da630ce382526a61166a8428ea20a8e547f5f450
under this new system, dialects can define their own custom attributes, complete with their own print() and parse() callbacks, which can then be used as values in an op's attribute dictionary. alongside custom dialect attributes, the former int, float, and string constant values have been converted to attributes provided by the arith and builtin dialects respectively. the caches for these attributes have also been moved from mie_ctx to their respective dialect data structures. this system will allow new types of attributes to be implemented, including dictionaries, arrays, and references to types themselves (rather than just particular values of a given type).
Table of Contents
About The Project
Mie is an extensible intermediate representation framework inspired by LLVM and MLIR. It is designed to accept high-level program representation from a compiler front-end, lower and optimise it into a flat stream of low-level instructions, and then convert it into a target-specific implementation.
(The name Mie is pronounced like "me", "bee", or "sea", and (despite the logo) is not pronounced like "my", "bye", or "pie")
License
Distributed under the BSD 3-Clause License. See LICENSE for more information.
Description
Languages
C
97.2%
CMake
1.8%
Vim Script
1%