CPU Operators Supported by VART ML Runtime#
When a model is compiled with CPU partition support, NPU-incompatible operators are routed to CPU implementations provided by the AMD Vitis™ AI compiler. The following operators are supported by VART ML Runtime for CPU execution. Each operator provides both an x86_64 and an AArch64/ARM implementation.
If a model contains NPU-incompatible operators that are not in this list, it cannot be compiled with CPU partition support and must use standard compilation with ONNX Runtime instead.
Operator |
x86_64 |
AArch64 (ARM) |
Notes |
|---|---|---|---|
Add |
Yes |
Yes |
|
ArgMax |
Yes |
Yes |
|
Cast |
Yes |
Yes |
|
Clip |
Yes |
Yes |
|
Concat |
Yes |
Yes |
|
Conv |
Yes |
Yes |
|
DequantizeLinear |
Yes |
Yes |
|
Div |
Yes |
Yes |
|
Erf |
Yes |
Yes |
|
Expand |
Yes |
Yes |
|
Flatten |
Yes |
Yes |
|
Gather |
Yes |
Yes |
|
Gemm |
Yes |
Yes |
|
Inverse |
Yes |
Yes |
|
Log |
Yes |
Yes |
|
MatMul |
Yes |
Yes |
|
MaxPool |
Yes |
Yes |
|
Mul |
Yes |
Yes |
|
NonMaxSuppression |
Yes |
Yes |
|
QuantizeLinear |
Yes |
Yes |
|
ReduceMax |
Yes |
Yes |
|
ReduceMean |
Yes |
Yes |
|
Relu |
Yes |
Yes |
|
Reshape |
Yes |
Yes |
|
Resize |
Yes |
Yes |
|
ScatterND |
Yes |
Yes |
|
Sigmoid |
Yes |
Yes |
|
Slice |
Yes |
Yes |
|
Softmax |
Yes |
Yes |
|
Split |
Yes |
Yes |
|
Squeeze |
Yes |
Yes |
|
Sub |
Yes |
Yes |
|
Transpose |
Yes |
Yes |
|
Unsqueeze |
Yes |
Yes |