API Changes

To optimize API naming for clarity and ease of understanding, some API names have been changed in the current version compared to CANN 8.5.0 and earlier versions. The following table describes the changes.

Original API Name	New API Name	Remarks
ScalarGetCountOfValue	GetBitCount	None
ScalarCountLeadingZero	CountLeadingZero	None
ScalarGetSFFValue	GetSFFValue	None
ScalarCast	Cast (float to half/int32_t)	None
ToBfloat16	Cast (float to bfloat16_t)	None
ToFloat	Cast (multiple types to float)	The return type of one overloaded version has been changed from float to a template parameter U, where U defaults to float.
FusedMulAddRelu	MulAddRelu	None
CastDeq	CastDequant	None
CompareScalar	Compares	None
GetAccVal	GetReduceRepeatSumSpr	None
GetReduceMaxMinCount	GetReduceRepeatMaxMinSpr	None
SetAtomicNone	DisableDmaAtomic	None
InitConstValue	Fill	None
SetMMLayoutTransform	SetMMRowMajor/SetMMColumnMajor	SetMMLayoutTransform (true) corresponds to SetMMRowMajor(). SetMMLayoutTransform (false) corresponds to SetMMColumnMajor().
SetHF32Mode	SetHF32Mode	The API name remains unchanged, and the parameter type is changed from bool to the enumeration HF32Mode.
SetHF32TransMode	SetHF32TransMode	The API name remains unchanged, and the parameter type is changed from bool to the enumeration HF32TransMode.
CalcTschBlockDim	CalcTschNumBlocks	None

Parent topic: Appendixes