Matrix multiplications in which matrix A and B have bfloat16 data elements.
More...
Matrix multiplications in which matrix A and B have bfloat16 data elements.
For an explanation how these operations works see Multiply Accumulate.
|
| v16accfloat | mul_4x8_8x4 (v32bfloat16 a, v32bfloat16 b) |
| |
| v16accfloat | negmul_4x8_8x4 (v32bfloat16 a, v32bfloat16 b) |
| |
| v16accfloat | mac_4x8_8x4 (v32bfloat16 a, v32bfloat16 b, v16accfloat acc1) |
| |
| v16accfloat | msc_4x8_8x4 (v32bfloat16 a, v32bfloat16 b, v16accfloat acc1) |
| |
| v16accfloat | addmac_4x8_8x4 (v32bfloat16 a, v32bfloat16 b, v16accfloat acc1, v16accfloat acc2) |
| |
| v16accfloat | addmsc_4x8_8x4 (v32bfloat16 a, v32bfloat16 b, v16accfloat acc1, v16accfloat acc2) |
| |
|
| v16accfloat | mac_4x8_8x4_conf (v32bfloat16 a, v32bfloat16 b, v16accfloat acc1, int zero_acc1, int sub_mul, int sub_acc1) |
| |
| v16accfloat | msc_4x8_8x4_conf (v32bfloat16 a, v32bfloat16 b, v16accfloat acc1, int zero_acc1, int sub_mul, int sub_acc1) |
| |
| v16accfloat | addmac_4x8_8x4_conf (v32bfloat16 a, v32bfloat16 b, v16accfloat acc1, v16accfloat acc2, int zero_acc1, int sub_mul, int sub_acc1, int sub_acc2) |
| |
| v16accfloat | addmsc_4x8_8x4_conf (v32bfloat16 a, v32bfloat16 b, v16accfloat acc1, v16accfloat acc2, int zero_acc1, int sub_mul, int sub_acc1, int sub_acc2) |
| |
◆ addmac_4x8_8x4()
- Parameters
-
| a | Matrix A |
| b | Matrix B |
| acc1 | Accumulator 1 input |
| acc2 | Accumulator 2 input |
- Returns
- Result of operation
◆ addmac_4x8_8x4_conf()
- Parameters
-
| a | Matrix A |
| b | Matrix B |
| acc1 | Accumulator 1 input |
| acc2 | Accumulator 2 input |
| zero_acc1 | Zeroing mask for acc1 |
| sub_mul | Negation mask of multiplication result |
| sub_acc1 | Negation mask of acc1 |
| sub_acc2 | Negation mask of acc2 |
- Returns
- Result of operation
◆ addmsc_4x8_8x4()
- Parameters
-
| a | Matrix A |
| b | Matrix B |
| acc1 | Accumulator 1 input |
| acc2 | Accumulator 2 input |
- Returns
- Result of operation
◆ addmsc_4x8_8x4_conf()
- Parameters
-
| a | Matrix A |
| b | Matrix B |
| acc1 | Accumulator 1 input |
| acc2 | Accumulator 2 input |
| zero_acc1 | Zeroing mask for acc1 |
| sub_mul | Negation mask of multiplication result |
| sub_acc1 | Negation mask of acc1 |
| sub_acc2 | Negation mask of acc2 |
- Returns
- Result of operation
◆ mac_4x8_8x4()
- Parameters
-
| a | Matrix A |
| b | Matrix B |
| acc1 | Accumulator 1 input |
- Returns
- Result of operation
◆ mac_4x8_8x4_conf()
- Parameters
-
| a | Matrix A |
| b | Matrix B |
| acc1 | Accumulator 1 input |
| zero_acc1 | Zeroing mask for acc1 |
| sub_mul | Negation mask of multiplication result |
| sub_acc1 | Negation mask of acc1 |
- Returns
- Result of operation
◆ msc_4x8_8x4()
- Parameters
-
| a | Matrix A |
| b | Matrix B |
| acc1 | Accumulator 1 input |
- Returns
- Result of operation
◆ msc_4x8_8x4_conf()
- Parameters
-
| a | Matrix A |
| b | Matrix B |
| acc1 | Accumulator 1 input |
| zero_acc1 | Zeroing mask for acc1 |
| sub_mul | Negation mask of multiplication result |
| sub_acc1 | Negation mask of acc1 |
- Returns
- Result of operation
◆ mul_4x8_8x4()
- Parameters
-
- Returns
- Result of operation
◆ mul_4x8_8x4_conf()
- Parameters
-
| a | Matrix A |
| b | Matrix B |
| sub_mul | Negation mask for multiplication result |
- Returns
- Result of operation
◆ negmul_4x8_8x4()
- Parameters
-
- Returns
- Result of operation
◆ negmul_4x8_8x4_conf()
- Parameters
-
| a | Matrix A |
| b | Matrix B |
| sub_mul | Negation mask for multiplication result. If a bit of sub_mul is set the corresponding vector lane of the output accumulator will be negated. |
- Returns
- Result of operation