...

Text file src/simd/archsimd/_gen/simdgen/ops/MLOps/go_amd64.yaml

Documentation: simd/archsimd/_gen/simdgen/ops/MLOps

     1!sum
     2- go: DotProductPairs
     3  asm: VPMADDWD
     4  in:
     5  - &int
     6    go: $t
     7    base: int
     8  - *int
     9  out:
    10  - &int2 # The elemBits are different
    11    go: $t2
    12    base: int
    13- go: DotProductPairsSaturated
    14  asm: VPMADDUBSW
    15  in:
    16  - &uint
    17    go: $t
    18    base: uint
    19    overwriteElementBits: 8
    20  - &int3
    21    go: $t3
    22    base: int
    23    overwriteElementBits: 8
    24  out:
    25  - *int2
    26# - go: DotProductBroadcast
    27#   asm: VDPP[SD]
    28#   in:
    29#   - &dpb_src
    30#     go: $t
    31#   - *dpb_src
    32#   - class: immediate
    33#     const: 127
    34#   out:
    35#   - *dpb_src
    36- go: AddDotProductPairs
    37  asm: "VPDPWSSD"
    38  in:
    39  - &pdpa_acc
    40    go: $t_acc
    41    base: int
    42    elemBits: 32
    43  - &pdpa_src1
    44    go: $t_src1
    45    base: int
    46    overwriteElementBits: 16
    47  - &pdpa_src2
    48    go: $t_src2
    49    base: int
    50    overwriteElementBits: 16
    51  out:
    52  - *pdpa_acc
    53- go: MulAdd
    54  asm: "VFMADD213PS|VFMADD213PD"
    55  documentation: !string |-
    56    // NAME performs a fused (x * y) + z.
    57  in:
    58  - &fma_op
    59   go: $t
    60   base: float
    61  - *fma_op
    62  - *fma_op
    63  out:
    64  - *fma_op
    65- go: MulAddEvenSubOdd
    66  asm: "VFMADDSUB213PS|VFMADDSUB213PD"
    67  in:
    68  - *fma_op
    69  - *fma_op
    70  - *fma_op
    71  out:
    72  - *fma_op
    73- go: MulAddOddSubEven
    74  asm: "VFMSUBADD213PS|VFMSUBADD213PD"
    75  in:
    76  - *fma_op
    77  - *fma_op
    78  - *fma_op
    79  out:
    80  - *fma_op
    81- go: SumOf8AbsDiff
    82  asm: "VPSADBW"
    83  in:
    84  - go: $t
    85    base: uint
    86  - go: $t
    87    base: uint
    88  out:
    89  - base: uint
    90    overwriteElementBits: 64

View as plain text