Snippets in its first generation can be seen as a generalization over a generic eltwise node. First generation of snippets does not have integration with oneDNN, and the patterns it supports should be kept orthogonal to what is fused with post-ops.
See the example of POC CPU implementation here.
First 8 kernel parameters are passed by structure which is unpacked inside a kernel into the registers. The rest are passed through the stack.
The loop trip count should be placed to a GP register, as well as the work amount. Moreover, you need to load all the parameters into GP registers. If you assume that you have enough registers, then it can be done before the loop body.
auto param0 = abi_params[0];
auto param1 = abi_params[1];
auto result = abi_params[2];
auto work_amount = abi_params[3];
A load could be Vector, Scalar, and Broadcast. Only the native vector size for an architecture is supported (for example, 16 on AVX-512).
Memory operation also generates post increments for the pointer it uses.
MemoryEmitterStoreEmitterScalarStoreEmitterLoadEmitter(post increment)BroadcastLoadEmitterScalarLoadEmitter(post increment)
All inputs and outputs should be the same layout. Re-layout operations are not included in the snippets dialect. Since current scope is limited to layout-oblivious operations no specific handling for blocking is required. Extending dialect with re-layout operations is a subject of further benchmarking. The following memory representation is assumed.
offset domain margin
+-------+-------------------------------+----------+
| | | |
| | | |
| | | |
| | | |
+-------+-------------------------------+----------+
Tensor data can be passed with strides.
Data corresponds to a constant table and wraps this entity for the CPU.