All of the regression (IR) tests
I suggest you use update_llc_test_checks.py to automatically generate the CHECK lines for the IR tests (see the RISC-V backend which does this pretty diligently). Otherwise it's a real nuisance to update them after CodeGen changes that affect tests.
- Fix / Update failed test cases. Now we have 100% passing rate except one XFAIL, yeah!
- Most of the previous failed test cases were caused by less-optimal results from previous version of codegen. For example, the codegen for jump tables.
- The only XFAIL is caused by (selection DAG) ISel which failed to select one instruction that supposed to have 16-bit operand but a 8-bit one was given. It seems to take more time to fix it so just mark XFAIL for now
- Rewrite utils/extract-section into utils/extract-section.py
- The new Python script is also more generic which doesn't have some M68k specific assumption (e.g. always group in 16-bits in hex mode) as the previous version
- Update M68k-local LIT config
If you're referring to this script as a whole, it is used extensively in M68k's test suite.
This script is designed for printing certain section in textual format. So I'm afraid your llvm-objcopy command might not be sufficient. As the comment below described, this script is similar to llvm-readobj but the latter only support one output format (i.e. hex with lots of redundant info like ASCii decoding), where in many cases, we want other formats like hex string with different width and textual bits.
This LGTM now, thanks!
I agree with Simon that using the default CHECK in FileCheck is more common and easier to read than x00-NEXT.
However, I imagine this is like that because there would be other tests, for other sub-architectures (x10, x20, ...) that would be tested on the same file and therefore would need special CHECK lines on their own.
If that's true, than it should be fine to keep it like this for now, even if there's only one. Otherwise, you'd have to regex-replace all files now, and then later again, which is wasteful.
I see very little by way of testing the return side of the calling convention, and nothing for sret.
Especially not one that's got lowercase letters in it...
You don't mean that...
Do you need uwtable on any of these?