Type refactor within MetaModel#92
Conversation
Merge Literal subclasses to one main DataType removed by PrimitiveKind
Try to use everywhere type.kind instead of signed Rename _width to size to get a constant access point in inferTypes
Use type.kind instead of signed size instead of _width
ETISS broken because of actual_size
Also resolve small merge issues
Common Subclass for - Alias - RegisterBanks - Memory - BitFieldDescr - Variables
50c68d4 to
dac6ea0
Compare
|
@PhilippvK I think this is so far rdy for merging/further extensions (see https://github.com/tum-ei-eda/M2-ISA-R/tree/behav_array_support) |
|
How can I test this @PhilippvK |
do you mean how to run the RISCV ISA Test suite on the generated ETISS arch? I can provide an example test script which I use until Friday… For the more special features which are only used by non-standard extensions we will have to come up with our own tests, which can be probably integrated in the etiss_riscv_examples repo |
I think that should suffice for now. I also don't expect any behavioral differences, since the changes are minimal compared to my generated code. I just want to merge this here somewhat fast, so that I can fork properly |
|
Here are the instructions to run the RISC-V test Suite using ETISS: The should be 5 failing tests, which is expected: |
|
@PhilippvK Ready for take off? |
|
Superseded by #98 |
Refactored the typing system to establish a single, consistent, and centralized type model.
Summary:
Typing has been moved out of the arch file and completely redesigned,
This lays hopefully the foundation for future extensions such as arrays, structs, and unions.