![]() |
AI Engine Intrinsics User Guide (AIE) v(2025.1)
|
Summary of the vector and accumulator registers. More...
Summary of the vector and accumulator registers.
The two main data types offered by the AIE cores are vector and accumulator registers.
Vector data-types pack multiple scalar data elements into a wider vector. There are vector data-types with standard sized elements (8-bit, 16-bit, 32-bit, 64-bit). The total data-width of the vector data-types can be 128-bits, 256-bits, 512-bits, or 1024-bits. A summary of the scalar types and the number of lanes supported is described in the following table. Please refer to Vector Data-types for more details.
| Scalar type | Supported vector lanes |
|---|---|
| int8/uint8 | 16/32/64/128 |
| int16/uint16/cint8 | 8/16/32/64 |
| int32/uint32/cint16 | 4/8/16/32 |
| int64/uint64/cint32 | 2/4/8/16 |
| int128/uint128/cint64 | 1/2/4/8 |
| cint128 | 1/2/4 |
| float | 4/8/16/32 |
| cfloat | 2/4/8/16 |
Accumulator data-types are stored in the accumulator registers within the AI Engine. The total data-width of the accumulator data-types can be 320/384-bits or 640/768-bits. A summary of the scalar types and the number of lanes supported is described in the following table. Please refer to Accumulator Types for more details.
| Scalar type | Supported vector lanes | Number of accumulator lanes for ADF cascade stream API |
|---|---|---|
| acc48 | 8/16 | 8 |
| cacc48 | 4/8 | 4 |
| accfloat | 4/8/16/32 | 8 |
| acc80 | 4/8 | 4 |
| cacc80 | 2/4 | 2 |
| caccfloat | 2/4/8/16/32 | 4 |
Modules | |
| Scalar Data-types | |
| All the standard C scalar data-types are supported. | |
| Vector Data-types | |