mlir 编译器学习笔记之三 -- 测试

1、完成mlir测试用例的check 自动更新

bishengir-opt xx.mlir --split-input-file -torch-decompose-complex-ops -convert-torch-to-hfusion | python mlir/utils/generate-test-checks.py --source xx.mlir -i

2、CHECK对 [[ 有特殊的含义，如果相应匹配，则写为 {{\[\[}}，]] 部分保持原样

3、onnx和torch的dense resource均以小端方式存储

module @test_dense_i64_resource {
func.func @test_dense_i64_resource() -> tensor<2xi64> {
%0 = arith.constant dense_resource<test_dense_i64_resource> : tensor<2xi64>
return %0 : tensor<2xi64>
}
}

// The dense resource data are encoded withlittle endian
{-#
dialect_resources: {
builtin: {
test_dense_i64_resource: "0x0200000000000000000000000100000000000000"
}
}
#-}