ReciprocalImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(),
482 (__ubuf__ PrimType*)src.GetPhyAddr(), mask, repeatTime, repeatParams);552 (__ubuf__ PrimType*)src.GetPhyAddr(), mask, repeatTime, repeatParams);
@@ -500,6 +570,9 @@ __aicore__ inline void Reciprocal(const LocalTensor<T>& dst, const LocalTensor<T
500 if (!CheckFunVecBinaryScalar(dst, src, static_cast<PrimType>(0), count, "Reciprocal")) {570 if (!CheckFunVecBinaryScalar(dst, src, static_cast<PrimType>(0), count, "Reciprocal")) {
501 ASCENDC_REPORT_CHECK_ERROR("Reciprocal", KernelFuncType::CALCOUNT_MODE);571 ASCENDC_REPORT_CHECK_ERROR("Reciprocal", KernelFuncType::CALCOUNT_MODE);
502 }572 }
573+#endif
574+#ifdef __MSTX_DFX_REPORT__
575+ MstxTensor::GetMstxVecUnaryInfo<T, T, true>(dst, src, "Reciprocal", count);
503#endif576#endif
504 ReciprocalImpl<PrimType, config>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(), count);577 ReciprocalImpl<PrimType, config>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(), count);
505}578}
@@ -513,6 +586,9 @@ __aicore__ inline void Reciprocal(const LocalTensor<T>& dst, const LocalTensor<T
513 if (!CheckFunVecBinaryScalar(dst, src, static_cast<PrimType>(0), count, "Reciprocal")) {586 if (!CheckFunVecBinaryScalar(dst, src, static_cast<PrimType>(0), count, "Reciprocal")) {
514 ASCENDC_REPORT_CHECK_ERROR("Reciprocal", KernelFuncType::CALCOUNT_MODE);587 ASCENDC_REPORT_CHECK_ERROR("Reciprocal", KernelFuncType::CALCOUNT_MODE);
515 }588 }
589+#endif
590+#ifdef __MSTX_DFX_REPORT__
591+ MstxTensor::GetMstxVecUnaryInfo<T, T, true>(dst, src, "Reciprocal", count);
516#endif592#endif
517 ReciprocalImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(), count);593 ReciprocalImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(), count);
518}594}
@@ -543,6 +619,9 @@ __aicore__ inline void Rsqrt(const LocalTensor<T>& dst, const LocalTensor<T>& sr
543 "Rsqrt")) {619 "Rsqrt")) {
544 ASCENDC_REPORT_CHECK_ERROR("Rsqrt", KernelFuncType::MASK_BIT_MODE);620 ASCENDC_REPORT_CHECK_ERROR("Rsqrt", KernelFuncType::MASK_BIT_MODE);
545 }621 }
622+#endif
623+#ifdef __MSTX_DFX_REPORT__
624+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask[0], mask[1], repeatTime, repeatParams, "Rsqrt");
546#endif625#endif
547 RsqrtImpl<PrimType, isSetMask, config>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),626 RsqrtImpl<PrimType, isSetMask, config>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),
548 mask, repeatTime, repeatParams);627 mask, repeatTime, repeatParams);
@@ -558,6 +637,9 @@ __aicore__ inline void Rsqrt(const LocalTensor<T>& dst, const LocalTensor<T>& sr
558 "Rsqrt")) {637 "Rsqrt")) {
559 ASCENDC_REPORT_CHECK_ERROR("Rsqrt", KernelFuncType::MASK_COUNT_MODE);638 ASCENDC_REPORT_CHECK_ERROR("Rsqrt", KernelFuncType::MASK_COUNT_MODE);
560 }639 }
640+#endif
641+#ifdef __MSTX_DFX_REPORT__
642+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask, repeatTime, repeatParams, "Rsqrt");
561#endif643#endif
562 RsqrtImpl<PrimType, isSetMask, config>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),644 RsqrtImpl<PrimType, isSetMask, config>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),
563 mask, repeatTime, repeatParams);645 mask, repeatTime, repeatParams);
@@ -574,6 +656,9 @@ __aicore__ inline void Rsqrt(const LocalTensor<T>& dst, const LocalTensor<T>& sr
574 "Rsqrt")) {656 "Rsqrt")) {
575 ASCENDC_REPORT_CHECK_ERROR("Rsqrt", KernelFuncType::MASK_BIT_MODE);657 ASCENDC_REPORT_CHECK_ERROR("Rsqrt", KernelFuncType::MASK_BIT_MODE);
576 }658 }
659+#endif
660+#ifdef __MSTX_DFX_REPORT__
661+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask[0], mask[1], repeatTime, repeatParams, "Rsqrt");
577#endif662#endif
578 RsqrtImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),663 RsqrtImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),
579 mask, repeatTime, repeatParams);664 mask, repeatTime, repeatParams);
@@ -589,6 +674,9 @@ __aicore__ inline void Rsqrt(const LocalTensor<T>& dst, const LocalTensor<T>& sr
589 "Rsqrt")) {674 "Rsqrt")) {
590 ASCENDC_REPORT_CHECK_ERROR("Rsqrt", KernelFuncType::MASK_COUNT_MODE);675 ASCENDC_REPORT_CHECK_ERROR("Rsqrt", KernelFuncType::MASK_COUNT_MODE);
591 }676 }
677+#endif
678+#ifdef __MSTX_DFX_REPORT__
679+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask, repeatTime, repeatParams, "Rsqrt");
592#endif680#endif
593 RsqrtImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),681 RsqrtImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),
594 mask, repeatTime, repeatParams);682 mask, repeatTime, repeatParams);
@@ -611,6 +699,9 @@ __aicore__ inline void Rsqrt(const LocalTensor<T>& dst, const LocalTensor<T>& sr
611 if (!CheckFunVecBinaryScalar(dst, src, static_cast<PrimType>(0), count, "Rsqrt")) {699 if (!CheckFunVecBinaryScalar(dst, src, static_cast<PrimType>(0), count, "Rsqrt")) {
612 ASCENDC_REPORT_CHECK_ERROR("Rsqrt", KernelFuncType::CALCOUNT_MODE);700 ASCENDC_REPORT_CHECK_ERROR("Rsqrt", KernelFuncType::CALCOUNT_MODE);
613 }701 }
702+#endif
703+#ifdef __MSTX_DFX_REPORT__
704+ MstxTensor::GetMstxVecUnaryInfo<T, T, true>(dst, src, "Rsqrt", count);
614#endif705#endif
615 RsqrtImpl<PrimType, config>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(), count);706 RsqrtImpl<PrimType, config>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(), count);
616}707}
@@ -623,6 +714,9 @@ __aicore__ inline void Rsqrt(const LocalTensor<T>& dst, const LocalTensor<T>& sr
623 if (!CheckFunVecBinaryScalar(dst, src, static_cast<PrimType>(0), count, "Rsqrt")) {714 if (!CheckFunVecBinaryScalar(dst, src, static_cast<PrimType>(0), count, "Rsqrt")) {
624 ASCENDC_REPORT_CHECK_ERROR("Rsqrt", KernelFuncType::CALCOUNT_MODE);715 ASCENDC_REPORT_CHECK_ERROR("Rsqrt", KernelFuncType::CALCOUNT_MODE);
625 }716 }
717+#endif
718+#ifdef __MSTX_DFX_REPORT__
719+ MstxTensor::GetMstxVecUnaryInfo<T, T, true>(dst, src, "Rsqrt", count);
626#endif720#endif
627 RsqrtImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(), count);721 RsqrtImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(), count);
628}722}
@@ -653,6 +747,9 @@ __aicore__ inline void Sqrt(const LocalTensor<T>& dst, const LocalTensor<T>& src
653 "Sqrt")) {747 "Sqrt")) {
654 ASCENDC_REPORT_CHECK_ERROR("Sqrt", KernelFuncType::MASK_BIT_MODE);748 ASCENDC_REPORT_CHECK_ERROR("Sqrt", KernelFuncType::MASK_BIT_MODE);
655 }749 }
750+#endif
751+#ifdef __MSTX_DFX_REPORT__
752+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask[0], mask[1], repeatTime, repeatParams, "Sqrt");
656#endif753#endif
657 SqrtImpl<PrimType, isSetMask, config>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),754 SqrtImpl<PrimType, isSetMask, config>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),
658 mask, repeatTime, repeatParams);755 mask, repeatTime, repeatParams);
@@ -668,6 +765,9 @@ __aicore__ inline void Sqrt(const LocalTensor<T>& dst, const LocalTensor<T>& src
668 "Sqrt")) {765 "Sqrt")) {
669 ASCENDC_REPORT_CHECK_ERROR("Sqrt", KernelFuncType::MASK_COUNT_MODE);766 ASCENDC_REPORT_CHECK_ERROR("Sqrt", KernelFuncType::MASK_COUNT_MODE);
670 }767 }
768+#endif
769+#ifdef __MSTX_DFX_REPORT__
770+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask, repeatTime, repeatParams, "Sqrt");
671#endif771#endif
672 SqrtImpl<PrimType, isSetMask, config>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),772 SqrtImpl<PrimType, isSetMask, config>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),
673 mask, repeatTime, repeatParams);773 mask, repeatTime, repeatParams);
@@ -684,6 +784,9 @@ __aicore__ inline void Sqrt(const LocalTensor<T>& dst, const LocalTensor<T>& src
684 "Sqrt")) {784 "Sqrt")) {
685 ASCENDC_REPORT_CHECK_ERROR("Sqrt", KernelFuncType::MASK_BIT_MODE);785 ASCENDC_REPORT_CHECK_ERROR("Sqrt", KernelFuncType::MASK_BIT_MODE);
686 }786 }
787+#endif
788+#ifdef __MSTX_DFX_REPORT__
789+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask[0], mask[1], repeatTime, repeatParams, "Sqrt");
687#endif790#endif
688 SqrtImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),791 SqrtImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),
689 mask, repeatTime, repeatParams);792 mask, repeatTime, repeatParams);
@@ -699,6 +802,9 @@ __aicore__ inline void Sqrt(const LocalTensor<T>& dst, const LocalTensor<T>& src
699 "Sqrt")) {802 "Sqrt")) {
700 ASCENDC_REPORT_CHECK_ERROR("Sqrt", KernelFuncType::MASK_COUNT_MODE);803 ASCENDC_REPORT_CHECK_ERROR("Sqrt", KernelFuncType::MASK_COUNT_MODE);
701 }804 }
805+#endif
806+#ifdef __MSTX_DFX_REPORT__
807+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask, repeatTime, repeatParams, "Sqrt");
702#endif808#endif
703 SqrtImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),809 SqrtImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),
704 mask, repeatTime, repeatParams);810 mask, repeatTime, repeatParams);
@@ -721,6 +827,9 @@ __aicore__ inline void Sqrt(const LocalTensor<T>& dst, const LocalTensor<T>& src
721 if (!CheckFunVecBinaryScalar(dst, src, static_cast<PrimType>(0), count, "Sqrt")) {827 if (!CheckFunVecBinaryScalar(dst, src, static_cast<PrimType>(0), count, "Sqrt")) {
722 ASCENDC_REPORT_CHECK_ERROR("Sqrt", KernelFuncType::CALCOUNT_MODE);828 ASCENDC_REPORT_CHECK_ERROR("Sqrt", KernelFuncType::CALCOUNT_MODE);
723 }829 }
830+#endif
831+#ifdef __MSTX_DFX_REPORT__
832+ MstxTensor::GetMstxVecUnaryInfo<T, T, true>(dst, src, "Sqrt", count);
724#endif833#endif
725 SqrtImpl<PrimType, config>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(), count);834 SqrtImpl<PrimType, config>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(), count);
726}835}
@@ -733,6 +842,9 @@ __aicore__ inline void Sqrt(const LocalTensor<T>& dst, const LocalTensor<T>& src
733 if (!CheckFunVecBinaryScalar(dst, src, static_cast<PrimType>(0), count, "Sqrt")) {842 if (!CheckFunVecBinaryScalar(dst, src, static_cast<PrimType>(0), count, "Sqrt")) {
734 ASCENDC_REPORT_CHECK_ERROR("Sqrt", KernelFuncType::CALCOUNT_MODE);843 ASCENDC_REPORT_CHECK_ERROR("Sqrt", KernelFuncType::CALCOUNT_MODE);
735 }844 }
845+#endif
846+#ifdef __MSTX_DFX_REPORT__
847+ MstxTensor::GetMstxVecUnaryInfo<T, T, true>(dst, src, "Sqrt", count);
736#endif848#endif
737 SqrtImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(), count);849 SqrtImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(), count);
738}850}
@@ -762,6 +874,9 @@ __aicore__ inline void Not(const LocalTensor<T>& dst, const LocalTensor<T>& src,
762 "Not")) {874 "Not")) {
763 ASCENDC_REPORT_CHECK_ERROR("Not", KernelFuncType::MASK_BIT_MODE);875 ASCENDC_REPORT_CHECK_ERROR("Not", KernelFuncType::MASK_BIT_MODE);
764 }876 }
877+#endif
878+#ifdef __MSTX_DFX_REPORT__
879+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask[0], mask[1], repeatTime, repeatParams, "Not");
765#endif880#endif
766 NotImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),881 NotImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),
767 mask, repeatTime, repeatParams);882 mask, repeatTime, repeatParams);
@@ -777,6 +892,9 @@ __aicore__ inline void Not(const LocalTensor<T>& dst, const LocalTensor<T>& src,
777 "Not")) {892 "Not")) {
778 ASCENDC_REPORT_CHECK_ERROR("Not", KernelFuncType::MASK_COUNT_MODE);893 ASCENDC_REPORT_CHECK_ERROR("Not", KernelFuncType::MASK_COUNT_MODE);
779 }894 }
895+#endif
896+#ifdef __MSTX_DFX_REPORT__
897+ MstxTensor::GetMstxVecUnaryInfo<T, T, isSetMask>(dst, src, mask, repeatTime, repeatParams, "Not");
780#endif898#endif
781 NotImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),899 NotImpl<PrimType, isSetMask>((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(),
782 mask, repeatTime, repeatParams);900 mask, repeatTime, repeatParams);
@@ -797,6 +915,9 @@ __aicore__ inline void Not(const LocalTensor<T>& dst, const LocalTensor<T>& src,
797 if (!CheckFunVecBinaryScalar(dst, src, static_cast<PrimType>(0), count, "Not")) {915 if (!CheckFunVecBinaryScalar(dst, src, static_cast<PrimType>(0), count, "Not")) {
798 ASCENDC_REPORT_CHECK_ERROR("Not", KernelFuncType::CALCOUNT_MODE);916 ASCENDC_REPORT_CHECK_ERROR("Not", KernelFuncType::CALCOUNT_MODE);
799 }917 }
918+#endif
919+#ifdef __MSTX_DFX_REPORT__
920+ MstxTensor::GetMstxVecUnaryInfo<T, T, true>(dst, src, "Not", count);
800#endif921#endif
801 NotImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(), count);922 NotImpl((__ubuf__ PrimType*)dst.GetPhyAddr(), (__ubuf__ PrimType*)src.GetPhyAddr(), count);
802}923}
Mimpl/basic_api/kernel_operator_vec_vconv_intf_impl.h+37-0
@@ -19,6 +19,7 @@
19#include "kernel_struct_binary.h"19#include "kernel_struct_binary.h"
20#include "kernel_struct_unary.h"20#include "kernel_struct_unary.h"
21#include "kernel_struct_vdeq.h"21#include "kernel_struct_vdeq.h"
22+#include "mstx_local_tensor_info.h"
22 23 
23#if __NPU_ARCH__ == 100124#if __NPU_ARCH__ == 1001
24#include "dav_c100/kernel_operator_vec_vconv_impl.h"25#include "dav_c100/kernel_operator_vec_vconv_impl.h"
@@ -64,6 +65,9 @@ __aicore__ inline void Cast(const LocalTensor<T>& dst, const LocalTensor<U>& src
64 const RoundMode& roundMode, const uint64_t mask[], const uint8_t repeatTime,65 const RoundMode& roundMode, const uint64_t mask[], const uint8_t repeatTime,
65 const UnaryRepeatParams& repeatParams)66 const UnaryRepeatParams& repeatParams)
66{67{
68+#ifdef __MSTX_DFX_REPORT__
69+ MstxTensor::GetMstxVecUnaryCastInfo<T, U, isSetMask>(dst, src, mask[0], mask[1], repeatTime, repeatParams, "Cast");
70+#endif
67 using DstPrimType = PrimT<T>;71 using DstPrimType = PrimT<T>;
68 using SrcPrimType = PrimT<U>;72 using SrcPrimType = PrimT<U>;
69#if ASCENDC_CPU_DEBUG73#if ASCENDC_CPU_DEBUG
@@ -90,6 +94,9 @@ template <typename T, typename U, bool isSetMask>
90__aicore__ inline void Cast(const LocalTensor<T>& dst, const LocalTensor<U>& src,94__aicore__ inline void Cast(const LocalTensor<T>& dst, const LocalTensor<U>& src,
91 const RoundMode& roundMode, const uint64_t mask, const uint8_t repeatTime, const UnaryRepeatParams& repeatParams)95 const RoundMode& roundMode, const uint64_t mask, const uint8_t repeatTime, const UnaryRepeatParams& repeatParams)
92{96{
97+#ifdef __MSTX_DFX_REPORT__
98+ MstxTensor::GetMstxVecUnaryCastInfo<T, U, isSetMask>(dst, src, mask, repeatTime, repeatParams, "Cast");
99+#endif
93 using DstPrimType = PrimT<T>;100 using DstPrimType = PrimT<T>;
94 using SrcPrimType = PrimT<U>;101 using SrcPrimType = PrimT<U>;
95#if ASCENDC_CPU_DEBUG102#if ASCENDC_CPU_DEBUG
@@ -123,6 +130,9 @@ template <typename T, typename U>
123__aicore__ inline void Cast(const LocalTensor<T>& dst, const LocalTensor<U>& src,130__aicore__ inline void Cast(const LocalTensor<T>& dst, const LocalTensor<U>& src,
124 const RoundMode& roundMode, const uint32_t count)131 const RoundMode& roundMode, const uint32_t count)
125{132{
133+#ifdef __MSTX_DFX_REPORT__
134+ MstxTensor::GetMstxVecUnaryCastInfo<T, U, true>(dst, src, "Cast", count);
135+#endif
126 using DstPrimType = PrimT<T>;136 using DstPrimType = PrimT<T>;
127 using SrcPrimType = PrimT<U>;137 using SrcPrimType = PrimT<U>;
128#if ASCENDC_CPU_DEBUG138#if ASCENDC_CPU_DEBUG
@@ -196,6 +206,9 @@ template <typename T, typename U, bool isSetMask, bool isVecDeq, bool halfBlock>
196__aicore__ inline void CastDeq(const LocalTensor<T>& dst, const LocalTensor<U>& src,206__aicore__ inline void CastDeq(const LocalTensor<T>& dst, const LocalTensor<U>& src,
197 const uint64_t mask[], uint8_t repeatTime, const UnaryRepeatParams& repeatParams)207 const uint64_t mask[], uint8_t repeatTime, const UnaryRepeatParams& repeatParams)
198{208{
209+#ifdef __MSTX_DFX_REPORT__
210+ MstxTensor::GetMstxVecUnaryCastDeqInfo<T, U, isSetMask>(dst, src, mask[0], mask[1], repeatTime, repeatParams, "CastDeq", halfBlock);
211+#endif
199 using DstPrimType = PrimT<T>;212 using DstPrimType = PrimT<T>;
200 using SrcPrimType = PrimT<U>;213 using SrcPrimType = PrimT<U>;
201#if ASCENDC_CPU_DEBUG214#if ASCENDC_CPU_DEBUG
@@ -214,6 +227,9 @@ template <typename T, typename U, bool isSetMask, bool isVecDeq, bool halfBlock>
214__aicore__ inline void CastDeq(const LocalTensor<T>& dst, const LocalTensor<U>& src,227__aicore__ inline void CastDeq(const LocalTensor<T>& dst, const LocalTensor<U>& src,
215 const int32_t mask, uint8_t repeatTime, const UnaryRepeatParams& repeatParams)228 const int32_t mask, uint8_t repeatTime, const UnaryRepeatParams& repeatParams)
216{229{
230+#ifdef __MSTX_DFX_REPORT__
231+ MstxTensor::GetMstxVecUnaryCastDeqInfo<T, U, isSetMask>(dst, src, mask, repeatTime, repeatParams, "CastDeq", halfBlock);
232+#endif
217 using DstPrimType = PrimT<T>;233 using DstPrimType = PrimT<T>;
218 using SrcPrimType = PrimT<U>;234 using SrcPrimType = PrimT<U>;
219#if ASCENDC_CPU_DEBUG235#if ASCENDC_CPU_DEBUG
@@ -259,6 +275,9 @@ template <typename T, typename U, bool isVecDeq, bool halfBlock>
259__aicore__ inline void CastDeq(const LocalTensor<T>& dst, const LocalTensor<U>& src,275__aicore__ inline void CastDeq(const LocalTensor<T>& dst, const LocalTensor<U>& src,
260 const uint32_t count)276 const uint32_t count)
261{277{
278+#ifdef __MSTX_DFX_REPORT__
279+ MstxTensor::GetMstxVecUnaryCastDeqInfo<T, U, true>(dst, src, "CastDeq", count, halfBlock);
280+#endif
262 using DstPrimType = PrimT<T>;281 using DstPrimType = PrimT<T>;
263 using SrcPrimType = PrimT<U>;282 using SrcPrimType = PrimT<U>;
264#if ASCENDC_CPU_DEBUG283#if ASCENDC_CPU_DEBUG
@@ -304,6 +323,9 @@ __aicore__ inline void AddReluCast(const LocalTensor<T>& dst, const LocalTensor<
304 if (!CheckFuncVecBinaryCmp(dst, src0, src1, mask, repeatTime, repeatParams, "AddReluCast")) {323 if (!CheckFuncVecBinaryCmp(dst, src0, src1, mask, repeatTime, repeatParams, "AddReluCast")) {
305 ASCENDC_REPORT_CHECK_ERROR("AddReluCast", KernelFuncType::MASK_COUNT_MODE);324 ASCENDC_REPORT_CHECK_ERROR("AddReluCast", KernelFuncType::MASK_COUNT_MODE);
306 }325 }
326+#endif
327+#ifdef __MSTX_DFX_REPORT__
328+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, mask, repeatTime, repeatParams, isSetMask, "AddReluCast");
307#endif329#endif
308 AddReluCastImpl<PrimT<T>, PrimT<U>, isSetMask>((__ubuf__ PrimT<T>*)dst.GetPhyAddr(), (__ubuf__ PrimT<U>*)src0.GetPhyAddr(),330 AddReluCastImpl<PrimT<T>, PrimT<U>, isSetMask>((__ubuf__ PrimT<T>*)dst.GetPhyAddr(), (__ubuf__ PrimT<U>*)src0.GetPhyAddr(),
309 (__ubuf__ PrimT<U>*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);331 (__ubuf__ PrimT<U>*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);
@@ -325,6 +347,9 @@ __aicore__ inline void AddReluCast(const LocalTensor<T>& dst, const LocalTensor<
325 if (!CheckFuncVecBinaryCmp(dst, src0, src1, mask, repeatTime, repeatParams, "AddReluCast")) {347 if (!CheckFuncVecBinaryCmp(dst, src0, src1, mask, repeatTime, repeatParams, "AddReluCast")) {
326 ASCENDC_REPORT_CHECK_ERROR("AddReluCast", KernelFuncType::MASK_BIT_MODE);348 ASCENDC_REPORT_CHECK_ERROR("AddReluCast", KernelFuncType::MASK_BIT_MODE);
327 }349 }
350+#endif
351+#ifdef __MSTX_DFX_REPORT__
352+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, mask[0], mask[1], repeatTime, repeatParams, isSetMask, "AddReluCast");
328#endif353#endif
329 AddReluCastImpl<PrimT<T>, PrimT<U>, isSetMask>((__ubuf__ PrimT<T>*)dst.GetPhyAddr(), (__ubuf__ PrimT<U>*)src0.GetPhyAddr(),354 AddReluCastImpl<PrimT<T>, PrimT<U>, isSetMask>((__ubuf__ PrimT<T>*)dst.GetPhyAddr(), (__ubuf__ PrimT<U>*)src0.GetPhyAddr(),
330 (__ubuf__ PrimT<U>*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);355 (__ubuf__ PrimT<U>*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);
@@ -351,6 +376,9 @@ __aicore__ inline void AddReluCast(const LocalTensor<T>& dst, const LocalTensor<
351 if (!CheckFuncVecBinaryCmp(dst, src0, src1, count, "AddReluCast")) {376 if (!CheckFuncVecBinaryCmp(dst, src0, src1, count, "AddReluCast")) {
352 ASCENDC_REPORT_CHECK_ERROR("AddReluCast", KernelFuncType::CALCOUNT_MODE);377 ASCENDC_REPORT_CHECK_ERROR("AddReluCast", KernelFuncType::CALCOUNT_MODE);
353 }378 }
379+#endif
380+#ifdef __MSTX_DFX_REPORT__
381+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, "AddReluCast", count);
354#endif382#endif
355 AddReluCastImpl((__ubuf__ PrimT<T>*)dst.GetPhyAddr(), (__ubuf__ PrimT<U>*)src0.GetPhyAddr(),383 AddReluCastImpl((__ubuf__ PrimT<T>*)dst.GetPhyAddr(), (__ubuf__ PrimT<U>*)src0.GetPhyAddr(),
356 (__ubuf__ PrimT<U>*)src1.GetPhyAddr(), count);384 (__ubuf__ PrimT<U>*)src1.GetPhyAddr(), count);
@@ -391,6 +419,9 @@ __aicore__ inline void SubReluCast(const LocalTensor<T>& dst, const LocalTensor<
391 if (!CheckFuncVecBinaryCmp(dst, src0, src1, mask, repeatTime, repeatParams, "SubReluCast")) {419 if (!CheckFuncVecBinaryCmp(dst, src0, src1, mask, repeatTime, repeatParams, "SubReluCast")) {
392 ASCENDC_REPORT_CHECK_ERROR("SubReluCast", KernelFuncType::MASK_COUNT_MODE);420 ASCENDC_REPORT_CHECK_ERROR("SubReluCast", KernelFuncType::MASK_COUNT_MODE);
393 }421 }
422+#endif
423+#ifdef __MSTX_DFX_REPORT__
424+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, mask, repeatTime, repeatParams, isSetMask, "SubReluCast");
394#endif425#endif
395 SubReluCastImpl<DstPrimType, SrcPrimType, isSetMask>((__ubuf__ DstPrimType*)dst.GetPhyAddr(), (__ubuf__ SrcPrimType*)src0.GetPhyAddr(),426 SubReluCastImpl<DstPrimType, SrcPrimType, isSetMask>((__ubuf__ DstPrimType*)dst.GetPhyAddr(), (__ubuf__ SrcPrimType*)src0.GetPhyAddr(),
396 (__ubuf__ SrcPrimType*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);427 (__ubuf__ SrcPrimType*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);
@@ -414,6 +445,9 @@ __aicore__ inline void SubReluCast(const LocalTensor<T>& dst, const LocalTensor<
414 if (!CheckFuncVecBinaryCmp(dst, src0, src1, mask, repeatTime, repeatParams, "SubReluCast")) {445 if (!CheckFuncVecBinaryCmp(dst, src0, src1, mask, repeatTime, repeatParams, "SubReluCast")) {
415 ASCENDC_REPORT_CHECK_ERROR("SubReluCast", KernelFuncType::MASK_BIT_MODE);446 ASCENDC_REPORT_CHECK_ERROR("SubReluCast", KernelFuncType::MASK_BIT_MODE);
416 }447 }
448+#endif
449+#ifdef __MSTX_DFX_REPORT__
450+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, mask[0], mask[1], repeatTime, repeatParams, isSetMask, "SubReluCast");
417#endif451#endif
418 SubReluCastImpl<DstPrimType, SrcPrimType, isSetMask>((__ubuf__ DstPrimType*)dst.GetPhyAddr(), (__ubuf__ SrcPrimType*)src0.GetPhyAddr(),452 SubReluCastImpl<DstPrimType, SrcPrimType, isSetMask>((__ubuf__ DstPrimType*)dst.GetPhyAddr(), (__ubuf__ SrcPrimType*)src0.GetPhyAddr(),
419 (__ubuf__ SrcPrimType*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);453 (__ubuf__ SrcPrimType*)src1.GetPhyAddr(), mask, repeatTime, repeatParams);
@@ -442,6 +476,9 @@ __aicore__ inline void SubReluCast(const LocalTensor<T>& dst, const LocalTensor<
442 if (!CheckFuncVecBinaryCmp(dst, src0, src1, count, "SubReluCast")) {476 if (!CheckFuncVecBinaryCmp(dst, src0, src1, count, "SubReluCast")) {
443 ASCENDC_REPORT_CHECK_ERROR("SubReluCast", KernelFuncType::CALCOUNT_MODE);477 ASCENDC_REPORT_CHECK_ERROR("SubReluCast", KernelFuncType::CALCOUNT_MODE);
444 }478 }
479+#endif
480+#ifdef __MSTX_DFX_REPORT__
481+ MstxTensor::GetMstxVecBinaryInfo(dst, src0, src1, "SubReluCast", count);
445#endif482#endif
446 SubReluCastImpl((__ubuf__ DstPrimType*)dst.GetPhyAddr(), (__ubuf__ SrcPrimType*)src0.GetPhyAddr(),483 SubReluCastImpl((__ubuf__ DstPrimType*)dst.GetPhyAddr(), (__ubuf__ SrcPrimType*)src0.GetPhyAddr(),
447 (__ubuf__ SrcPrimType*)src1.GetPhyAddr(), count);484 (__ubuf__ SrcPrimType*)src1.GetPhyAddr(), count);
Aimpl/basic_api/mstx_local_tensor_info.h+1825-0
@@ -0,0 +1,1825 @@
1+/**
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 of
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.
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.
8+* See LICENSE in the root of the software repository for the full text of the License.
9+*/
10+ 
11+/*!
12+ * \file mstx_local_tensor_info.h
13+ * \brief
14+ */
15+#ifndef MSTX_TENSOR_INFO_H
16+#define MSTX_TENSOR_INFO_H
17+#include "kernel_tensor.h"
18+#include "kernel_struct_unary.h"
19+#include "kernel_struct_brcb.h"
20+#include "kernel_struct_gather.h"
21+ 
22+#ifdef __MSTX_DFX_REPORT__
23+ 
24+namespace AscendC {
25+namespace MstxTensor {
26+ 
27+enum MstxTensorAddressSpace {
28+ MSTX_TENSOR_AT_PRIVATE = 0,
29+ MSTX_TENSOR_AT_GM,
30+ MSTX_TENSOR_AT_L1,
31+ MSTX_TENSOR_AT_L0A,
32+ MSTX_TENSOR_AT_L0B,
33+ MSTX_TENSOR_AT_L0C,
34+ MSTX_TENSOR_AT_UB,
35+};
36+ 
37+struct MstxTensorDesc {
38+ MstxTensorAddressSpace space;
39+ uint64_t addr;
40+ uint64_t size;
41+ uint8_t dataBits;
42+};
43+ 
44+template <typename T>
45+__aicore__ inline MstxTensorDesc From(const LocalTensor<T>& dst) {
46+ MstxTensorAddressSpace space;
47+ Hardware dstHWPos = GetPhyType((TPosition)dst.GetPosition());
48+ if (dstHWPos == Hardware::GM) {
49+ space = MSTX_TENSOR_AT_GM;
50+ } else if (dstHWPos == Hardware::UB) {
51+ space = MSTX_TENSOR_AT_UB;
52+ } else if (dstHWPos == Hardware::L1) {
53+ space = MSTX_TENSOR_AT_L1;
54+ } else if (dstHWPos == Hardware::L0A) {
55+ space = MSTX_TENSOR_AT_L0A;
56+ } else if (dstHWPos == Hardware::L0B) {
57+ space = MSTX_TENSOR_AT_L0B;
58+ } else if (dstHWPos == Hardware::L0C) {
59+ space = MSTX_TENSOR_AT_L0C;
60+ } else {
61+ space = MSTX_TENSOR_AT_PRIVATE;
62+ }
63+ 
64+ uint8_t sizebit;
65+ if constexpr (IsSameType<PrimT<T>, int4b_t>::value) {
66+ sizebit = static_cast<uint8_t>(4 * sizeof(PrimT<T>));
67+ } else {
68+ sizebit = static_cast<uint8_t>(8 * sizeof(PrimT<T>));
69+ }
70+ 
71+ return MstxTensorDesc{
72+ space,
73+ reinterpret_cast<uint64_t>(dst.GetPhyAddr()),
74+ dst.GetSize(),
75+ sizebit
76+ };
77+}
78+ 
79+template <typename T>
80+__aicore__ inline MstxTensorDesc FromGm(const GlobalTensor<T>& dst) {
81+ uint8_t sizebit;
82+ if constexpr (IsSameType<PrimT<T>, int4b_t>::value) {
83+ sizebit = static_cast<uint8_t>(4 * sizeof(PrimT<T>));
84+ } else {
85+ sizebit = static_cast<uint8_t>(8 * sizeof(PrimT<T>));
86+ }
87+ return MstxTensorDesc{
88+ MSTX_TENSOR_AT_GM,
89+ reinterpret_cast<uint64_t>(dst.GetPhyAddr()),
90+ dst.GetSize(),
91+ sizebit
92+ };
93+}
94+ 
95+enum class MstxReportType : uint32_t {
96+ MSTX_VEC_UNARY = 3000,
97+ MSTX_VEC_BINARY = 3001,
98+ MSTX_VEC_GATHER = 3002,
99+ MSTX_VEC_BINARY_SCALAR,
100+ MSTX_VEC_BILINEAR_INTERPOLATION,
101+ MSTX_VEC_TENARY,
102+ MSTX_VEC_CAST,
103+ MSTX_VEC_CASTDEQ,
104+ MSTX_VEC_SET_DEQ_SCALE,
105+ MSTX_VEC_CMP,
106+ MSTX_VEC_CMPS,
107+ MSTX_VEC_SEL,
108+ MSTX_VEC_GATHER_MASK,
109+ MSTX_VEC_TRANSDATA,
110+ MSTX_VEC_TRANSPOSE,
111+ MSTX_VEC_WHOLE_REDUCE,
112+ MSTX_VEC_BLK_REDUCE,
113+ MSTX_VEC_REDUCE,
114+ MSTX_VEC_PAIR_REDUCE,
115+ MSTX_VEC_REPEAT_REDUCE,
116+ MSTX_VEC_DUP,
117+ MSTX_VEC_BROADCAST,
118+ MSTX_VEC_VCI,
119+ MSTX_VEC_COPY,
120+ MSTX_DATA_COPY = 4001,
121+ MSTX_DATA_COPY_PAD = 4002,
122+};
123+ 
124+enum MstxMaskMode: uint32_t {
125+ MSTX_MASK_NORM = 0,
126+ MSTX_MASK_COUNT,
127+ MSTX_MASK_FROM_REG = 0xff,
128+};
129+ 
130+struct MstxVectorMask {
131+ uint64_t mask0;
132+ uint64_t mask1;
133+};
134+ 
135+struct MstxVecWrapper {
136+ MstxMaskMode maskMode;
137+ MstxVectorMask mask;
138+ uint32_t reserveBufSize;
139+ bool useMask;
140+};
141+ 
142+__aicore__ inline MstxVecWrapper WrapperFrom(MstxMaskMode maskMode, uint64_t mask0, uint64_t mask1, bool isSetMask, uint32_t reserveBufSize) {
143+ return MstxVecWrapper{
144+ maskMode,
145+ {mask0, mask1},
146+ reserveBufSize,
147+ isSetMask
148+ };
149+}
150+ 
151+template <typename T>
152+__aicore__ inline MstxVecWrapper WrapperFrom(MstxMaskMode maskMode, uint64_t mask, bool isSetMask, uint32_t reserveBufSize) {
153+ int32_t typeLen = 0;
154+ constexpr int32_t halfTypeLen = 64; // 1 register -> 64 bits -> 64 elements
155+ constexpr int32_t lenCoeff = 2;
156+ uint64_t trueMask0, trueMask1;
157+ if constexpr (IsSameType<T, int4b_t>::value) {
158+ typeLen = DEFAULT_BLOCK_SIZE * INT4_TWO;
159+ } else {
160+ typeLen = DEFAULT_BLOCK_SIZE / sizeof(T);
161+ }
162+ if (mask == halfTypeLen) {
163+ trueMask0 = FULL_MASK;
164+ trueMask1 = 0;
165+ } else if (mask == typeLen || mask >= halfTypeLen * lenCoeff) {
166+ trueMask0 = FULL_MASK;
167+ trueMask1 = FULL_MASK;
168+ } else {
169+ trueMask0 = (mask > halfTypeLen) ? FULL_MASK : (((static_cast<uint64_t>(1)) << static_cast<uint32_t>(mask)) - 1);
170+ trueMask1 = (mask > halfTypeLen) ? (((static_cast<uint64_t>(1)) << static_cast<uint32_t>(mask - halfTypeLen)) - 1) : 0;
171+ }
172+ 
173+ return MstxVecWrapper{
174+ maskMode,
175+ {trueMask0, trueMask1},
176+ reserveBufSize,
177+ isSetMask
178+ };
179+}
180+ 
181+struct MstxVecUnaryDesc {
182+ MstxTensorDesc dst;
183+ MstxTensorDesc src;
184+ MstxVecWrapper wrapper;
185+ uint32_t blockNum;
186+ uint32_t dstBlockStride;
187+ uint32_t srcBlockStride;
188+ uint32_t repeatTimes;
189+ uint32_t dstRepeatStride;
190+ uint32_t srcRepeatStride;
191+ char name[64];
192+};
193+ 
194+struct MstxVecReduceDesc {
195+ MstxTensorDesc dst;
196+ MstxTensorDesc src;
197+ MstxVecWrapper wrapper;
198+ uint32_t srcBlockStride;
199+ uint32_t repeatTimes;
200+ uint32_t dstRepeatStride;
201+ uint32_t srcRepeatStride;
202+ char name[64];
203+};
204+ 
205+struct MstxVecBrcbDesc {
206+ MstxTensorDesc dst;
207+ MstxTensorDesc src;
208+ MstxVecWrapper wrapper;
209+ uint32_t dstBlockStride;
210+ uint32_t repeatTimes;
211+ uint32_t dstRepeatStride;
212+ char name[64];
213+};
214+ 
215+struct MstxVecComplexReduceDesc {
216+ MstxTensorDesc dst;
217+ MstxTensorDesc src;
218+ MstxTensorDesc tmp;
219+ MstxVecWrapper wrapper;
220+ uint32_t repeatTimes;
221+ uint32_t srcRepeatStride;
222+ char name[64];
223+};
224+ 
225+struct MstxVecDupDesc {
226+ MstxTensorDesc dst;
227+ MstxVecWrapper wrapper;
228+ uint32_t dstBlockStride;
229+ uint32_t repeatTimes;
230+ uint32_t dstRepeatStride;
231+ char name[64];
232+};
233+ 
234+struct MstxVecCopy {
235+ MstxTensorDesc dst;
236+ MstxTensorDesc src;
237+ MstxVecWrapper wrapper;
238+ uint32_t repeatTimes;
239+ uint32_t dstStride;
240+ uint32_t srcStride;
241+ uint32_t dstRepeatSize;
242+ uint32_t srcRepeatSize;
243+ char name[64];
244+};
245+ 
246+struct MstxVecCastDeqDesc {
247+ MstxTensorDesc dst;
248+ MstxTensorDesc src;
249+ MstxVecWrapper wrapper;
250+ uint32_t blockNum;
251+ uint32_t dstBlockStride;
252+ uint32_t srcBlockStride;
253+ uint32_t repeatTimes;
254+ uint32_t dstRepeatStride;
255+ uint32_t srcRepeatStride;
256+ bool halfBlock;
257+ char name[64];
258+};
259+ 
260+struct MstxDataCopyDesc {
261+ MstxTensorDesc dst;
262+ MstxTensorDesc src;
263+ uint32_t lenBurst;
264+ uint32_t nBurst;
265+ uint32_t srcGap;
266+ uint32_t dstGap;
267+ char name[64];
268+};
269+ 
270+struct MstxDataCopyPadDesc {
271+ MstxTensorDesc dst;
272+ MstxTensorDesc src;
273+ uint32_t lenBurst;
274+ uint32_t nBurst;
275+ uint32_t srcGap;
276+ uint32_t dstGap;
277+ uint32_t leftPad;
278+ uint32_t rightPad;
279+ char name[64];
280+};
281+ 
282+struct MstxVecBinaryDesc {
283+ MstxTensorDesc dst;
284+ MstxTensorDesc src0;
285+ MstxTensorDesc src1;
286+ MstxVecWrapper wrapper;
287+ uint32_t blockNum;
288+ uint32_t dstBlockStride;
289+ uint32_t src0BlockStride;
290+ uint32_t src1BlockStride;
291+ uint32_t repeatTimes;
292+ uint32_t dstRepeatStride;
293+ uint32_t src0RepeatStride;
294+ uint32_t src1RepeatStride;
295+ char name[64];
296+};
297+ 
298+struct MstxVecBilinearInterpolation {
299+ MstxTensorDesc dst;
300+ MstxTensorDesc src0;
301+ MstxTensorDesc src1;
302+ MstxTensorDesc src0Offset;
303+ MstxTensorDesc shared;
304+ MstxVecWrapper wrapper;
305+ uint32_t hRepeat;
306+ bool repeatMode;
307+ uint32_t dstBlockStride;
308+ uint32_t vROffset;
309+ uint32_t vRepeat;
310+ char name[64];
311+};
312+ 
313+struct MstxVecTranspose {
314+ MstxTensorDesc dst;
315+ MstxTensorDesc src;
316+ MstxTensorDesc shared;
317+ bool temp;
318+ char name[64];
319+};
320+ 
321+struct MstxVecSelDesc {
322+ MstxTensorDesc dst;
323+ MstxTensorDesc src0;
324+ MstxTensorDesc src1;
325+ MstxTensorDesc mask;
326+ MstxVecWrapper wrapper;
327+ bool scalarMode;
328+ uint32_t blockNum;
329+ uint32_t dstBlockStride;
330+ uint32_t src0BlockStride;
331+ uint32_t src1BlockStride;
332+ uint32_t repeatTimes;
333+ uint32_t dstRepeatStride;
334+ uint32_t src0RepeatStride;
335+ uint32_t src1RepeatStride;
336+ char name[64];
337+};
338+ 
339+enum class MstxGatherMaskMode {
340+ V1,
341+ V2
342+};
343+ 
344+struct MstxVecGatherMaskDesc {
345+ MstxTensorDesc dst;
346+ MstxTensorDesc src;
347+ MstxVecWrapper wrapper;
348+ MstxGatherMaskMode mode;
349+ uint32_t repeatTimes;
350+ uint32_t src0BlockStride;
351+ uint32_t src0RepeatStride;
352+ uint32_t src1RepeatStride;
353+ char name[64];
354+};
355+ 
356+__aicore__ inline void CopyName(char b[64], __gm__ const char* a)
357+{
358+ uint32_t i = 0;
359+ for (; i < 63; ++i){
360+ b[i] = a[i];
361+ if (a[i] == '\0') {
362+ break;
363+ }
364+ }
365+ b[i] = '\0';
366+}
367+ 
368+ 
369+template <typename T, typename U, bool isSetMask>
370+__aicore__ inline void GetMstxVecUnaryInfo(const LocalTensor<T>& dst, const LocalTensor<U>& src, uint64_t mask0, uint64_t mask1,
371+ const uint8_t repeatTime, const UnaryRepeatParams& repeatParams, __gm__ const char* name)
372+{
373+ MstxVecUnaryDesc mstxVecUnaryDesc;
374+ mstxVecUnaryDesc.dst = From(dst);
375+ mstxVecUnaryDesc.src = From(src);
376+ mstxVecUnaryDesc.wrapper = WrapperFrom(MSTX_MASK_FROM_REG, mask0, mask1, isSetMask, static_cast<uint32_t>(0));
377+ mstxVecUnaryDesc.blockNum = repeatParams.blockNumber;
378+ mstxVecUnaryDesc.repeatTimes = repeatTime;
379+ mstxVecUnaryDesc.dstBlockStride = repeatParams.dstBlkStride;
380+ mstxVecUnaryDesc.srcBlockStride = repeatParams.srcBlkStride;
381+ mstxVecUnaryDesc.dstRepeatStride = repeatParams.dstRepStride;
382+ mstxVecUnaryDesc.srcRepeatStride = repeatParams.srcRepStride;
383+ CopyName(mstxVecUnaryDesc.name, name);
384+ 
385+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_UNARY), sizeof(mstxVecUnaryDesc), &mstxVecUnaryDesc);
386+}
387+ 
388+template <typename T, typename U, bool isSetMask>
389+__aicore__ inline void GetMstxVecUnaryInfo(const LocalTensor<T>& dst, const LocalTensor<U>& src, uint64_t mask,
390+ const uint8_t repeatTime, const UnaryRepeatParams& repeatParams, __gm__ const char* name)
391+{
392+ MstxVecUnaryDesc mstxVecUnaryDesc;
393+ mstxVecUnaryDesc.dst = From(dst);
394+ mstxVecUnaryDesc.src = From(src);
395+ mstxVecUnaryDesc.wrapper = WrapperFrom<T>(MSTX_MASK_FROM_REG, mask, isSetMask, static_cast<uint32_t>(0));
396+ mstxVecUnaryDesc.blockNum = repeatParams.blockNumber;
397+ mstxVecUnaryDesc.repeatTimes = repeatTime;
398+ mstxVecUnaryDesc.dstBlockStride = repeatParams.dstBlkStride;
399+ mstxVecUnaryDesc.srcBlockStride = repeatParams.srcBlkStride;
400+ mstxVecUnaryDesc.dstRepeatStride = repeatParams.dstRepStride;
401+ mstxVecUnaryDesc.srcRepeatStride = repeatParams.srcRepStride;
402+ CopyName(mstxVecUnaryDesc.name, name);
403+ 
404+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_UNARY), sizeof(mstxVecUnaryDesc), &mstxVecUnaryDesc);
405+}
406+ 
407+template <typename T, typename U, bool isSetMask>
408+__aicore__ inline void GetMstxVecUnaryInfo(const LocalTensor<T>& dst, const LocalTensor<U>& src, __gm__ const char* name,
409+ const int32_t count)
410+{
411+ MstxVecUnaryDesc mstxVecUnaryDesc;
412+ mstxVecUnaryDesc.dst = From(dst);
413+ mstxVecUnaryDesc.src = From(src);
414+ mstxVecUnaryDesc.wrapper = WrapperFrom(MSTX_MASK_COUNT, count, static_cast<uint64_t>(0), isSetMask, static_cast<uint32_t>(0));
415+ mstxVecUnaryDesc.blockNum = 8;
416+ mstxVecUnaryDesc.repeatTimes = 1;
417+ mstxVecUnaryDesc.dstBlockStride = 1;
418+ mstxVecUnaryDesc.srcBlockStride = 1;
419+ mstxVecUnaryDesc.dstRepeatStride = 8;
420+ mstxVecUnaryDesc.srcRepeatStride = 8;
421+ CopyName(mstxVecUnaryDesc.name, name);
422+ 
423+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_UNARY), sizeof(mstxVecUnaryDesc), &mstxVecUnaryDesc);
424+}
425+ 
426+template <typename T, typename U, bool isSetMask>
427+__aicore__ inline void GetMstxVecUnaryTenaryInfo(const LocalTensor<T>& dst, const LocalTensor<U>& src, uint64_t mask0, uint64_t mask1,
428+ const uint8_t repeatTime, const UnaryRepeatParams& repeatParams, __gm__ const char* name)
429+{
430+ MstxVecUnaryDesc mstxVecUnaryDesc;
431+ mstxVecUnaryDesc.dst = From(dst);
432+ mstxVecUnaryDesc.src = From(src);
433+ mstxVecUnaryDesc.wrapper = WrapperFrom(MSTX_MASK_FROM_REG, mask0, mask1, isSetMask, static_cast<uint32_t>(0));
434+ mstxVecUnaryDesc.blockNum = repeatParams.blockNumber;
435+ mstxVecUnaryDesc.repeatTimes = repeatTime;
436+ mstxVecUnaryDesc.dstBlockStride = repeatParams.dstBlkStride;
437+ mstxVecUnaryDesc.srcBlockStride = repeatParams.srcBlkStride;
438+ mstxVecUnaryDesc.dstRepeatStride = repeatParams.dstRepStride;
439+ mstxVecUnaryDesc.srcRepeatStride = repeatParams.srcRepStride;
440+ CopyName(mstxVecUnaryDesc.name, name);
441+ 
442+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_TENARY), sizeof(mstxVecUnaryDesc), &mstxVecUnaryDesc);
443+}
444+ 
445+template <typename T, typename U, bool isSetMask>
446+__aicore__ inline void GetMstxVecUnaryTenaryInfo(const LocalTensor<T>& dst, const LocalTensor<U>& src, uint64_t mask,
447+ const uint8_t repeatTime, const UnaryRepeatParams& repeatParams, __gm__ const char* name)
448+{
449+ MstxVecUnaryDesc mstxVecUnaryDesc;
450+ mstxVecUnaryDesc.dst = From(dst);
451+ mstxVecUnaryDesc.src = From(src);
452+ mstxVecUnaryDesc.wrapper = WrapperFrom<T>(MSTX_MASK_FROM_REG, mask, isSetMask, static_cast<uint32_t>(0));
453+ mstxVecUnaryDesc.blockNum = repeatParams.blockNumber;
454+ mstxVecUnaryDesc.repeatTimes = repeatTime;
455+ mstxVecUnaryDesc.dstBlockStride = repeatParams.dstBlkStride;
456+ mstxVecUnaryDesc.srcBlockStride = repeatParams.srcBlkStride;
457+ mstxVecUnaryDesc.dstRepeatStride = repeatParams.dstRepStride;
458+ mstxVecUnaryDesc.srcRepeatStride = repeatParams.srcRepStride;
459+ CopyName(mstxVecUnaryDesc.name, name);
460+ 
461+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_TENARY), sizeof(mstxVecUnaryDesc), &mstxVecUnaryDesc);
462+}
463+ 
464+template <typename T, typename U, bool isSetMask>
465+__aicore__ inline void GetMstxVecUnaryTenaryInfo(const LocalTensor<T>& dst, const LocalTensor<U>& src, __gm__ const char* name,
466+ const int32_t count)
467+{
468+ MstxVecUnaryDesc mstxVecUnaryDesc;
469+ mstxVecUnaryDesc.dst = From(dst);
470+ mstxVecUnaryDesc.src = From(src);
471+ mstxVecUnaryDesc.wrapper = WrapperFrom(MSTX_MASK_COUNT, count, static_cast<uint64_t>(0), isSetMask, static_cast<uint32_t>(0));
472+ mstxVecUnaryDesc.blockNum = 8;
473+ mstxVecUnaryDesc.repeatTimes = 1;
474+ mstxVecUnaryDesc.dstBlockStride = 1;
475+ mstxVecUnaryDesc.srcBlockStride = 1;
476+ mstxVecUnaryDesc.dstRepeatStride = 8;
477+ if constexpr (sizeof(T) > sizeof(U)) {
478+ mstxVecUnaryDesc.srcRepeatStride = DEFAULT_REPEAT_STRIDE / 2 ;
479+ } else {
480+ mstxVecUnaryDesc.srcRepeatStride = DEFAULT_REPEAT_STRIDE;
481+ }
482+ CopyName(mstxVecUnaryDesc.name, name);
483+ 
484+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_TENARY), sizeof(mstxVecUnaryDesc), &mstxVecUnaryDesc);
485+}
486+ 
487+template <typename T, typename U, bool isSetMask>
488+__aicore__ inline void GetMstxVecUnaryCastInfo(const LocalTensor<T>& dst, const LocalTensor<U>& src, uint64_t mask0, uint64_t mask1,
489+ const uint8_t repeatTime, const UnaryRepeatParams& repeatParams, __gm__ const char* name)
490+{
491+ MstxVecUnaryDesc mstxVecUnaryDesc;
C
Cchenyiyuan2月9日

这个也不是默认的

likedislike
492+ mstxVecUnaryDesc.dst = From(dst);
493+ mstxVecUnaryDesc.src = From(src);
494+ mstxVecUnaryDesc.wrapper = WrapperFrom(MSTX_MASK_FROM_REG, mask0, mask1, isSetMask, static_cast<uint32_t>(0));
495+ mstxVecUnaryDesc.blockNum = repeatParams.blockNumber;
496+ mstxVecUnaryDesc.repeatTimes = repeatTime;
497+ mstxVecUnaryDesc.dstBlockStride = repeatParams.dstBlkStride;
498+ mstxVecUnaryDesc.srcBlockStride = repeatParams.srcBlkStride;
499+ mstxVecUnaryDesc.dstRepeatStride = repeatParams.dstRepStride;
500+ mstxVecUnaryDesc.srcRepeatStride = repeatParams.srcRepStride;
501+ CopyName(mstxVecUnaryDesc.name, name);
502+ 
503+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_CAST), sizeof(mstxVecUnaryDesc), &mstxVecUnaryDesc);
504+}
505+ 
506+template <typename T, typename U, bool isSetMask>
507+__aicore__ inline void GetMstxVecUnaryCastInfo(const LocalTensor<T>& dst, const LocalTensor<U>& src, uint64_t mask,
508+ const uint8_t repeatTime, const UnaryRepeatParams& repeatParams, __gm__ const char* name)
509+{
510+ MstxVecUnaryDesc mstxVecUnaryDesc;
511+ mstxVecUnaryDesc.dst = From(dst);
512+ mstxVecUnaryDesc.src = From(src);
513+ mstxVecUnaryDesc.wrapper = WrapperFrom<T>(MSTX_MASK_FROM_REG, mask, isSetMask, static_cast<uint32_t>(0));
514+ mstxVecUnaryDesc.blockNum = repeatParams.blockNumber;
515+ mstxVecUnaryDesc.repeatTimes = repeatTime;
516+ mstxVecUnaryDesc.dstBlockStride = repeatParams.dstBlkStride;
517+ mstxVecUnaryDesc.srcBlockStride = repeatParams.srcBlkStride;
518+ mstxVecUnaryDesc.dstRepeatStride = repeatParams.dstRepStride;
519+ mstxVecUnaryDesc.srcRepeatStride = repeatParams.srcRepStride;
520+ CopyName(mstxVecUnaryDesc.name, name);
521+ 
522+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_CAST), sizeof(mstxVecUnaryDesc), &mstxVecUnaryDesc);
523+}
524+ 
525+template <typename T, typename U, bool isSetMask>
526+__aicore__ inline void GetMstxVecUnaryCastInfo(const LocalTensor<T>& dst, const LocalTensor<U>& src, __gm__ const char* name,
527+ const int32_t count)
528+{
529+ MstxVecUnaryDesc mstxVecUnaryDesc;
530+ mstxVecUnaryDesc.dst = From(dst);
531+ mstxVecUnaryDesc.src = From(src);
532+ mstxVecUnaryDesc.wrapper = WrapperFrom(MSTX_MASK_COUNT, count, static_cast<uint64_t>(0), isSetMask, static_cast<uint32_t>(0));
533+ mstxVecUnaryDesc.blockNum = 8;
534+ mstxVecUnaryDesc.repeatTimes = 1;
535+ mstxVecUnaryDesc.dstBlockStride = 1;
536+ mstxVecUnaryDesc.srcBlockStride = 1;
537+ if constexpr (sizeof(T) > sizeof(U)) {
538+ if constexpr (IsSameType<U, int4b_t>::value) {
539+ mstxVecUnaryDesc.dstRepeatStride = DEFAULT_REPEAT_STRIDE;
540+ mstxVecUnaryDesc.srcRepeatStride = ONE_FOURTH_DEFAULT_REPEAT_STRIDE;
541+ } else {
542+ mstxVecUnaryDesc.dstRepeatStride = DEFAULT_REPEAT_STRIDE;
543+ mstxVecUnaryDesc.srcRepeatStride = DEFAULT_REPEAT_STRIDE / 2;
544+ }
545+ } else if constexpr (sizeof(T) < sizeof(U)) {
546+ if constexpr (IsSameType<T, int4b_t>::value) {
547+ mstxVecUnaryDesc.dstRepeatStride = ONE_FOURTH_DEFAULT_REPEAT_STRIDE;
548+ mstxVecUnaryDesc.srcRepeatStride = DEFAULT_REPEAT_STRIDE;
549+ } else {
550+ mstxVecUnaryDesc.dstRepeatStride = DEFAULT_REPEAT_STRIDE / 2;
551+ mstxVecUnaryDesc.srcRepeatStride = DEFAULT_REPEAT_STRIDE;
552+ }
553+ } else {
554+ mstxVecUnaryDesc.dstRepeatStride = DEFAULT_REPEAT_STRIDE;
555+ mstxVecUnaryDesc.srcRepeatStride = DEFAULT_REPEAT_STRIDE;
556+ }
557+ CopyName(mstxVecUnaryDesc.name, name);
558+ 
559+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_CAST), sizeof(mstxVecUnaryDesc), &mstxVecUnaryDesc);
560+}
561+ 
562+template <typename T, typename U, bool isSetMask>
563+__aicore__ inline void GetMstxVecUnaryCastDeqInfo(const LocalTensor<T>& dst, const LocalTensor<U>& src, uint64_t mask0, uint64_t mask1,
564+ const uint8_t repeatTime, const UnaryRepeatParams& repeatParams, __gm__ const char* name, bool halfBlock)
565+{
566+ MstxVecCastDeqDesc mstxVecCastDeqDesc;
567+ mstxVecCastDeqDesc.dst = From(dst);
568+ mstxVecCastDeqDesc.src = From(src);
569+ mstxVecCastDeqDesc.wrapper = WrapperFrom(MSTX_MASK_FROM_REG, mask0, mask1, isSetMask, static_cast<uint32_t>(0));
570+ mstxVecCastDeqDesc.blockNum = repeatParams.blockNumber;
571+ mstxVecCastDeqDesc.repeatTimes = repeatTime;
572+ mstxVecCastDeqDesc.dstBlockStride = repeatParams.dstBlkStride;
573+ mstxVecCastDeqDesc.srcBlockStride = repeatParams.srcBlkStride;
574+ mstxVecCastDeqDesc.dstRepeatStride = repeatParams.dstRepStride;
575+ mstxVecCastDeqDesc.srcRepeatStride = repeatParams.srcRepStride;
576+ mstxVecCastDeqDesc.halfBlock = halfBlock;
577+ CopyName(mstxVecCastDeqDesc.name, name);
578+ 
579+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_CASTDEQ), sizeof(mstxVecCastDeqDesc), &mstxVecCastDeqDesc);
580+}
581+ 
582+template <typename T, typename U, bool isSetMask>
583+__aicore__ inline void GetMstxVecUnaryCastDeqInfo(const LocalTensor<T>& dst, const LocalTensor<U>& src, uint64_t mask,
584+ const uint8_t repeatTime, const UnaryRepeatParams& repeatParams, __gm__ const char* name, bool halfBlock)
585+{
586+ MstxVecCastDeqDesc mstxVecCastDeqDesc;
587+ mstxVecCastDeqDesc.dst = From(dst);
588+ mstxVecCastDeqDesc.src = From(src);
589+ mstxVecCastDeqDesc.wrapper = WrapperFrom<U>(MSTX_MASK_FROM_REG, mask, isSetMask, static_cast<uint32_t>(0));
590+ mstxVecCastDeqDesc.blockNum = repeatParams.blockNumber;
591+ mstxVecCastDeqDesc.repeatTimes = repeatTime;
592+ mstxVecCastDeqDesc.dstBlockStride = repeatParams.dstBlkStride;
593+ mstxVecCastDeqDesc.srcBlockStride = repeatParams.srcBlkStride;
594+ mstxVecCastDeqDesc.dstRepeatStride = repeatParams.dstRepStride;
595+ mstxVecCastDeqDesc.srcRepeatStride = repeatParams.srcRepStride;
596+ mstxVecCastDeqDesc.halfBlock = halfBlock;
597+ CopyName(mstxVecCastDeqDesc.name, name);
598+ 
599+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_CASTDEQ), sizeof(mstxVecCastDeqDesc), &mstxVecCastDeqDesc);
600+}
601+ 
602+template <typename T, typename U, bool isSetMask>
603+__aicore__ inline void GetMstxVecUnaryCastDeqInfo(const LocalTensor<T>& dst, const LocalTensor<U>& src, __gm__ const char* name,
604+ const int32_t count, bool halfBlock)
605+{
606+ MstxVecCastDeqDesc mstxVecCastDeqDesc;
607+ mstxVecCastDeqDesc.dst = From(dst);
608+ mstxVecCastDeqDesc.src = From(src);
609+ mstxVecCastDeqDesc.wrapper = WrapperFrom(MSTX_MASK_COUNT, count, static_cast<uint64_t>(0), isSetMask, static_cast<uint32_t>(0));
610+ mstxVecCastDeqDesc.blockNum = 8;
611+ mstxVecCastDeqDesc.repeatTimes = 1;
612+ mstxVecCastDeqDesc.dstBlockStride = 1;
613+ mstxVecCastDeqDesc.srcBlockStride = 1;
614+ mstxVecCastDeqDesc.dstRepeatStride = 8;
615+ mstxVecCastDeqDesc.srcRepeatStride = 8;
616+ if constexpr (sizeof(T) > sizeof(U)) {
617+ if constexpr (IsSameType<U, int4b_t>::value) {
618+ mstxVecCastDeqDesc.dstRepeatStride = DEFAULT_REPEAT_STRIDE;
619+ mstxVecCastDeqDesc.srcRepeatStride = ONE_FOURTH_DEFAULT_REPEAT_STRIDE;
620+ } else {
621+ mstxVecCastDeqDesc.dstRepeatStride = DEFAULT_REPEAT_STRIDE;
622+ mstxVecCastDeqDesc.srcRepeatStride = DEFAULT_REPEAT_STRIDE / 2;
623+ }
624+ } else if constexpr (sizeof(T) < sizeof(U)) {
625+ if constexpr (IsSameType<T, int4b_t>::value) {
626+ mstxVecCastDeqDesc.dstRepeatStride = ONE_FOURTH_DEFAULT_REPEAT_STRIDE;
627+ mstxVecCastDeqDesc.srcRepeatStride = DEFAULT_REPEAT_STRIDE;
628+ } else {
629+ mstxVecCastDeqDesc.dstRepeatStride = DEFAULT_REPEAT_STRIDE / 2;
630+ mstxVecCastDeqDesc.srcRepeatStride = DEFAULT_REPEAT_STRIDE;
631+ }
632+ } else {
633+ mstxVecCastDeqDesc.dstRepeatStride = DEFAULT_REPEAT_STRIDE;
634+ mstxVecCastDeqDesc.srcRepeatStride = DEFAULT_REPEAT_STRIDE;
635+ }
636+ 
637+ mstxVecCastDeqDesc.halfBlock = halfBlock;
638+ CopyName(mstxVecCastDeqDesc.name, name);
639+ 
640+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_CASTDEQ), sizeof(mstxVecCastDeqDesc), &mstxVecCastDeqDesc);
641+}
642+ 
C
Cchenyiyuan2月9日

涉及count的输入不一致的T, U要确认是否采用1，1，1，8，8，8模式

likedislike
643+//comparescalar双目的
644+template <typename T, typename U, bool isSetMask>
645+__aicore__ inline void GetMstxVecUnaryCmpsInfo(const LocalTensor<T>& dst, const LocalTensor<U>& src, uint64_t mask0, uint64_t mask1,
646+ const uint8_t repeatTime, const UnaryRepeatParams& repeatParams, __gm__ const char* name)
647+{
648+ MstxVecUnaryDesc mstxVecUnaryDesc;
649+ mstxVecUnaryDesc.dst = From(dst);
650+ mstxVecUnaryDesc.src = From(src);
651+ mstxVecUnaryDesc.wrapper = WrapperFrom(MSTX_MASK_FROM_REG, mask0, mask1, isSetMask, static_cast<uint32_t>(0));
652+ mstxVecUnaryDesc.blockNum = repeatParams.blockNumber;
653+ mstxVecUnaryDesc.repeatTimes = repeatTime;
654+ mstxVecUnaryDesc.dstBlockStride = repeatParams.dstBlkStride;
655+ mstxVecUnaryDesc.srcBlockStride = repeatParams.srcBlkStride;
656+ mstxVecUnaryDesc.dstRepeatStride = repeatParams.dstRepStride;
657+ mstxVecUnaryDesc.srcRepeatStride = repeatParams.srcRepStride;
658+ CopyName(mstxVecUnaryDesc.name, name);
659+ 
660+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_CMPS), sizeof(mstxVecUnaryDesc), &mstxVecUnaryDesc);
661+}
662+ 
663+template <typename T, typename U, bool isSetMask>
664+__aicore__ inline void GetMstxVecUnaryCmpsInfo(const LocalTensor<T>& dst, const LocalTensor<U>& src, uint64_t mask,
665+ const uint8_t repeatTime, const UnaryRepeatParams& repeatParams, __gm__ const char* name)
666+{
667+ MstxVecUnaryDesc mstxVecUnaryDesc;
668+ mstxVecUnaryDesc.dst = From(dst);
669+ mstxVecUnaryDesc.src = From(src);
670+ mstxVecUnaryDesc.wrapper = WrapperFrom<T>(MSTX_MASK_FROM_REG, mask, isSetMask, static_cast<uint32_t>(0));
671+ mstxVecUnaryDesc.blockNum = repeatParams.blockNumber;
672+ mstxVecUnaryDesc.repeatTimes = repeatTime;
673+ mstxVecUnaryDesc.dstBlockStride = repeatParams.dstBlkStride;
674+ mstxVecUnaryDesc.srcBlockStride = repeatParams.srcBlkStride;
675+ mstxVecUnaryDesc.dstRepeatStride = repeatParams.dstRepStride;
676+ mstxVecUnaryDesc.srcRepeatStride = repeatParams.srcRepStride;
677+ CopyName(mstxVecUnaryDesc.name, name);
678+ 
679+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_CMPS), sizeof(mstxVecUnaryDesc), &mstxVecUnaryDesc);
680+}
681+ 
682+template <typename T, typename U, bool isSetMask>
683+__aicore__ inline void GetMstxVecUnaryCmpsInfo(const LocalTensor<T>& dst, const LocalTensor<U>& src, __gm__ const char* name,
684+ const int32_t count)
685+{
686+ MstxVecUnaryDesc mstxVecUnaryDesc;
687+ mstxVecUnaryDesc.dst = From(dst);
688+ mstxVecUnaryDesc.src = From(src);
689+ mstxVecUnaryDesc.wrapper = WrapperFrom(MSTX_MASK_COUNT, count, static_cast<uint64_t>(0), isSetMask, static_cast<uint32_t>(0));
690+ mstxVecUnaryDesc.blockNum = 8;
691+ mstxVecUnaryDesc.repeatTimes = 1;
692+ mstxVecUnaryDesc.dstBlockStride = 1;
693+ mstxVecUnaryDesc.srcBlockStride = 1;
694+ mstxVecUnaryDesc.dstRepeatStride = 8;
695+ mstxVecUnaryDesc.srcRepeatStride = 8;
696+ CopyName(mstxVecUnaryDesc.name, name);
697+ 
698+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_CMPS), sizeof(mstxVecUnaryDesc), &mstxVecUnaryDesc);
699+}
700+ 
701+template <typename T, typename U>
702+__aicore__ inline void GetMstxVecReduceInfo(const LocalTensor<U>& dst, const LocalTensor<T>& src, uint64_t mask0, uint64_t mask1,
703+ const int32_t repeatTime, const int32_t dstRepStride, const int32_t srcBlkStride, const int32_t srcRepStride, bool isSetMask, __gm__ const char* name)
704+{
705+ MstxVecReduceDesc mstxVecReduceDesc;
706+ mstxVecReduceDesc.dst = From(dst);
707+ mstxVecReduceDesc.src = From(src);
708+ mstxVecReduceDesc.wrapper = WrapperFrom(MSTX_MASK_FROM_REG, mask0, mask1, isSetMask, static_cast<uint32_t>(0));
709+ mstxVecReduceDesc.repeatTimes = repeatTime;
710+ mstxVecReduceDesc.srcBlockStride = srcBlkStride;
711+ mstxVecReduceDesc.dstRepeatStride = dstRepStride;
712+ mstxVecReduceDesc.srcRepeatStride = srcRepStride;
713+ CopyName(mstxVecReduceDesc.name, name);
714+ 
715+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_WHOLE_REDUCE), sizeof(mstxVecReduceDesc), &mstxVecReduceDesc);
716+}
717+ 
718+template <typename T, typename U>
719+__aicore__ inline void GetMstxVecReduceInfo(const LocalTensor<U>& dst, const LocalTensor<T>& src, uint64_t mask,
720+ const int32_t repeatTime, const int32_t dstRepStride, const int32_t srcBlkStride, const int32_t srcRepStride, bool isSetMask, __gm__ const char* name)
721+{
722+ MstxVecReduceDesc mstxVecReduceDesc;
723+ mstxVecReduceDesc.dst = From(dst);
724+ mstxVecReduceDesc.src = From(src);
725+ mstxVecReduceDesc.wrapper = WrapperFrom<T>(MSTX_MASK_FROM_REG, mask, isSetMask, static_cast<uint32_t>(0));
726+ mstxVecReduceDesc.repeatTimes = repeatTime;
727+ mstxVecReduceDesc.srcBlockStride = srcBlkStride;
728+ mstxVecReduceDesc.dstRepeatStride = dstRepStride;
729+ mstxVecReduceDesc.srcRepeatStride = srcRepStride;
730+ CopyName(mstxVecReduceDesc.name, name);
731+ 
732+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_WHOLE_REDUCE), sizeof(mstxVecReduceDesc), &mstxVecReduceDesc);
733+}
734+ 
735+template <typename T>
736+__aicore__ inline void GetMstxVecCopyInfo(const LocalTensor<T>& dst, const LocalTensor<T>& src, uint64_t mask0, uint64_t mask1,
737+ const int32_t repeatTime, const CopyRepeatParams& repeatParams, bool isSetMask, __gm__ const char* name)
738+{
739+ MstxVecCopy mstxVecCopy;
740+ mstxVecCopy.dst = From(dst);
741+ mstxVecCopy.src = From(src);
742+ mstxVecCopy.wrapper = WrapperFrom(MSTX_MASK_FROM_REG, mask0, mask1, isSetMask, static_cast<uint32_t>(0));
743+ mstxVecCopy.repeatTimes = repeatTime;
744+ mstxVecCopy.dstStride = repeatParams.dstStride;
745+ mstxVecCopy.srcStride = repeatParams.srcStride;
746+ mstxVecCopy.dstRepeatSize = repeatParams.dstRepeatSize;
747+ mstxVecCopy.srcRepeatSize = repeatParams.srcRepeatSize;
748+ CopyName(mstxVecCopy.name, name);
749+ 
750+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_COPY), sizeof(mstxVecCopy), &mstxVecCopy);
751+}
752+ 
753+template <typename T>
754+__aicore__ inline void GetMstxVecCopyInfo(const LocalTensor<T>& dst, const LocalTensor<T>& src, uint64_t mask,
755+ const int32_t repeatTime, const CopyRepeatParams& repeatParams, bool isSetMask, __gm__ const char* name)
756+{
757+ MstxVecCopy mstxVecCopy;
758+ mstxVecCopy.dst = From(dst);
759+ mstxVecCopy.src = From(src);
760+ mstxVecCopy.wrapper = WrapperFrom<T>(MSTX_MASK_FROM_REG, mask, isSetMask, static_cast<uint32_t>(0));
761+ mstxVecCopy.repeatTimes = repeatTime;
762+ mstxVecCopy.dstStride = repeatParams.dstStride;
763+ mstxVecCopy.srcStride = repeatParams.srcStride;
764+ mstxVecCopy.dstRepeatSize = repeatParams.dstRepeatSize;
765+ mstxVecCopy.srcRepeatSize = repeatParams.srcRepeatSize;
766+ CopyName(mstxVecCopy.name, name);
767+ 
768+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_COPY), sizeof(mstxVecCopy), &mstxVecCopy);
769+}
770+ 
771+template <typename T, typename U>
772+__aicore__ inline void GetMstxVecReduceBlkInfo(const LocalTensor<U>& dst, const LocalTensor<T>& src, uint64_t mask0, uint64_t mask1,
773+ const int32_t repeatTime, const int32_t dstRepStride, const int32_t srcBlkStride, const int32_t srcRepStride, bool isSetMask, __gm__ const char* name)
774+{
775+ MstxVecReduceDesc mstxVecReduceDesc;
776+ mstxVecReduceDesc.dst = From(dst);
777+ mstxVecReduceDesc.src = From(src);
778+ mstxVecReduceDesc.wrapper = WrapperFrom(MSTX_MASK_FROM_REG, mask0, mask1, isSetMask, static_cast<uint32_t>(0));
779+ mstxVecReduceDesc.repeatTimes = repeatTime;
780+ mstxVecReduceDesc.srcBlockStride = srcBlkStride;
781+ mstxVecReduceDesc.dstRepeatStride = dstRepStride;
782+ mstxVecReduceDesc.srcRepeatStride = srcRepStride;
783+ CopyName(mstxVecReduceDesc.name, name);
784+ 
785+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_BLK_REDUCE), sizeof(mstxVecReduceDesc), &mstxVecReduceDesc);
786+}
787+ 
788+template <typename T, typename U>
789+__aicore__ inline void GetMstxVecReduceBlkInfo(const LocalTensor<U>& dst, const LocalTensor<T>& src, uint64_t mask,
790+ const int32_t repeatTime, const int32_t dstRepStride, const int32_t srcBlkStride, const int32_t srcRepStride, bool isSetMask, __gm__ const char* name)
791+{
792+ MstxVecReduceDesc mstxVecReduceDesc;
793+ mstxVecReduceDesc.dst = From(dst);
794+ mstxVecReduceDesc.src = From(src);
795+ mstxVecReduceDesc.wrapper = WrapperFrom<T>(MSTX_MASK_FROM_REG, mask, isSetMask, static_cast<uint32_t>(0));
796+ mstxVecReduceDesc.repeatTimes = repeatTime;
797+ mstxVecReduceDesc.srcBlockStride = srcBlkStride;
798+ mstxVecReduceDesc.dstRepeatStride = dstRepStride;
799+ mstxVecReduceDesc.srcRepeatStride = srcRepStride;
800+ CopyName(mstxVecReduceDesc.name, name);
801+ 
802+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_BLK_REDUCE), sizeof(mstxVecReduceDesc), &mstxVecReduceDesc);
803+}
804+ 
805+template <typename T, typename U>
806+__aicore__ inline void GetMstxVecReducePairInfo(const LocalTensor<U>& dst, const LocalTensor<T>& src, uint64_t mask0, uint64_t mask1,
807+ const int32_t repeatTime, const int32_t dstRepStride, const int32_t srcBlkStride, const int32_t srcRepStride, bool isSetMask, __gm__ const char* name)
808+{
809+ MstxVecReduceDesc mstxVecReduceDesc;
810+ mstxVecReduceDesc.dst = From(dst);
811+ mstxVecReduceDesc.src = From(src);
812+ mstxVecReduceDesc.wrapper = WrapperFrom(MSTX_MASK_FROM_REG, mask0, mask1, isSetMask, static_cast<uint32_t>(0));
813+ mstxVecReduceDesc.repeatTimes = repeatTime;
814+ mstxVecReduceDesc.srcBlockStride = srcBlkStride;
815+ mstxVecReduceDesc.dstRepeatStride = dstRepStride;
816+ mstxVecReduceDesc.srcRepeatStride = srcRepStride;
817+ CopyName(mstxVecReduceDesc.name, name);
818+ 
819+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_PAIR_REDUCE), sizeof(mstxVecReduceDesc), &mstxVecReduceDesc);
820+}
821+ 
822+template <typename T, typename U>
823+__aicore__ inline void GetMstxVecReducePairInfo(const LocalTensor<U>& dst, const LocalTensor<T>& src, uint64_t mask,
824+ const int32_t repeatTime, const int32_t dstRepStride, const int32_t srcBlkStride, const int32_t srcRepStride, bool isSetMask, __gm__ const char* name)
825+{
826+ MstxVecReduceDesc mstxVecReduceDesc;
827+ mstxVecReduceDesc.dst = From(dst);
828+ mstxVecReduceDesc.src = From(src);
829+ mstxVecReduceDesc.wrapper = WrapperFrom<T>(MSTX_MASK_FROM_REG, mask, isSetMask, static_cast<uint32_t>(0));
830+ mstxVecReduceDesc.repeatTimes = repeatTime;
831+ mstxVecReduceDesc.srcBlockStride = srcBlkStride;
832+ mstxVecReduceDesc.dstRepeatStride = dstRepStride;
833+ mstxVecReduceDesc.srcRepeatStride = srcRepStride;
834+ CopyName(mstxVecReduceDesc.name, name);
835+ 
836+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_PAIR_REDUCE), sizeof(mstxVecReduceDesc), &mstxVecReduceDesc);
837+}
838+ 
839+template <typename T, typename U>
840+__aicore__ inline void GetMstxVecReduceRepeatInfo(const LocalTensor<U>& dst, const LocalTensor<T>& src, uint64_t mask0, uint64_t mask1,
841+ const int32_t repeatTime, const int32_t dstRepStride, const int32_t srcBlkStride, const int32_t srcRepStride, bool isSetMask, __gm__ const char* name)
842+{
843+ MstxVecReduceDesc mstxVecReduceDesc;
844+ mstxVecReduceDesc.dst = From(dst);
845+ mstxVecReduceDesc.src = From(src);
846+ mstxVecReduceDesc.wrapper = WrapperFrom(MSTX_MASK_FROM_REG, mask0, mask1, isSetMask, static_cast<uint32_t>(0));
847+ mstxVecReduceDesc.repeatTimes = repeatTime;
848+ mstxVecReduceDesc.srcBlockStride = srcBlkStride;
849+ mstxVecReduceDesc.dstRepeatStride = dstRepStride;
850+ mstxVecReduceDesc.srcRepeatStride = srcRepStride;
851+ CopyName(mstxVecReduceDesc.name, name);
852+ 
853+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_REPEAT_REDUCE), sizeof(mstxVecReduceDesc), &mstxVecReduceDesc);
854+}
855+ 
856+template <typename T, typename U>
857+__aicore__ inline void GetMstxVecReduceRepeatInfo(const LocalTensor<U>& dst, const LocalTensor<T>& src, uint64_t mask,
858+ const int32_t repeatTime, const int32_t dstRepStride, const int32_t srcBlkStride, const int32_t srcRepStride, bool isSetMask, __gm__ const char* name)
859+{
860+ MstxVecReduceDesc mstxVecReduceDesc;
861+ mstxVecReduceDesc.dst = From(dst);
862+ mstxVecReduceDesc.src = From(src);
863+ mstxVecReduceDesc.wrapper = WrapperFrom<T>(MSTX_MASK_FROM_REG, mask, isSetMask, static_cast<uint32_t>(0));
864+ mstxVecReduceDesc.repeatTimes = repeatTime;
865+ mstxVecReduceDesc.srcBlockStride = srcBlkStride;
866+ mstxVecReduceDesc.dstRepeatStride = dstRepStride;
867+ mstxVecReduceDesc.srcRepeatStride = srcRepStride;
868+ CopyName(mstxVecReduceDesc.name, name);
869+ 
870+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_REPEAT_REDUCE), sizeof(mstxVecReduceDesc), &mstxVecReduceDesc);
871+}
872+ 
873+template <typename T>
874+__aicore__ inline void GetMstxVecBrcbInfo(const LocalTensor<T>& dst, const LocalTensor<T>& src, const uint8_t repeatTime,
875+ const BrcbRepeatParams& repeatParams, __gm__ const char* name)
876+{
877+ MstxVecBrcbDesc mstxVecBrcbDesc;
878+ mstxVecBrcbDesc.dst = From(dst);
879+ mstxVecBrcbDesc.src = From(src);
880+ mstxVecBrcbDesc.wrapper = WrapperFrom(MSTX_MASK_FROM_REG, 0, 0, false, static_cast<uint32_t>(0));
881+ mstxVecBrcbDesc.repeatTimes = repeatTime;
882+ mstxVecBrcbDesc.dstBlockStride = repeatParams.dstBlkStride;
883+ mstxVecBrcbDesc.dstRepeatStride = repeatParams.dstRepStride;
884+ CopyName(mstxVecBrcbDesc.name, name);
885+ 
886+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_BROADCAST), sizeof(mstxVecBrcbDesc), &mstxVecBrcbDesc);
887+}
888+ 
889+template <typename T>
890+__aicore__ inline void GetMstxVecTransposeInfo(const LocalTensor<T>& dst, const LocalTensor<T>& src, __gm__ const char* name)
891+{
892+ MstxVecTranspose mstxVecTranspose;
893+ mstxVecTranspose.dst = From(dst);
894+ mstxVecTranspose.src = From(src);
895+ mstxVecTranspose.temp = false;
896+ CopyName(mstxVecTranspose.name, name);
897+ 
898+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_TRANSPOSE), sizeof(mstxVecTranspose), &mstxVecTranspose);
899+}
900+ 
901+template <typename T, typename U>
902+__aicore__ inline void GetMstxVecTransposeTempInfo(const LocalTensor<T>& dst, const LocalTensor<T>& src,
903+ const LocalTensor<U> &sharedTmpBuffer, __gm__ const char* name)
904+{
905+ MstxVecTranspose mstxVecTranspose;
906+ mstxVecTranspose.dst = From(dst);
907+ mstxVecTranspose.src = From(src);
908+ mstxVecTranspose.shared = From(sharedTmpBuffer);
909+ mstxVecTranspose.temp = true;
910+ CopyName(mstxVecTranspose.name, name);
911+ 
912+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_TRANSPOSE), sizeof(mstxVecTranspose), &mstxVecTranspose);
913+}
914+ 
915+template <typename T>
916+__aicore__ inline void GetMstxVecDupInfo(const LocalTensor<T>& dst, uint64_t mask0, uint64_t mask1,
917+ const uint8_t repeatTime, const uint16_t dstBlockStride, const uint8_t dstRepeatStride, bool isSetMask, __gm__ const char* name)
918+{
919+ MstxVecDupDesc mstxVecDupDesc;
920+ mstxVecDupDesc.dst = From(dst);
921+ mstxVecDupDesc.wrapper = WrapperFrom(MSTX_MASK_FROM_REG, mask0, mask1, isSetMask, static_cast<uint32_t>(0));
922+ mstxVecDupDesc.repeatTimes = repeatTime;
923+ mstxVecDupDesc.dstBlockStride = dstBlockStride;
924+ mstxVecDupDesc.dstRepeatStride = dstRepeatStride;
925+ CopyName(mstxVecDupDesc.name, name);
926+ 
927+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_DUP), sizeof(mstxVecDupDesc), &mstxVecDupDesc);
928+}
929+ 
930+template <typename T>
931+__aicore__ inline void GetMstxVecDupInfo(const LocalTensor<T>& dst, uint64_t mask,
932+ const uint8_t repeatTime, const uint16_t dstBlockStride, const uint8_t dstRepeatStride, bool isSetMask, __gm__ const char* name)
933+{
934+ MstxVecDupDesc mstxVecDupDesc;
935+ mstxVecDupDesc.dst = From(dst);
936+ mstxVecDupDesc.wrapper = WrapperFrom<T>(MSTX_MASK_FROM_REG, mask, isSetMask, static_cast<uint32_t>(0));
937+ mstxVecDupDesc.repeatTimes = repeatTime;
938+ mstxVecDupDesc.dstBlockStride = dstBlockStride;
939+ mstxVecDupDesc.dstRepeatStride = dstRepeatStride;
940+ CopyName(mstxVecDupDesc.name, name);
941+ 
942+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_DUP), sizeof(mstxVecDupDesc), &mstxVecDupDesc);
943+}
944+ 
945+template <typename T>
946+__aicore__ inline void GetMstxVecDupInfo(const LocalTensor<T>& dst, const int32_t& count, __gm__ const char* name)
947+{
948+ MstxVecDupDesc mstxVecDupDesc;
949+ mstxVecDupDesc.dst = From(dst);
950+ mstxVecDupDesc.wrapper = WrapperFrom(MSTX_MASK_COUNT, count, static_cast<uint64_t>(0), true, static_cast<uint32_t>(0));
951+ mstxVecDupDesc.repeatTimes = 1;
952+ mstxVecDupDesc.dstBlockStride = 1;
953+ mstxVecDupDesc.dstRepeatStride = 8;
954+ CopyName(mstxVecDupDesc.name, name);
955+ 
956+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_DUP), sizeof(mstxVecDupDesc), &mstxVecDupDesc);
957+}
958+ 
959+template <typename T>
960+__aicore__ inline void GetMstxVecIndexInfo(const LocalTensor<T>& dst, uint64_t mask0, uint64_t mask1,
961+ const uint8_t repeatTime, const uint16_t dstBlockStride, const uint8_t dstRepeatStride, __gm__ const char* name)
962+{
963+ MstxVecDupDesc mstxVecDupDesc;
964+ mstxVecDupDesc.dst = From(dst);
965+ mstxVecDupDesc.wrapper = WrapperFrom(MSTX_MASK_FROM_REG, mask0, mask1, true, static_cast<uint32_t>(8192));
966+ mstxVecDupDesc.repeatTimes = repeatTime;
967+ mstxVecDupDesc.dstBlockStride = dstBlockStride;
968+ mstxVecDupDesc.dstRepeatStride = dstRepeatStride;
969+ CopyName(mstxVecDupDesc.name, name);
970+ 
971+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_VCI), sizeof(mstxVecDupDesc), &mstxVecDupDesc);
972+}
973+ 
974+template <typename T>
975+__aicore__ inline void GetMstxVecIndexInfo(const LocalTensor<T>& dst, uint64_t mask,
976+ const uint8_t repeatTime, const uint16_t dstBlockStride, const uint8_t dstRepeatStride, __gm__ const char* name)
977+{
978+ MstxVecDupDesc mstxVecDupDesc;
979+ mstxVecDupDesc.dst = From(dst);
980+ mstxVecDupDesc.wrapper = WrapperFrom<T>(MSTX_MASK_FROM_REG, mask, true, static_cast<uint32_t>(8192));
981+ mstxVecDupDesc.repeatTimes = repeatTime;
982+ mstxVecDupDesc.dstBlockStride = dstBlockStride;
983+ mstxVecDupDesc.dstRepeatStride = dstRepeatStride;
984+ CopyName(mstxVecDupDesc.name, name);
985+ 
986+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_VCI), sizeof(mstxVecDupDesc), &mstxVecDupDesc);
987+}
988+ 
989+template <typename T>
990+__aicore__ inline void GetMstxVecIndexInfo(const LocalTensor<T>& dst, uint32_t count, __gm__ const char* name)
991+{
992+ MstxVecDupDesc mstxVecDupDesc;
993+ mstxVecDupDesc.dst = From(dst);
994+ mstxVecDupDesc.wrapper = WrapperFrom(MSTX_MASK_COUNT, count, static_cast<uint64_t>(0), true, static_cast<uint32_t>(8192));
995+ mstxVecDupDesc.repeatTimes = 1;
996+ mstxVecDupDesc.dstBlockStride = 1;
997+ mstxVecDupDesc.dstRepeatStride = 8;
998+ CopyName(mstxVecDupDesc.name, name);
999+ 
1000+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_VCI), sizeof(mstxVecDupDesc), &mstxVecDupDesc);
1001+}
1002+ 
1003+//软仿
1004+template <typename T, typename U>
1005+__aicore__ inline void GetMstxVecReduceComplexInfo(const LocalTensor<U>& dst, const LocalTensor<T>& src,
1006+ const LocalTensor<T>& sharedTmpBuffer, const int32_t mask0, const int32_t mask1,
1007+ const int32_t repeatTime, const int32_t srcRepStride, __gm__ const char* name)
1008+{
1009+ MstxVecComplexReduceDesc mstxVecComplexReduceDesc;
1010+ mstxVecComplexReduceDesc.dst = From(dst);
1011+ mstxVecComplexReduceDesc.src = From(src);
1012+ mstxVecComplexReduceDesc.tmp = From(sharedTmpBuffer);
1013+ mstxVecComplexReduceDesc.wrapper = WrapperFrom(MSTX_MASK_FROM_REG, mask0, mask1, true, static_cast<uint32_t>(8192));
1014+ mstxVecComplexReduceDesc.repeatTimes = repeatTime;
1015+ mstxVecComplexReduceDesc.srcRepeatStride = srcRepStride;
1016+ CopyName(mstxVecComplexReduceDesc.name, name);
1017+ 
1018+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_REDUCE), sizeof(mstxVecComplexReduceDesc), &mstxVecComplexReduceDesc);
1019+}
1020+ 
1021+template <typename T, typename U>
1022+__aicore__ inline void GetMstxVecReduceComplexInfo(const LocalTensor<U>& dst, const LocalTensor<T>& src,
1023+ const LocalTensor<T>& sharedTmpBuffer, const int32_t mask,
1024+ const int32_t repeatTime, const int32_t srcRepStride, __gm__ const char* name)
1025+{
1026+ MstxVecComplexReduceDesc mstxVecComplexReduceDesc;
1027+ mstxVecComplexReduceDesc.dst = From(dst);
1028+ mstxVecComplexReduceDesc.src = From(src);
1029+ mstxVecComplexReduceDesc.tmp = From(sharedTmpBuffer);
1030+ mstxVecComplexReduceDesc.wrapper = WrapperFrom<T>(MSTX_MASK_FROM_REG, mask, true, static_cast<uint32_t>(8192));
1031+ mstxVecComplexReduceDesc.repeatTimes = repeatTime;
1032+ mstxVecComplexReduceDesc.srcRepeatStride = srcRepStride;
1033+ CopyName(mstxVecComplexReduceDesc.name, name);
1034+ 
1035+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_REDUCE), sizeof(mstxVecComplexReduceDesc), &mstxVecComplexReduceDesc);
1036+}
1037+ 
1038+template <typename T, typename U>
1039+__aicore__ inline void GetMstxVecReduceComplexInfo(const LocalTensor<U>& dst, const LocalTensor<T>& src,
1040+ const LocalTensor<T>& sharedTmpBuffer, const int32_t count, __gm__ const char* name)
1041+{
1042+ MstxVecComplexReduceDesc mstxVecComplexReduceDesc;
1043+ mstxVecComplexReduceDesc.dst = From(dst);
1044+ mstxVecComplexReduceDesc.src = From(src);
1045+ mstxVecComplexReduceDesc.tmp = From(sharedTmpBuffer);
1046+ mstxVecComplexReduceDesc.wrapper = WrapperFrom(MSTX_MASK_COUNT, count, static_cast<uint64_t>(0), true, static_cast<uint32_t>(8192));
1047+ mstxVecComplexReduceDesc.repeatTimes = 1;
1048+ mstxVecComplexReduceDesc.srcRepeatStride = 8;
1049+ CopyName(mstxVecComplexReduceDesc.name, name);
1050+ 
1051+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_REDUCE), sizeof(mstxVecComplexReduceDesc), &mstxVecComplexReduceDesc);
1052+}
1053+ 
1054+template <typename T, typename U, typename V>
1055+__aicore__ inline void GetMstxVecBinaryInfo(const LocalTensor<T>& dst, const LocalTensor<U>& src0,
1056+ const LocalTensor<V>& src1, uint64_t mask0, uint64_t mask1, const uint8_t repeatTime,
1057+ const BinaryRepeatParams& repeatParams, bool isSetMask, __gm__ const char* name)
1058+{
1059+ MstxVecBinaryDesc mstxVecBinaryDesc;
1060+ mstxVecBinaryDesc.dst = From(dst);
1061+ mstxVecBinaryDesc.src0 = From(src0);
1062+ mstxVecBinaryDesc.src1 = From(src1);
1063+ mstxVecBinaryDesc.wrapper = WrapperFrom(MSTX_MASK_FROM_REG, mask0, mask1, isSetMask, static_cast<uint32_t>(0));
1064+ mstxVecBinaryDesc.blockNum = repeatParams.blockNumber;
1065+ mstxVecBinaryDesc.repeatTimes = repeatTime;
1066+ mstxVecBinaryDesc.dstBlockStride = repeatParams.dstBlkStride;
1067+ mstxVecBinaryDesc.src0BlockStride = repeatParams.src0BlkStride;
1068+ mstxVecBinaryDesc.src1BlockStride = repeatParams.src1BlkStride;
1069+ mstxVecBinaryDesc.dstRepeatStride = repeatParams.dstRepStride;
1070+ mstxVecBinaryDesc.src0RepeatStride = repeatParams.src0RepStride;
1071+ mstxVecBinaryDesc.src1RepeatStride = repeatParams.src1RepStride;
1072+ CopyName(mstxVecBinaryDesc.name, name);
1073+ 
1074+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_BINARY), sizeof(mstxVecBinaryDesc), &mstxVecBinaryDesc);
1075+}
1076+ 
1077+template <typename T, typename U, typename V>
1078+__aicore__ inline void GetMstxVecBinaryInfo(const LocalTensor<T>& dst, const LocalTensor<U>& src0,
1079+ const LocalTensor<V>& src1, uint64_t mask, const uint8_t repeatTime,
1080+ const BinaryRepeatParams& repeatParams, bool isSetMask, __gm__ const char* name)
1081+{
1082+ MstxVecBinaryDesc mstxVecBinaryDesc;
1083+ mstxVecBinaryDesc.dst = From(dst);
1084+ mstxVecBinaryDesc.src0 = From(src0);
1085+ mstxVecBinaryDesc.src1 = From(src1);
1086+ mstxVecBinaryDesc.wrapper = WrapperFrom<T>(MSTX_MASK_FROM_REG, mask, isSetMask, static_cast<uint32_t>(0));
1087+ mstxVecBinaryDesc.blockNum = repeatParams.blockNumber;
1088+ mstxVecBinaryDesc.repeatTimes = repeatTime;
1089+ mstxVecBinaryDesc.dstBlockStride = repeatParams.dstBlkStride;
1090+ mstxVecBinaryDesc.src0BlockStride = repeatParams.src0BlkStride;
1091+ mstxVecBinaryDesc.src1BlockStride = repeatParams.src1BlkStride;
1092+ mstxVecBinaryDesc.dstRepeatStride = repeatParams.dstRepStride;
1093+ mstxVecBinaryDesc.src0RepeatStride = repeatParams.src0RepStride;
1094+ mstxVecBinaryDesc.src1RepeatStride = repeatParams.src1RepStride;
1095+ CopyName(mstxVecBinaryDesc.name, name);
1096+ 
1097+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_BINARY), sizeof(mstxVecBinaryDesc), &mstxVecBinaryDesc);
1098+}
1099+ 
1100+template <typename T, typename U, typename V>
1101+__aicore__ inline void GetMstxVecBinaryInfo(const LocalTensor<T>& dst, const LocalTensor<U>& src0,
1102+ const LocalTensor<V>& src1, __gm__ const char* name, const int32_t count)
1103+{
1104+ MstxVecBinaryDesc mstxVecBinaryDesc;
1105+ mstxVecBinaryDesc.dst = From(dst);
1106+ mstxVecBinaryDesc.src0 = From(src0);
1107+ mstxVecBinaryDesc.src1 = From(src1);
1108+ mstxVecBinaryDesc.wrapper = WrapperFrom(MSTX_MASK_COUNT, count, static_cast<uint64_t>(0), true, static_cast<uint32_t>(0));
1109+ mstxVecBinaryDesc.blockNum = 8;
1110+ mstxVecBinaryDesc.repeatTimes = 1;
1111+ mstxVecBinaryDesc.dstBlockStride = 1;
1112+ mstxVecBinaryDesc.src0BlockStride = 1;
1113+ mstxVecBinaryDesc.src1BlockStride = 1;
1114+ mstxVecBinaryDesc.dstRepeatStride = 8;
1115+ mstxVecBinaryDesc.src0RepeatStride = 8;
1116+ mstxVecBinaryDesc.src1RepeatStride = 8;
1117+ CopyName(mstxVecBinaryDesc.name, name);
1118+ 
1119+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_BINARY), sizeof(mstxVecBinaryDesc), &mstxVecBinaryDesc);
1120+}
1121+ 
1122+template <typename T, typename U, typename V>
1123+__aicore__ inline void GetMstxVecBilinearInterpolationInfo(const LocalTensor<T> &dst, const LocalTensor<T> &src0,
1124+ const LocalTensor<U> &src0Offset, const LocalTensor<T> &src1, uint64_t mask0, uint64_t mask1, uint8_t hRepeat,
1125+ bool repeatMode, uint16_t dstBlkStride, uint16_t vROffset, uint8_t vRepeat, const LocalTensor<V> &sharedTmpBuffer, __gm__ const char* name)
1126+{
1127+ MstxVecBilinearInterpolation mstxVecBilinearInterpolation;
1128+ mstxVecBilinearInterpolation.dst = From(dst);
1129+ mstxVecBilinearInterpolation.src0 = From(src0);
1130+ mstxVecBilinearInterpolation.src1 = From(src1);
1131+ mstxVecBilinearInterpolation.src0Offset = From(src0Offset);
1132+ mstxVecBilinearInterpolation.shared = From(sharedTmpBuffer);
1133+ mstxVecBilinearInterpolation.wrapper = WrapperFrom(MSTX_MASK_FROM_REG, mask0, mask1, true, static_cast<uint32_t>(8192));
1134+ mstxVecBilinearInterpolation.hRepeat = hRepeat;
1135+ mstxVecBilinearInterpolation.repeatMode = repeatMode;
1136+ mstxVecBilinearInterpolation.dstBlockStride = dstBlkStride;
1137+ mstxVecBilinearInterpolation.vROffset = vROffset;
1138+ mstxVecBilinearInterpolation.vRepeat = vRepeat;
1139+ CopyName(mstxVecBilinearInterpolation.name, name);
1140+ 
1141+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_BILINEAR_INTERPOLATION), sizeof(mstxVecBilinearInterpolation), &mstxVecBilinearInterpolation);
1142+}
1143+ 
1144+template <typename T, typename U, typename V>
1145+__aicore__ inline void GetMstxVecBilinearInterpolationInfo(const LocalTensor<T> &dst, const LocalTensor<T> &src0,
1146+ const LocalTensor<U> &src0Offset, const LocalTensor<T> &src1, uint64_t mask, uint8_t hRepeat,
1147+ bool repeatMode, uint16_t dstBlkStride, uint16_t vROffset, uint8_t vRepeat, const LocalTensor<V> &sharedTmpBuffer, __gm__ const char* name)
1148+{
1149+ MstxVecBilinearInterpolation mstxVecBilinearInterpolation;
1150+ mstxVecBilinearInterpolation.dst = From(dst);
1151+ mstxVecBilinearInterpolation.src0 = From(src0);
1152+ mstxVecBilinearInterpolation.src1 = From(src1);
1153+ mstxVecBilinearInterpolation.src0Offset = From(src0Offset);
1154+ mstxVecBilinearInterpolation.shared = From(sharedTmpBuffer);
1155+ mstxVecBilinearInterpolation.wrapper = WrapperFrom<T>(MSTX_MASK_FROM_REG, mask, true, static_cast<uint32_t>(8192));
1156+ mstxVecBilinearInterpolation.hRepeat = hRepeat;
1157+ mstxVecBilinearInterpolation.repeatMode = repeatMode;
1158+ mstxVecBilinearInterpolation.dstBlockStride = dstBlkStride;
1159+ mstxVecBilinearInterpolation.vROffset = vROffset;
1160+ mstxVecBilinearInterpolation.vRepeat = vRepeat;
1161+ CopyName(mstxVecBilinearInterpolation.name, name);
1162+ 
1163+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_BILINEAR_INTERPOLATION), sizeof(mstxVecBilinearInterpolation), &mstxVecBilinearInterpolation);
1164+}
1165+ 
1166+//select
1167+//无软仿
1168+template <typename T, typename U>
1169+__aicore__ inline void GetMstxVecBinarySelInfo(const LocalTensor<T>& dst, const LocalTensor<U>& selMask,
1170+ const LocalTensor<T>& src0, const LocalTensor<T>& src1, uint64_t mask0, uint64_t mask1, const uint8_t repeatTime,
1171+ const BinaryRepeatParams& repeatParams, bool isSetMask, SELMODE selMode, __gm__ const char* name)
1172+{
1173+ MstxVecSelDesc mstxVecSelDesc;
1174+ mstxVecSelDesc.dst = From(dst);
1175+ mstxVecSelDesc.src0 = From(src0);
1176+ mstxVecSelDesc.src1 = From(src1);
1177+ mstxVecSelDesc.mask = From(selMask);
1178+ mstxVecSelDesc.scalarMode = (selMode == AscendC::SELMODE::VSEL_TENSOR_SCALAR_MODE);
1179+ if (mstxVecSelDesc.scalarMode) {
1180+ mstxVecSelDesc.wrapper = WrapperFrom(MSTX_MASK_FROM_REG, mask0, mask1, isSetMask, static_cast<uint32_t>(8192));
1181+ } else {
1182+ mstxVecSelDesc.wrapper = WrapperFrom(MSTX_MASK_FROM_REG, mask0, mask1, isSetMask, static_cast<uint32_t>(0));
1183+ }
1184+ mstxVecSelDesc.blockNum = repeatParams.blockNumber;
1185+ mstxVecSelDesc.repeatTimes = repeatTime;
1186+ mstxVecSelDesc.dstBlockStride = repeatParams.dstBlkStride;
1187+ mstxVecSelDesc.src0BlockStride = repeatParams.src0BlkStride;
1188+ mstxVecSelDesc.src1BlockStride = repeatParams.src1BlkStride;
1189+ mstxVecSelDesc.dstRepeatStride = repeatParams.dstRepStride;
1190+ mstxVecSelDesc.src0RepeatStride = repeatParams.src0RepStride;
1191+ mstxVecSelDesc.src1RepeatStride = repeatParams.src1RepStride;
1192+ CopyName(mstxVecSelDesc.name, name);
1193+ 
1194+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_SEL), sizeof(mstxVecSelDesc), &mstxVecSelDesc);
1195+}
1196+ 
1197+template <typename T, typename U>
1198+__aicore__ inline void GetMstxVecBinarySelInfo(const LocalTensor<T>& dst, const LocalTensor<U>& selMask,
1199+ const LocalTensor<T>& src0, const LocalTensor<T>& src1, uint64_t mask, const uint8_t repeatTime,
1200+ const BinaryRepeatParams& repeatParams, bool isSetMask, SELMODE selMode, __gm__ const char* name)
1201+{
1202+ MstxVecSelDesc mstxVecSelDesc;
1203+ mstxVecSelDesc.dst = From(dst);
1204+ mstxVecSelDesc.src0 = From(src0);
1205+ mstxVecSelDesc.src1 = From(src1);
1206+ mstxVecSelDesc.mask = From(selMask);
1207+ mstxVecSelDesc.scalarMode = (selMode == AscendC::SELMODE::VSEL_TENSOR_SCALAR_MODE);
1208+ if (mstxVecSelDesc.scalarMode) {
1209+ mstxVecSelDesc.wrapper = WrapperFrom<T>(MSTX_MASK_FROM_REG, mask, isSetMask, static_cast<uint32_t>(8192));
1210+ } else {
1211+ mstxVecSelDesc.wrapper = WrapperFrom<T>(MSTX_MASK_FROM_REG, mask, isSetMask, static_cast<uint32_t>(0));
1212+ }
1213+ mstxVecSelDesc.blockNum = repeatParams.blockNumber;
1214+ mstxVecSelDesc.repeatTimes = repeatTime;
1215+ mstxVecSelDesc.dstBlockStride = repeatParams.dstBlkStride;
1216+ mstxVecSelDesc.src0BlockStride = repeatParams.src0BlkStride;
1217+ mstxVecSelDesc.src1BlockStride = repeatParams.src1BlkStride;
1218+ mstxVecSelDesc.dstRepeatStride = repeatParams.dstRepStride;
1219+ mstxVecSelDesc.src0RepeatStride = repeatParams.src0RepStride;
1220+ mstxVecSelDesc.src1RepeatStride = repeatParams.src1RepStride;
1221+ CopyName(mstxVecSelDesc.name, name);
1222+ 
1223+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_SEL), sizeof(mstxVecSelDesc), &mstxVecSelDesc);
1224+}
1225+ 
1226+template <typename T>
1227+__aicore__ inline void GetMstxVecBinarySelInfo(const LocalTensor<T>& dst, const LocalTensor<T>& src0,
1228+ const LocalTensor<T>& src1, const uint8_t repeatTime,
1229+ const BinaryRepeatParams& repeatParams, SELMODE selMode, __gm__ const char* name)
1230+{
1231+ MstxVecSelDesc mstxVecSelDesc;
1232+ mstxVecSelDesc.dst = From(dst);
1233+ mstxVecSelDesc.src0 = From(src0);
1234+ mstxVecSelDesc.src1 = From(src1);
1235+ mstxVecSelDesc.scalarMode = (selMode == AscendC::SELMODE::VSEL_TENSOR_SCALAR_MODE);
1236+ if (mstxVecSelDesc.scalarMode) {
1237+ mstxVecSelDesc.wrapper = WrapperFrom(MSTX_MASK_FROM_REG, 0, 0, false, static_cast<uint32_t>(8192));
1238+ } else {
1239+ mstxVecSelDesc.wrapper = WrapperFrom(MSTX_MASK_FROM_REG, 0, 0, false, static_cast<uint32_t>(0));
1240+ }
1241+ mstxVecSelDesc.blockNum = repeatParams.blockNumber;
1242+ mstxVecSelDesc.repeatTimes = repeatTime;
1243+ mstxVecSelDesc.dstBlockStride = repeatParams.dstBlkStride;
1244+ mstxVecSelDesc.src0BlockStride = repeatParams.src0BlkStride;
1245+ mstxVecSelDesc.src1BlockStride = repeatParams.src1BlkStride;
1246+ mstxVecSelDesc.dstRepeatStride = repeatParams.dstRepStride;
1247+ mstxVecSelDesc.src0RepeatStride = repeatParams.src0RepStride;
1248+ mstxVecSelDesc.src1RepeatStride = repeatParams.src1RepStride;
1249+ CopyName(mstxVecSelDesc.name, name);
1250+ 
1251+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_SEL), sizeof(mstxVecSelDesc), &mstxVecSelDesc);
1252+}
1253+ 
1254+template <typename T, typename U>
1255+__aicore__ inline void GetMstxVecBinarySelInfo(const LocalTensor<T>& dst, const LocalTensor<U>& selMask,
1256+ const LocalTensor<T>& src0, const uint8_t repeatTime,
1257+ const BinaryRepeatParams& repeatParams, SELMODE selMode, __gm__ const char* name)
1258+{
1259+ MstxVecSelDesc mstxVecSelDesc;
1260+ mstxVecSelDesc.dst = From(dst);
1261+ mstxVecSelDesc.src0 = From(src0);
1262+ mstxVecSelDesc.mask = From(selMask);
1263+ mstxVecSelDesc.scalarMode = (selMode == AscendC::SELMODE::VSEL_TENSOR_SCALAR_MODE);
1264+ if (mstxVecSelDesc.scalarMode) {
1265+ mstxVecSelDesc.wrapper = WrapperFrom(MSTX_MASK_FROM_REG, 0, 0, false, static_cast<uint32_t>(8192));
1266+ } else {
1267+ mstxVecSelDesc.wrapper = WrapperFrom(MSTX_MASK_FROM_REG, 0, 0, false, static_cast<uint32_t>(0));
1268+ }
1269+ mstxVecSelDesc.blockNum = repeatParams.blockNumber;
1270+ mstxVecSelDesc.repeatTimes = repeatTime;
1271+ mstxVecSelDesc.dstBlockStride = repeatParams.dstBlkStride;
1272+ mstxVecSelDesc.src0BlockStride = repeatParams.src0BlkStride;
1273+ mstxVecSelDesc.src1BlockStride = repeatParams.src1BlkStride;
1274+ mstxVecSelDesc.dstRepeatStride = repeatParams.dstRepStride;
1275+ mstxVecSelDesc.src0RepeatStride = repeatParams.src0RepStride;
1276+ mstxVecSelDesc.src1RepeatStride = repeatParams.src1RepStride;
1277+ CopyName(mstxVecSelDesc.name, name);
1278+ 
1279+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_SEL), sizeof(mstxVecSelDesc), &mstxVecSelDesc);
1280+}
1281+ 
1282+template <typename T, typename U>
1283+__aicore__ inline void GetMstxVecBinarySelInfo(const LocalTensor<T>& dst, const LocalTensor<U>& selMask,
1284+ const LocalTensor<T>& src0, const LocalTensor<T>& src1, uint32_t count, bool isSetMask, SELMODE selMode, __gm__ const char* name)
1285+{
1286+ MstxVecSelDesc mstxVecSelDesc;
1287+ mstxVecSelDesc.dst = From(dst);
1288+ mstxVecSelDesc.src0 = From(src0);
1289+ mstxVecSelDesc.src1 = From(src1);
1290+ mstxVecSelDesc.mask = From(selMask);
1291+ mstxVecSelDesc.scalarMode = (selMode == AscendC::SELMODE::VSEL_TENSOR_SCALAR_MODE);
1292+ if (mstxVecSelDesc.scalarMode) {
1293+ mstxVecSelDesc.wrapper = WrapperFrom(MSTX_MASK_COUNT, count, static_cast<uint64_t>(0), isSetMask, static_cast<uint32_t>(8192));
1294+ } else {
1295+ mstxVecSelDesc.wrapper = WrapperFrom(MSTX_MASK_COUNT, count, static_cast<uint64_t>(0), isSetMask, static_cast<uint32_t>(0));
1296+ }
1297+ mstxVecSelDesc.blockNum = 8;
1298+ mstxVecSelDesc.repeatTimes = 1;
1299+ mstxVecSelDesc.dstBlockStride = 1;
1300+ mstxVecSelDesc.src0BlockStride = 1;
1301+ mstxVecSelDesc.src1BlockStride = 1;
1302+ mstxVecSelDesc.dstRepeatStride = 8;
1303+ mstxVecSelDesc.src0RepeatStride = 8;
1304+ mstxVecSelDesc.src1RepeatStride = 8;
1305+ CopyName(mstxVecSelDesc.name, name);
1306+ 
1307+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_SEL), sizeof(mstxVecSelDesc), &mstxVecSelDesc);
1308+}
1309+ 
1310+template <typename T, typename U>
1311+__aicore__ inline void GetMstxVecBinarySelInfo(const LocalTensor<T>& dst, const LocalTensor<U>& selMask,
1312+ const LocalTensor<T>& src0, uint64_t mask0, uint64_t mask1, const uint8_t repeatTime,
1313+ const BinaryRepeatParams& repeatParams, bool isSetMask, SELMODE selMode, __gm__ const char* name)
1314+{
1315+ MstxVecSelDesc mstxVecSelDesc;
1316+ mstxVecSelDesc.dst = From(dst);
1317+ mstxVecSelDesc.src0 = From(src0);
1318+ mstxVecSelDesc.mask = From(selMask);
1319+ mstxVecSelDesc.wrapper = WrapperFrom(MSTX_MASK_FROM_REG, mask0, mask1, isSetMask, static_cast<uint32_t>(8192));
1320+ mstxVecSelDesc.blockNum = repeatParams.blockNumber;
1321+ mstxVecSelDesc.repeatTimes = repeatTime;
1322+ mstxVecSelDesc.dstBlockStride = repeatParams.dstBlkStride;
1323+ mstxVecSelDesc.src0BlockStride = repeatParams.src0BlkStride;
1324+ mstxVecSelDesc.src1BlockStride = repeatParams.src1BlkStride;
1325+ mstxVecSelDesc.dstRepeatStride = repeatParams.dstRepStride;
1326+ mstxVecSelDesc.src0RepeatStride = repeatParams.src0RepStride;
1327+ mstxVecSelDesc.src1RepeatStride = repeatParams.src1RepStride;
1328+ mstxVecSelDesc.scalarMode = (selMode == AscendC::SELMODE::VSEL_TENSOR_SCALAR_MODE);
1329+ CopyName(mstxVecSelDesc.name, name);
1330+ 
1331+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_SEL), sizeof(mstxVecSelDesc), &mstxVecSelDesc);
1332+}
1333+ 
1334+template <typename T, typename U>
1335+__aicore__ inline void GetMstxVecBinarySelInfo(const LocalTensor<T>& dst, const LocalTensor<U>& selMask,
1336+ const LocalTensor<T>& src0, uint64_t mask, const uint8_t repeatTime,
1337+ const BinaryRepeatParams& repeatParams, bool isSetMask, SELMODE selMode, __gm__ const char* name)
1338+{
1339+ MstxVecSelDesc mstxVecSelDesc;
1340+ mstxVecSelDesc.dst = From(dst);
1341+ mstxVecSelDesc.src0 = From(src0);
1342+ mstxVecSelDesc.mask = From(selMask);
1343+ mstxVecSelDesc.wrapper = WrapperFrom<T>(MSTX_MASK_FROM_REG, mask, isSetMask, static_cast<uint32_t>(8192));
1344+ mstxVecSelDesc.blockNum = repeatParams.blockNumber;
1345+ mstxVecSelDesc.repeatTimes = repeatTime;
1346+ mstxVecSelDesc.dstBlockStride = repeatParams.dstBlkStride;
1347+ mstxVecSelDesc.src0BlockStride = repeatParams.src0BlkStride;
1348+ mstxVecSelDesc.src1BlockStride = repeatParams.src1BlkStride;
1349+ mstxVecSelDesc.dstRepeatStride = repeatParams.dstRepStride;
1350+ mstxVecSelDesc.src0RepeatStride = repeatParams.src0RepStride;
1351+ mstxVecSelDesc.src1RepeatStride = repeatParams.src1RepStride;
1352+ mstxVecSelDesc.scalarMode = (selMode == AscendC::SELMODE::VSEL_TENSOR_SCALAR_MODE);
1353+ CopyName(mstxVecSelDesc.name, name);
1354+ 
1355+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_SEL), sizeof(mstxVecSelDesc), &mstxVecSelDesc);
1356+}
1357+ 
1358+template <typename T, typename U>
1359+__aicore__ inline void GetMstxVecBinarySelInfo(const LocalTensor<T>& dst, const LocalTensor<U>& selMask,
1360+ const LocalTensor<T>& src0, uint32_t count, bool isSetMask, SELMODE selMode, __gm__ const char* name)
1361+{
1362+ MstxVecSelDesc mstxVecSelDesc;
1363+ mstxVecSelDesc.dst = From(dst);
1364+ mstxVecSelDesc.src0 = From(src0);
1365+ mstxVecSelDesc.mask = From(selMask);
1366+ mstxVecSelDesc.wrapper = WrapperFrom(MSTX_MASK_COUNT, count, static_cast<uint64_t>(0), isSetMask, static_cast<uint32_t>(8192));
1367+ mstxVecSelDesc.blockNum = 8;
1368+ mstxVecSelDesc.repeatTimes = 1;
1369+ mstxVecSelDesc.dstBlockStride = 1;
1370+ mstxVecSelDesc.src0BlockStride = 1;
1371+ mstxVecSelDesc.src1BlockStride = 1;
1372+ mstxVecSelDesc.dstRepeatStride = 8;
1373+ mstxVecSelDesc.src0RepeatStride = 8;
1374+ mstxVecSelDesc.src1RepeatStride = 8;
1375+ mstxVecSelDesc.scalarMode = (selMode == AscendC::SELMODE::VSEL_TENSOR_SCALAR_MODE);
1376+ CopyName(mstxVecSelDesc.name, name);
1377+ 
1378+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_SEL), sizeof(mstxVecSelDesc), &mstxVecSelDesc);
1379+}
1380+ 
1381+//compare
1382+template <typename T, typename U, typename V>
1383+__aicore__ inline void GetMstxVecBinaryCmpInfo(const LocalTensor<T>& dst, const LocalTensor<U>& src0,
1384+ const LocalTensor<V>& src1, uint64_t mask0, uint64_t mask1, const uint8_t repeatTime,
1385+ const BinaryRepeatParams& repeatParams, bool isSetMask, __gm__ const char* name)
1386+{
1387+ MstxVecBinaryDesc mstxVecBinaryDesc;
1388+ mstxVecBinaryDesc.dst = From(dst);
1389+ mstxVecBinaryDesc.src0 = From(src0);
1390+ mstxVecBinaryDesc.src1 = From(src1);
1391+ mstxVecBinaryDesc.wrapper = WrapperFrom(MSTX_MASK_FROM_REG, mask0, mask1, isSetMask, static_cast<uint32_t>(0));
1392+ mstxVecBinaryDesc.blockNum = repeatParams.blockNumber;
1393+ mstxVecBinaryDesc.repeatTimes = repeatTime;
1394+ mstxVecBinaryDesc.dstBlockStride = repeatParams.dstBlkStride;
1395+ mstxVecBinaryDesc.src0BlockStride = repeatParams.src0BlkStride;
1396+ mstxVecBinaryDesc.src1BlockStride = repeatParams.src1BlkStride;
1397+ mstxVecBinaryDesc.dstRepeatStride = repeatParams.dstRepStride;
1398+ mstxVecBinaryDesc.src0RepeatStride = repeatParams.src0RepStride;
1399+ mstxVecBinaryDesc.src1RepeatStride = repeatParams.src1RepStride;
1400+ CopyName(mstxVecBinaryDesc.name, name);
1401+ 
1402+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_CMP), sizeof(mstxVecBinaryDesc), &mstxVecBinaryDesc);
1403+}
1404+ 
1405+template <typename T, typename U, typename V>
1406+__aicore__ inline void GetMstxVecBinaryCmpInfo(const LocalTensor<T>& dst, const LocalTensor<U>& src0,
1407+ const LocalTensor<V>& src1, uint64_t mask, const uint8_t repeatTime,
1408+ const BinaryRepeatParams& repeatParams, bool isSetMask, __gm__ const char* name)
1409+{
1410+ MstxVecBinaryDesc mstxVecBinaryDesc;
1411+ mstxVecBinaryDesc.dst = From(dst);
1412+ mstxVecBinaryDesc.src0 = From(src0);
1413+ mstxVecBinaryDesc.src1 = From(src1);
1414+ mstxVecBinaryDesc.wrapper = WrapperFrom<T>(MSTX_MASK_FROM_REG, mask, isSetMask, static_cast<uint32_t>(0));
1415+ mstxVecBinaryDesc.blockNum = repeatParams.blockNumber;
1416+ mstxVecBinaryDesc.repeatTimes = repeatTime;
1417+ mstxVecBinaryDesc.dstBlockStride = repeatParams.dstBlkStride;
1418+ mstxVecBinaryDesc.src0BlockStride = repeatParams.src0BlkStride;
1419+ mstxVecBinaryDesc.src1BlockStride = repeatParams.src1BlkStride;
1420+ mstxVecBinaryDesc.dstRepeatStride = repeatParams.dstRepStride;
1421+ mstxVecBinaryDesc.src0RepeatStride = repeatParams.src0RepStride;
1422+ mstxVecBinaryDesc.src1RepeatStride = repeatParams.src1RepStride;
1423+ CopyName(mstxVecBinaryDesc.name, name);
1424+ 
1425+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_CMP), sizeof(mstxVecBinaryDesc), &mstxVecBinaryDesc);
1426+}
1427+ 
1428+template <typename T, typename U, typename V>
1429+__aicore__ inline void GetMstxVecBinaryCmpInfo(const LocalTensor<T>& dst, const LocalTensor<U>& src0,
1430+ const LocalTensor<V>& src1, __gm__ const char* name, const int32_t count)
1431+{
1432+ MstxVecBinaryDesc mstxVecBinaryDesc;
1433+ mstxVecBinaryDesc.dst = From(dst);
1434+ mstxVecBinaryDesc.src0 = From(src0);
1435+ mstxVecBinaryDesc.src1 = From(src1);
1436+ mstxVecBinaryDesc.wrapper = WrapperFrom(MSTX_MASK_COUNT, count, static_cast<uint64_t>(0), true, static_cast<uint32_t>(0));
1437+ mstxVecBinaryDesc.blockNum = 8;
1438+ mstxVecBinaryDesc.repeatTimes = 1;
1439+ mstxVecBinaryDesc.dstBlockStride = 1;
1440+ mstxVecBinaryDesc.src0BlockStride = 1;
1441+ mstxVecBinaryDesc.src1BlockStride = 1;
1442+ mstxVecBinaryDesc.dstRepeatStride = 8;
1443+ mstxVecBinaryDesc.src0RepeatStride = 8;
1444+ mstxVecBinaryDesc.src1RepeatStride = 8;
1445+ CopyName(mstxVecBinaryDesc.name, name);
1446+ 
1447+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_CMP), sizeof(mstxVecBinaryDesc), &mstxVecBinaryDesc);
1448+}
1449+ 
1450+//软仿的
1451+template <typename T, typename U, typename V>
1452+__aicore__ inline void GetMstxVecBinaryAddReqReluInfo(const LocalTensor<T>& dst, const LocalTensor<U>& src0,
1453+ const LocalTensor<V>& src1, uint64_t mask0, uint64_t mask1, const uint8_t repeatTime,
1454+ const BinaryRepeatParams& repeatParams, bool isSetMask, __gm__ const char* name)
1455+{
1456+ MstxVecBinaryDesc mstxVecBinaryDesc;
1457+ mstxVecBinaryDesc.dst = From(dst);
1458+ mstxVecBinaryDesc.src0 = From(src0);
1459+ mstxVecBinaryDesc.src1 = From(src1);
1460+ mstxVecBinaryDesc.wrapper = WrapperFrom(MSTX_MASK_FROM_REG, mask0, mask1, isSetMask, static_cast<uint32_t>(8192));
1461+ mstxVecBinaryDesc.blockNum = repeatParams.blockNumber;
1462+ mstxVecBinaryDesc.repeatTimes = repeatTime;
1463+ mstxVecBinaryDesc.dstBlockStride = repeatParams.dstBlkStride;
1464+ mstxVecBinaryDesc.src0BlockStride = repeatParams.src0BlkStride;
1465+ mstxVecBinaryDesc.src1BlockStride = repeatParams.src1BlkStride;
1466+ mstxVecBinaryDesc.dstRepeatStride = repeatParams.dstRepStride;
1467+ mstxVecBinaryDesc.src0RepeatStride = repeatParams.src0RepStride;
1468+ mstxVecBinaryDesc.src1RepeatStride = repeatParams.src1RepStride;
1469+ CopyName(mstxVecBinaryDesc.name, name);
1470+ 
1471+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_BINARY), sizeof(mstxVecBinaryDesc), &mstxVecBinaryDesc);
1472+}
1473+ 
1474+template <typename T, typename U, typename V>
1475+__aicore__ inline void GetMstxVecBinaryAddReqReluInfo(const LocalTensor<T>& dst, const LocalTensor<U>& src0,
1476+ const LocalTensor<V>& src1, uint64_t mask, const uint8_t repeatTime,
1477+ const BinaryRepeatParams& repeatParams, bool isSetMask, __gm__ const char* name)
1478+{
1479+ MstxVecBinaryDesc mstxVecBinaryDesc;
1480+ mstxVecBinaryDesc.dst = From(dst);
1481+ mstxVecBinaryDesc.src0 = From(src0);
1482+ mstxVecBinaryDesc.src1 = From(src1);
1483+ mstxVecBinaryDesc.wrapper = WrapperFrom<T>(MSTX_MASK_FROM_REG, mask, isSetMask, static_cast<uint32_t>(8192));
1484+ mstxVecBinaryDesc.blockNum = repeatParams.blockNumber;
1485+ mstxVecBinaryDesc.repeatTimes = repeatTime;
1486+ mstxVecBinaryDesc.dstBlockStride = repeatParams.dstBlkStride;
1487+ mstxVecBinaryDesc.src0BlockStride = repeatParams.src0BlkStride;
1488+ mstxVecBinaryDesc.src1BlockStride = repeatParams.src1BlkStride;
1489+ mstxVecBinaryDesc.dstRepeatStride = repeatParams.dstRepStride;
1490+ mstxVecBinaryDesc.src0RepeatStride = repeatParams.src0RepStride;
1491+ mstxVecBinaryDesc.src1RepeatStride = repeatParams.src1RepStride;
1492+ CopyName(mstxVecBinaryDesc.name, name);
1493+ 
1494+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_BINARY), sizeof(mstxVecBinaryDesc), &mstxVecBinaryDesc);
1495+}
1496+ 
1497+template <typename T, typename U, typename V>
1498+__aicore__ inline void GetMstxVecBinaryAddReqReluInfo(const LocalTensor<T>& dst, const LocalTensor<U>& src0,
1499+ const LocalTensor<V>& src1, __gm__ const char* name, const int32_t count)
1500+{
1501+ MstxVecBinaryDesc mstxVecBinaryDesc;
1502+ mstxVecBinaryDesc.dst = From(dst);
1503+ mstxVecBinaryDesc.src0 = From(src0);
1504+ mstxVecBinaryDesc.src1 = From(src1);
1505+ mstxVecBinaryDesc.wrapper = WrapperFrom(MSTX_MASK_COUNT, count, static_cast<uint64_t>(0), true, static_cast<uint32_t>(8192));
1506+ mstxVecBinaryDesc.blockNum = 8;
1507+ mstxVecBinaryDesc.repeatTimes = 1;
1508+ mstxVecBinaryDesc.dstBlockStride = 1;
1509+ mstxVecBinaryDesc.src0BlockStride = 1;
1510+ mstxVecBinaryDesc.src1BlockStride = 1;
1511+ mstxVecBinaryDesc.dstRepeatStride = 8;
1512+ mstxVecBinaryDesc.src0RepeatStride = 8;
1513+ mstxVecBinaryDesc.src1RepeatStride = 8;
1514+ CopyName(mstxVecBinaryDesc.name, name);
1515+ 
1516+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_BINARY), sizeof(mstxVecBinaryDesc), &mstxVecBinaryDesc);
1517+}
1518+ 
1519+template <typename T>
1520+__aicore__ inline void GetMstxVecBinaryScalarInfo(const LocalTensor<T>& dst, const LocalTensor<T>& src0,
1521+ const LocalTensor<T>& src1, uint64_t mask0, uint64_t mask1, const uint8_t repeatTime,
1522+ const UnaryRepeatParams& repeatParams, bool isSetMask, __gm__ const char* name)
1523+{
1524+ MstxVecBinaryDesc mstxVecBinaryDesc;
1525+ mstxVecBinaryDesc.dst = From(dst);
1526+ mstxVecBinaryDesc.src0 = From(src0);
1527+ mstxVecBinaryDesc.src1 = From(src1);
1528+ mstxVecBinaryDesc.wrapper = WrapperFrom(MSTX_MASK_FROM_REG, mask0, mask1, isSetMask, static_cast<uint32_t>(0));
1529+ mstxVecBinaryDesc.blockNum = repeatParams.blockNumber;
1530+ mstxVecBinaryDesc.repeatTimes = repeatTime;
1531+ mstxVecBinaryDesc.dstBlockStride = repeatParams.dstBlkStride;
1532+ mstxVecBinaryDesc.src0BlockStride = repeatParams.srcBlkStride;
1533+ mstxVecBinaryDesc.src1BlockStride = repeatParams.srcBlkStride;
1534+ mstxVecBinaryDesc.dstRepeatStride = repeatParams.dstRepStride;
1535+ mstxVecBinaryDesc.src0RepeatStride = repeatParams.srcRepStride;
1536+ mstxVecBinaryDesc.src1RepeatStride = repeatParams.srcRepStride;
1537+ CopyName(mstxVecBinaryDesc.name, name);
1538+ 
1539+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_BINARY_SCALAR), sizeof(mstxVecBinaryDesc), &mstxVecBinaryDesc);
1540+}
1541+ 
1542+template <typename T>
1543+__aicore__ inline void GetMstxVecBinaryScalarInfo(const LocalTensor<T>& dst, const LocalTensor<T>& src0,
1544+ const LocalTensor<T>& src1, uint64_t mask, const uint8_t repeatTime,
1545+ const UnaryRepeatParams& repeatParams, bool isSetMask, __gm__ const char* name)
1546+{
1547+ MstxVecBinaryDesc mstxVecBinaryDesc;
1548+ mstxVecBinaryDesc.dst = From(dst);
1549+ mstxVecBinaryDesc.src0 = From(src0);
1550+ mstxVecBinaryDesc.src1 = From(src1);
1551+ mstxVecBinaryDesc.wrapper = WrapperFrom<T>(MSTX_MASK_FROM_REG, mask, isSetMask, static_cast<uint32_t>(0));
1552+ mstxVecBinaryDesc.blockNum = repeatParams.blockNumber;
1553+ mstxVecBinaryDesc.repeatTimes = repeatTime;
1554+ mstxVecBinaryDesc.dstBlockStride = repeatParams.dstBlkStride;
1555+ mstxVecBinaryDesc.src0BlockStride = repeatParams.srcBlkStride;
1556+ mstxVecBinaryDesc.src1BlockStride = repeatParams.srcBlkStride;
1557+ mstxVecBinaryDesc.dstRepeatStride = repeatParams.dstRepStride;
1558+ mstxVecBinaryDesc.src0RepeatStride = repeatParams.srcRepStride;
1559+ mstxVecBinaryDesc.src1RepeatStride = repeatParams.srcRepStride;
1560+ CopyName(mstxVecBinaryDesc.name, name);
1561+ 
1562+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_BINARY_SCALAR), sizeof(mstxVecBinaryDesc), &mstxVecBinaryDesc);
1563+}
1564+ 
1565+template <typename T>
1566+__aicore__ inline void GetMstxVecBinaryScalarInfo(const LocalTensor<T>& dst, const LocalTensor<T>& src0,
1567+ const LocalTensor<T>& src1, bool isSetMask, __gm__ const char* name, const int32_t count)
1568+{
1569+ MstxVecBinaryDesc mstxVecBinaryDesc;
1570+ mstxVecBinaryDesc.dst = From(dst);
1571+ mstxVecBinaryDesc.src0 = From(src0);
1572+ mstxVecBinaryDesc.src1 = From(src1);
1573+ mstxVecBinaryDesc.wrapper = WrapperFrom(MSTX_MASK_COUNT, count, static_cast<uint64_t>(0), isSetMask, static_cast<uint32_t>(0));
1574+ mstxVecBinaryDesc.blockNum = 8;
1575+ mstxVecBinaryDesc.repeatTimes = 1;
1576+ mstxVecBinaryDesc.dstBlockStride = 1;
1577+ mstxVecBinaryDesc.src0BlockStride = 1;
1578+ mstxVecBinaryDesc.src1BlockStride = 1;
1579+ mstxVecBinaryDesc.dstRepeatStride = 8;
1580+ mstxVecBinaryDesc.src0RepeatStride = 8;
1581+ mstxVecBinaryDesc.src1RepeatStride = 8;
1582+ CopyName(mstxVecBinaryDesc.name, name);
1583+ 
1584+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_BINARY_SCALAR), sizeof(mstxVecBinaryDesc), &mstxVecBinaryDesc);
1585+}
1586+ 
1587+//comparescalar的三目的
1588+template <typename T>
1589+__aicore__ inline void GetMstxVecBinaryScalarCmpsInfo(const LocalTensor<T>& dst, const LocalTensor<T>& src0,
1590+ const LocalTensor<T>& src1, uint64_t mask0, uint64_t mask1, const uint8_t repeatTime,
1591+ const UnaryRepeatParams& repeatParams, bool isSetMask, __gm__ const char* name)
1592+{
1593+ MstxVecBinaryDesc mstxVecBinaryDesc;
1594+ mstxVecBinaryDesc.dst = From(dst);
1595+ mstxVecBinaryDesc.src0 = From(src0);
1596+ mstxVecBinaryDesc.src1 = From(src1);
1597+ mstxVecBinaryDesc.wrapper = WrapperFrom(MSTX_MASK_FROM_REG, mask0, mask1, isSetMask, static_cast<uint32_t>(0));
1598+ mstxVecBinaryDesc.blockNum = repeatParams.blockNumber;
1599+ mstxVecBinaryDesc.repeatTimes = repeatTime;
1600+ mstxVecBinaryDesc.dstBlockStride = repeatParams.dstBlkStride;
1601+ mstxVecBinaryDesc.src0BlockStride = repeatParams.srcBlkStride;
1602+ mstxVecBinaryDesc.src1BlockStride = repeatParams.srcBlkStride;
1603+ mstxVecBinaryDesc.dstRepeatStride = repeatParams.dstRepStride;
1604+ mstxVecBinaryDesc.src0RepeatStride = repeatParams.srcRepStride;
1605+ mstxVecBinaryDesc.src1RepeatStride = repeatParams.srcRepStride;
1606+ CopyName(mstxVecBinaryDesc.name, name);
1607+ 
1608+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_CMPS), sizeof(mstxVecBinaryDesc), &mstxVecBinaryDesc);
1609+}
1610+ 
1611+template <typename T>
1612+__aicore__ inline void GetMstxVecBinaryScalarCmpsInfo(const LocalTensor<T>& dst, const LocalTensor<T>& src0,
1613+ const LocalTensor<T>& src1, uint64_t mask, const uint8_t repeatTime,
1614+ const UnaryRepeatParams& repeatParams, bool isSetMask, __gm__ const char* name)
1615+{
1616+ MstxVecBinaryDesc mstxVecBinaryDesc;
1617+ mstxVecBinaryDesc.dst = From(dst);
1618+ mstxVecBinaryDesc.src0 = From(src0);
1619+ mstxVecBinaryDesc.src1 = From(src1);
1620+ mstxVecBinaryDesc.wrapper = WrapperFrom<T>(MSTX_MASK_FROM_REG, mask, isSetMask, static_cast<uint32_t>(0));
1621+ mstxVecBinaryDesc.blockNum = repeatParams.blockNumber;
1622+ mstxVecBinaryDesc.repeatTimes = repeatTime;
1623+ mstxVecBinaryDesc.dstBlockStride = repeatParams.dstBlkStride;
1624+ mstxVecBinaryDesc.src0BlockStride = repeatParams.srcBlkStride;
1625+ mstxVecBinaryDesc.src1BlockStride = repeatParams.srcBlkStride;
1626+ mstxVecBinaryDesc.dstRepeatStride = repeatParams.dstRepStride;
1627+ mstxVecBinaryDesc.src0RepeatStride = repeatParams.srcRepStride;
1628+ mstxVecBinaryDesc.src1RepeatStride = repeatParams.srcRepStride;
1629+ CopyName(mstxVecBinaryDesc.name, name);
1630+ 
1631+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_CMPS), sizeof(mstxVecBinaryDesc), &mstxVecBinaryDesc);
1632+}
1633+ 
1634+template <typename T>
1635+__aicore__ inline void GetMstxVecBinaryScalarCmpsInfo(const LocalTensor<T>& dst, const LocalTensor<T>& src0,
1636+ const LocalTensor<T>& src1, bool isSetMask, __gm__ const char* name, const int32_t count)
1637+{
1638+ MstxVecBinaryDesc mstxVecBinaryDesc;
1639+ mstxVecBinaryDesc.dst = From(dst);
1640+ mstxVecBinaryDesc.src0 = From(src0);
1641+ mstxVecBinaryDesc.src1 = From(src1);
1642+ mstxVecBinaryDesc.wrapper = WrapperFrom(MSTX_MASK_COUNT, count, static_cast<uint64_t>(0), isSetMask, static_cast<uint32_t>(0));
1643+ mstxVecBinaryDesc.blockNum = 8;
1644+ mstxVecBinaryDesc.repeatTimes = 1;
1645+ mstxVecBinaryDesc.dstBlockStride = 1;
1646+ mstxVecBinaryDesc.src0BlockStride = 1;
1647+ mstxVecBinaryDesc.src1BlockStride = 1;
1648+ mstxVecBinaryDesc.dstRepeatStride = 8;
1649+ mstxVecBinaryDesc.src0RepeatStride = 8;
1650+ mstxVecBinaryDesc.src1RepeatStride = 8;
1651+ CopyName(mstxVecBinaryDesc.name, name);
1652+ 
1653+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_CMPS), sizeof(mstxVecBinaryDesc), &mstxVecBinaryDesc);
1654+}
1655+ 
1656+//GatherMask
1657+template <typename T>
1658+__aicore__ inline void GetMstxVecGatherMaskInfo(const LocalTensor<T>& dst, const LocalTensor<T>& src,
1659+ uint32_t mask0, uint32_t mask1, const GatherMaskParams& gatherMaskParams, GatherMaskMode mode, __gm__ const char* name)
1660+{
1661+ MstxVecGatherMaskDesc mstxVecGatherMaskDesc;
1662+ mstxVecGatherMaskDesc.dst = From(dst);
1663+ mstxVecGatherMaskDesc.src = From(src);
1664+ mstxVecGatherMaskDesc.wrapper = WrapperFrom(MSTX_MASK_FROM_REG, mask0, mask1, true, static_cast<uint32_t>(0));
1665+ mstxVecGatherMaskDesc.mode = (mode == GatherMaskMode::VERSION_V2) ? AscendC::MstxTensor::MstxGatherMaskMode::V2 : AscendC::MstxTensor::MstxGatherMaskMode::V1;
1666+ mstxVecGatherMaskDesc.repeatTimes = gatherMaskParams.repeatTimes;
1667+ mstxVecGatherMaskDesc.src0BlockStride = gatherMaskParams.src0BlockStride;
1668+ mstxVecGatherMaskDesc.src0RepeatStride = gatherMaskParams.src0RepeatStride;
1669+ mstxVecGatherMaskDesc.src1RepeatStride = gatherMaskParams.src1RepeatStride;
1670+ CopyName(mstxVecGatherMaskDesc.name, name);
1671+ 
1672+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_GATHER_MASK), sizeof(mstxVecGatherMaskDesc), &mstxVecGatherMaskDesc);
1673+}
1674+ 
1675+template <typename T>
1676+__aicore__ inline void GetMstxVecGatherMaskInfo(const LocalTensor<T>& dst, const LocalTensor<T>& src,
1677+ uint32_t mask, const GatherMaskParams& gatherMaskParams, GatherMaskMode mode, __gm__ const char* name)
1678+{
1679+ MstxVecGatherMaskDesc mstxVecGatherMaskDesc;
1680+ mstxVecGatherMaskDesc.dst = From(dst);
1681+ mstxVecGatherMaskDesc.src = From(src);
1682+ mstxVecGatherMaskDesc.wrapper = WrapperFrom<T>(MSTX_MASK_FROM_REG, mask, true, static_cast<uint32_t>(0));
1683+ mstxVecGatherMaskDesc.mode = (mode == GatherMaskMode::VERSION_V2) ? AscendC::MstxTensor::MstxGatherMaskMode::V2 : AscendC::MstxTensor::MstxGatherMaskMode::V1;
1684+ mstxVecGatherMaskDesc.repeatTimes = gatherMaskParams.repeatTimes;
1685+ mstxVecGatherMaskDesc.src0BlockStride = gatherMaskParams.src0BlockStride;
1686+ mstxVecGatherMaskDesc.src0RepeatStride = gatherMaskParams.src0RepeatStride;
1687+ mstxVecGatherMaskDesc.src1RepeatStride = gatherMaskParams.src1RepeatStride;
1688+ CopyName(mstxVecGatherMaskDesc.name, name);
1689+ 
1690+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_VEC_GATHER_MASK), sizeof(mstxVecGatherMaskDesc), &mstxVecGatherMaskDesc);
1691+}
1692+ 
1693+template <typename T, typename U>
1694+__aicore__ inline void GetMstxDataCopyInfo(const LocalTensor<T>& dst, const GlobalTensor<U>& src,
1695+ const DataCopyParams& repeatParams, __gm__ const char* name)
1696+{
1697+ MstxDataCopyDesc mstxDataCopyDesc;
1698+ mstxDataCopyDesc.dst = From(dst);
1699+ mstxDataCopyDesc.src = FromGm(src);
1700+ mstxDataCopyDesc.nBurst = repeatParams.blockCount;
1701+ mstxDataCopyDesc.lenBurst = repeatParams.blockLen;
1702+ mstxDataCopyDesc.srcGap = repeatParams.srcStride;
1703+ mstxDataCopyDesc.dstGap = repeatParams.dstStride;
1704+ CopyName(mstxDataCopyDesc.name, name);
1705+ 
1706+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_DATA_COPY), sizeof(mstxDataCopyDesc), &mstxDataCopyDesc);
1707+}
1708+ 
1709+template <typename T, typename U>
1710+__aicore__ inline void GetMstxDataCopyInfo(const GlobalTensor<T>& dst, const LocalTensor<U>& src,
1711+ const DataCopyParams& repeatParams, __gm__ const char* name)
1712+{
1713+ MstxDataCopyDesc mstxDataCopyDesc;
1714+ mstxDataCopyDesc.dst = FromGm(dst);
1715+ mstxDataCopyDesc.src = From(src);
1716+ mstxDataCopyDesc.nBurst = repeatParams.blockCount;
1717+ mstxDataCopyDesc.lenBurst = repeatParams.blockLen;
1718+ mstxDataCopyDesc.srcGap = repeatParams.srcStride;
1719+ mstxDataCopyDesc.dstGap = repeatParams.dstStride;
1720+ CopyName(mstxDataCopyDesc.name, name);
1721+ 
1722+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_DATA_COPY), sizeof(mstxDataCopyDesc), &mstxDataCopyDesc);
1723+}
1724+ 
1725+template <typename T, typename U>
1726+__aicore__ inline void GetMstxDataCopyInfo(const LocalTensor<T>& dst, const LocalTensor<U>& src,
1727+ const DataCopyParams& repeatParams, __gm__ const char* name)
1728+{
1729+ MstxDataCopyDesc mstxDataCopyDesc;
1730+ mstxDataCopyDesc.dst = From(dst);
1731+ mstxDataCopyDesc.src = From(src);
1732+ mstxDataCopyDesc.nBurst = repeatParams.blockCount;
1733+ mstxDataCopyDesc.lenBurst = repeatParams.blockLen;
1734+ mstxDataCopyDesc.srcGap = repeatParams.srcStride;
1735+ mstxDataCopyDesc.dstGap = repeatParams.dstStride;
1736+ CopyName(mstxDataCopyDesc.name, name);
1737+ 
1738+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_DATA_COPY), sizeof(mstxDataCopyDesc), &mstxDataCopyDesc);
1739+}
1740+ 
1741+template <typename T, typename U>
1742+__aicore__ inline void GetMstxDataCopyPadInfo(const LocalTensor<T>& dst, const GlobalTensor<U>& src,
1743+ const DataCopyParams& dataCopyParams, const DataCopyPadParams &padParams, __gm__ const char* name)
1744+{
1745+ MstxDataCopyPadDesc mstxDataCopyPadDesc;
1746+ mstxDataCopyPadDesc.dst = From(dst);
1747+ mstxDataCopyPadDesc.src = FromGm(src);
1748+ mstxDataCopyPadDesc.nBurst = dataCopyParams.blockCount;
1749+ mstxDataCopyPadDesc.lenBurst = dataCopyParams.blockLen;
1750+ mstxDataCopyPadDesc.srcGap = dataCopyParams.srcStride;
1751+ mstxDataCopyPadDesc.dstGap = dataCopyParams.dstStride;
1752+ mstxDataCopyPadDesc.leftPad = padParams.leftPadding;
1753+ mstxDataCopyPadDesc.rightPad = padParams.rightPadding;
1754+ CopyName(mstxDataCopyPadDesc.name, name);
1755+ 
1756+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_DATA_COPY_PAD), sizeof(mstxDataCopyPadDesc), &mstxDataCopyPadDesc);
1757+}
1758+ 
1759+template <typename T, typename U>
1760+__aicore__ inline void GetMstxDataCopyPadInfo(const GlobalTensor<T>& dst, const LocalTensor<U>& src,
1761+ const DataCopyParams& dataCopyParams, __gm__ const char* name)
1762+{
1763+ MstxDataCopyPadDesc mstxDataCopyPadDesc;
1764+ mstxDataCopyPadDesc.dst = FromGm(dst);
1765+ mstxDataCopyPadDesc.src = From(src);
1766+ mstxDataCopyPadDesc.nBurst = dataCopyParams.blockCount;
1767+ mstxDataCopyPadDesc.lenBurst = dataCopyParams.blockLen;
1768+ mstxDataCopyPadDesc.srcGap = dataCopyParams.srcStride;
1769+ mstxDataCopyPadDesc.dstGap = dataCopyParams.dstStride;
1770+ mstxDataCopyPadDesc.leftPad = 0;
1771+ mstxDataCopyPadDesc.rightPad = 0;
1772+ CopyName(mstxDataCopyPadDesc.name, name);
1773+ 
1774+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_DATA_COPY_PAD), sizeof(mstxDataCopyPadDesc), &mstxDataCopyPadDesc);
1775+}
1776+ 
1777+template <typename T, typename U>
1778+__aicore__ inline void GetMstxDataCopyPadInfo(const LocalTensor<T>& dst, const GlobalTensor<U>& src,
1779+ const DataCopyExtParams &dataCopyParams, const DataCopyPadExtParams<T> &padParams, __gm__ const char* name)
1780+{
1781+ MstxDataCopyPadDesc mstxDataCopyPadDesc;
1782+ mstxDataCopyPadDesc.dst = From(dst);
1783+ mstxDataCopyPadDesc.src = FromGm(src);
1784+ mstxDataCopyPadDesc.nBurst = dataCopyParams.blockCount;
1785+ mstxDataCopyPadDesc.lenBurst = dataCopyParams.blockLen;
1786+ mstxDataCopyPadDesc.srcGap = dataCopyParams.srcStride;
1787+ mstxDataCopyPadDesc.dstGap = dataCopyParams.dstStride;
1788+ mstxDataCopyPadDesc.leftPad = padParams.leftPadding;
1789+ mstxDataCopyPadDesc.rightPad = padParams.rightPadding;
1790+ CopyName(mstxDataCopyPadDesc.name, name);
1791+ 
1792+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_DATA_COPY_PAD), sizeof(mstxDataCopyPadDesc), &mstxDataCopyPadDesc);
1793+}
1794+ 
1795+template <typename T, typename U>
1796+__aicore__ inline void GetMstxDataCopyPadInfo(const GlobalTensor<T>& dst, const LocalTensor<U>& src,
1797+ const DataCopyExtParams& dataCopyParams, __gm__ const char* name)
1798+{
1799+ MstxDataCopyPadDesc mstxDataCopyPadDesc;
1800+ mstxDataCopyPadDesc.dst = FromGm(dst);
1801+ mstxDataCopyPadDesc.src = From(src);
1802+ mstxDataCopyPadDesc.nBurst = dataCopyParams.blockCount;
1803+ mstxDataCopyPadDesc.lenBurst = dataCopyParams.blockLen;
1804+ mstxDataCopyPadDesc.srcGap = dataCopyParams.srcStride;
1805+ mstxDataCopyPadDesc.dstGap = dataCopyParams.dstStride;
1806+ mstxDataCopyPadDesc.leftPad = 0;
1807+ mstxDataCopyPadDesc.rightPad = 0;
1808+ CopyName(mstxDataCopyPadDesc.name, name);
1809+ 
1810+ __mstx_dfx_report_stub(static_cast<uint32_t>(MstxReportType::MSTX_DATA_COPY_PAD), sizeof(mstxDataCopyPadDesc), &mstxDataCopyPadDesc);
1811+}
1812+ 
1813+ 
1814+ 
1815+ 
1816+ 
1817+ 
1818+ 
1819+ 
1820+ 
1821+ 
1822+}
1823+}
1824+#endif //__MSTX_DFX_REPORT__
1825+#endif //MSTX_TENSOR_INFO_H
Mtests/api/basic_api/CMakeLists.txt+10-0
@@ -20,6 +20,7 @@ set(BASIC_UT_PRODUCT_TYPE_LIST_INIT
20 ascend310p20 ascend310p
21 ascend910B1_AIC21 ascend910B1_AIC
22 ascend910B1_AIV22 ascend910B1_AIV
23+ ascend910B1_AIV_MSTX
23 ascend310B124 ascend310B1
24 ascend950pr_9599_AIC25 ascend950pr_9599_AIC
25 ascend950pr_9599_AIV_BASIC26 ascend950pr_9599_AIV_BASIC
@@ -72,6 +73,11 @@ file(GLOB ASCENDC_TEST_ascend910B1_AIV_CASE_SRC_FILES
72 ascendc_case_ascend910b1/ascendc_case_ascend910b1_aiv/*.cpp73 ascendc_case_ascend910b1/ascendc_case_ascend910b1_aiv/*.cpp
73)74)
74 75 
76+# ascend910B1 aiv mstx test cases
77+file(GLOB ASCENDC_TEST_ascend910B1_AIV_MSTX_CASE_SRC_FILES
78+ ascendc_case_ascend910b1/ascendc_case_ascend910b1_aiv_mstx/test_operator_vec_mstx.cpp
79+)
80+ 
75# ascend910B1 aic test cases81# ascend910B1 aic test cases
76file(GLOB ASCENDC_TEST_ascend910B1_AIC_CASE_SRC_FILES82file(GLOB ASCENDC_TEST_ascend910B1_AIC_CASE_SRC_FILES
77 ascendc_case_ascend910b1/ascendc_case_ascend910b1_aic/*.cpp83 ascendc_case_ascend910b1/ascendc_case_ascend910b1_aic/*.cpp
@@ -150,6 +156,7 @@ foreach(product_type ${BASIC_UT_PRODUCT_TYPE_LIST})
150 $<$<STREQUAL:${product_type},ascend310p>:../common/tik_pv_wrapper.cpp>156 $<$<STREQUAL:${product_type},ascend310p>:../common/tik_pv_wrapper.cpp>
151 $<$<STREQUAL:${product_type},ascend910B1_AIC>:../common/k3_pvwrap.cpp>157 $<$<STREQUAL:${product_type},ascend910B1_AIC>:../common/k3_pvwrap.cpp>
152 $<$<STREQUAL:${product_type},ascend910B1_AIV>:../common/k3_pvwrap.cpp>158 $<$<STREQUAL:${product_type},ascend910B1_AIV>:../common/k3_pvwrap.cpp>
159+ $<$<STREQUAL:${product_type},ascend910B1_AIV_MSTX>:../common/k3_pvwrap.cpp>
153 $<$<STREQUAL:${product_type},ascend310B1>:../common/k3_pvwrap.cpp>160 $<$<STREQUAL:${product_type},ascend310B1>:../common/k3_pvwrap.cpp>
154 $<$<STREQUAL:${product_type},ascend610Lite>:../common/k3_pvwrap.cpp>161 $<$<STREQUAL:${product_type},ascend610Lite>:../common/k3_pvwrap.cpp>
155 $<$<STREQUAL:${product_type},ascend950pr_9599_AIC>:../common/k3_pvwrap.cpp>162 $<$<STREQUAL:${product_type},ascend950pr_9599_AIC>:../common/k3_pvwrap.cpp>
@@ -159,6 +166,7 @@ foreach(product_type ${BASIC_UT_PRODUCT_TYPE_LIST})
159 $<$<STREQUAL:${product_type},mc62cm12aa_BASIC>:../common/k3_pvwrap.cpp>166 $<$<STREQUAL:${product_type},mc62cm12aa_BASIC>:../common/k3_pvwrap.cpp>
160 $<$<STREQUAL:${product_type},mc62cm12aa_FRAMEWORK>:../common/k3_pvwrap.cpp>167 $<$<STREQUAL:${product_type},mc62cm12aa_FRAMEWORK>:../common/k3_pvwrap.cpp>
161 ${ASCENDC_TEST_${product_type}_CASE_SRC_FILES}168 ${ASCENDC_TEST_${product_type}_CASE_SRC_FILES}
169+ $<$<STREQUAL:${product_type},ascend910B1_AIV_MSTX>:${ASCENDC_TEST_ascend910B1_AIV_MSTX_CASE_SRC_FILES}>
162 )170 )
163 171 
164 target_compile_definitions(ascendc_ut_basic_api_${product_type} PRIVATE172 target_compile_definitions(ascendc_ut_basic_api_${product_type} PRIVATE
@@ -170,6 +178,7 @@ foreach(product_type ${BASIC_UT_PRODUCT_TYPE_LIST})
170 $<$<STREQUAL:${product_type},ascend610>:__NPU_ARCH__=2002;__DAV_M200__>178 $<$<STREQUAL:${product_type},ascend610>:__NPU_ARCH__=2002;__DAV_M200__>
171 $<$<STREQUAL:${product_type},ascend910B1_AIC>:__NPU_ARCH__=2201;__DAV_C220__;__DAV_C220_CUBE__;__DAV_CUBE__>179 $<$<STREQUAL:${product_type},ascend910B1_AIC>:__NPU_ARCH__=2201;__DAV_C220__;__DAV_C220_CUBE__;__DAV_CUBE__>
172 $<$<STREQUAL:${product_type},ascend910B1_AIV>:__NPU_ARCH__=2201;__DAV_C220__;__DAV_C220_VEC__;__DAV_VEC__>180 $<$<STREQUAL:${product_type},ascend910B1_AIV>:__NPU_ARCH__=2201;__DAV_C220__;__DAV_C220_VEC__;__DAV_VEC__>
181+ $<$<STREQUAL:${product_type},ascend910B1_AIV_MSTX>:__NPU_ARCH__=2201;__DAV_C220__;__DAV_C220_VEC__;__DAV_VEC__>
173 $<$<STREQUAL:${product_type},ascend310B1>:__NPU_ARCH__=3002;__DAV_M300__>182 $<$<STREQUAL:${product_type},ascend310B1>:__NPU_ARCH__=3002;__DAV_M300__>
174 $<$<STREQUAL:${product_type},ascend610Lite>:__NPU_ARCH__=3102;__DAV_M310__>183 $<$<STREQUAL:${product_type},ascend610Lite>:__NPU_ARCH__=3102;__DAV_M310__>
175 $<$<STREQUAL:${product_type},ascend950pr_9599_AIC>:__NPU_ARCH__=3101;__DAV_C310__;__DAV_C310_CUBE__>184 $<$<STREQUAL:${product_type},ascend950pr_9599_AIC>:__NPU_ARCH__=3101;__DAV_C310__;__DAV_C310_CUBE__>
@@ -200,6 +209,7 @@ foreach(product_type ${BASIC_UT_PRODUCT_TYPE_LIST})
200 $<$<STREQUAL:${product_type},ascend310p>:tikicpulib::ascend310p>209 $<$<STREQUAL:${product_type},ascend310p>:tikicpulib::ascend310p>
201 $<$<STREQUAL:${product_type},ascend910B1_AIC>:tikicpulib::ascend910B1>210 $<$<STREQUAL:${product_type},ascend910B1_AIC>:tikicpulib::ascend910B1>
202 $<$<STREQUAL:${product_type},ascend910B1_AIV>:tikicpulib::ascend910B1>211 $<$<STREQUAL:${product_type},ascend910B1_AIV>:tikicpulib::ascend910B1>
212+ $<$<STREQUAL:${product_type},ascend910B1_AIV_MSTX>:tikicpulib::ascend910B1>
203 $<$<STREQUAL:${product_type},ascend310B1>:tikicpulib::ascend310B1>213 $<$<STREQUAL:${product_type},ascend310B1>:tikicpulib::ascend310B1>
204 $<$<STREQUAL:${product_type},ascend610Lite>:tikicpulib::ascend610Lite>214 $<$<STREQUAL:${product_type},ascend610Lite>:tikicpulib::ascend610Lite>
205 $<$<STREQUAL:${product_type},ascend950pr_9599_AIC>:tikicpulib::ascend950pr_9599>215 $<$<STREQUAL:${product_type},ascend950pr_9599_AIC>:tikicpulib::ascend950pr_9599>
Mtests/test_parts.sh+1-0
@@ -53,6 +53,7 @@ basic_test_one_targets=(
53 # ascendc_ut_basic_api_ascend610Lite53 # ascendc_ut_basic_api_ascend610Lite
54 ascendc_ut_basic_api_ascend910B1_AIC54 ascendc_ut_basic_api_ascend910B1_AIC
55 ascendc_ut_basic_api_ascend910B1_AIV55 ascendc_ut_basic_api_ascend910B1_AIV
56+ ascendc_ut_basic_api_ascend910B1_AIV_MSTX
56)57)
57 58 
58basic_test_two_targets=(59basic_test_two_targets=(