Mutual Derivation Relationship

Derivation Rules

When the input aclTensor data types of an API (such as aclnnAdd, aclnnMul, etc.) are inconsistent, the API internally derives a data type and converts the input data to that data type for computation.

For the data types supported by aclTensor, see Data Types. Some of these types satisfy the following derivation rules, and the derivation principle is similar to PyTorch's Type Promotion.

NOTE

For ease of description, the data types used in the table are in abbreviated form, with the following meanings: ACL_FLOAT(f32), ACL_FLOAT16(f16), ACL_DOUBLE(f64), ACL_BF16(bf16), ACL_INT8(s8), ACL_UINT8(u8), ACL_INT16(s16), ACL_UINT16(u16), ACL_INT32(s32), ACL_UINT32(u32), ACL_INT64(s64), ACL_UINT64(u64), ACL_BOOL(bool), ACL_COMPLEX32(c32), ACL_COMPLEX64(c64), ACL_COMPLEX128(c128).

The table header and the leftmost column represent the two input data types to be derived, and the corresponding cell in the table indicates the derived data type.

A cross (×) in the table indicates that the two types cannot be used for derivation computation.

Table 1 Data type derivation relationship

Data Type	f32	f16	f64	bf16	s8	u8	s16	u16	s32	u32	s64	u64	bool	c32	c64	c128
f32	f32	f32	f64	f32	f32	f32	f32	N/A	f32	N/A	f32	N/A	f32	c64	c64	c128
f16	f32	f16	f64	f32	f16	f16	f16	N/A	f16	N/A	f16	N/A	f16	c32	c64	c128
f64	f64	f64	f64	f64	f64	f64	f64	N/A	f64	N/A	f64	N/A	f64	c128	c128	c128
bf16	f32	f32	f64	bf16	bf16	bf16	bf16	N/A	bf16	N/A	bf16	N/A	bf16	c32	c64	c128
s8	f32	f16	f64	bf16	s8	s16	s16	N/A	s32	N/A	s64	N/A	s8	c32	c64	c128
u8	f32	f16	f64	bf16	s16	u8	s16	N/A	s32	N/A	s64	N/A	u8	c32	c64	c128
s16	f32	f16	f64	bf16	s16	s16	s16	N/A	s32	N/A	s64	N/A	s16	c32	c64	c128
u16	N/A	N/A	N/A	N/A	N/A	N/A	N/A	u16	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A
s32	f32	f16	f64	bf16	s32	s32	s32	N/A	s32	N/A	s64	N/A	s32	c32	c64	c128
u32	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	u32	N/A	N/A	N/A	N/A	N/A	N/A
s64	f32	f16	f64	bf16	s64	s64	s64	N/A	s64	N/A	s64	N/A	s64	c32	c64	c128
u64	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	u64	N/A	N/A	N/A	N/A
bool	f32	f16	f64	bf16	s8	u8	s16	N/A	s32	N/A	s64	N/A	bool	c32	c64	c128
c32	c64	c32	c128	c32	c32	c32	c32	N/A	c32	N/A	c32	N/A	c32	c32	c64	c128
c64	c64	c64	c128	c64	c64	c64	c64	N/A	c64	N/A	c64	N/A	c64	c64	c64	c128
c128	c128	c128	c128	c128	c128	c128	c128	N/A	c128	N/A	c128	N/A	c128	c128	c128	c128

Derivation Examples

When calling the aclnnAdd API, if the input parameters have inconsistent data types — one being float16 and the other float32 — the API internally converts the float16 data type to float32 and then performs the computation.
When calling the aclnnAdd API, if the input parameters have inconsistent data types — one being float32 and the other bool — the API internally converts the bool data type to float32 and then performs the computation.