The "dot" operator seems to be missing from the spec.

[seanmcl]

We're trying to implement a spec/parser/etc for StableHLO in Lean. Reading https://openxla.org/stablehlo/spec#ops and the corresponding github file, I don't see docs for dot. There is dot_generic. Should the later somehow subsume the former, or should dot be in the spec? If the former, there are some examples with dot in the stablehlo repo that should be updated.

Thanks!

[ghpvnist]

StableHLO historically inherited dot from MHLO, and since dot_general can fully express dot, we plan to deprecate the op from StableHLO. See openxla/stablehlo#3 for original issue, RFC for its future, and current documentation here.

Edit: We also have a MHLO pass to convert dot to dot_general.

[seanmcl]

I see. Thanks!

I got the dot from the first example in the spec. I guess it will just take a while to expunge all these operators from the different docs and tools.

```mlir
func.func @main(
  %image: tensor<28x28xf32>,
  %weights: tensor<784x10xf32>,
  %bias: tensor<1x10xf32>
) -> tensor<1x10xf32> {
  %0 = "stablehlo.reshape"(%image) : (tensor<28x28xf32>) -> tensor<1x784xf32>
  %1 = "stablehlo.dot"(%0, %weights) : (tensor<1x784xf32>, tensor<784x10xf32>) -> tensor<1x10xf32>
  %2 = "stablehlo.add"(%1, %bias) : (tensor<1x10xf32>, tensor<1x10xf32>) -> tensor<1x10xf32>
  %3 = "stablehlo.constant"() {value = dense<0.0> : tensor<1x10xf32>} : () -> tensor<1x10xf32>
  %4 = "stablehlo.maximum"(%2, %3) : (tensor<1x10xf32>, tensor<1x10xf32>) -> tensor<1x10xf32>
  "func.return"(%4): (tensor<1x10xf32>) -> ()
}