[aievec] add I8/I16/I32 reduce + aievec.min/max e2e to-llvm tests

Xilinx / mlir-aie

An MLIR-based toolchain for AMD AI Engine-enabled devices.

Other

260 stars 76 forks source link

Closed jamestcl-amd closed 2 weeks ago

jamestcl-amd commented 2 weeks ago

This PR updates the existing I8/I16/I32 element-wise aievec.min/max e2e test to support the to-llvm flow.

Changes:

Add I8/I16/I32 vector.reduction + aievec.min/max e2e tests for the to-llvm flow.
Update bf16/float mul_elem tests to use -convert-func-to-llvm='use-bare-ptr-memref-call-conv'. See https://github.com/Xilinx/mlir-aie/pull/1555