[llvm] [llvm-mca] Abort on parse error without -skip-unsupported-instructions (PR #90474)

Peter Waller via llvm-commits llvm-commits at lists.llvm.org
Mon Apr 29 09:44:13 PDT 2024


peterwaller-arm wrote:

OK, I propose to:

* make llvm-mca report an error and exit by default
* make `-skip-unsupported-instructions` take `no-sched-info`, `parse-error` and `all`.
* `parse-error` would skip the case `Parser->Run(false)` returning true (indicating error).
* Add a test case to cover your suggestion 'I also wonder if it would be hygienic to have a test case where you run it on multiple mcpu but one CPU cannot accept that program.'

I just want to clarify what you're asking there; it's that this should test the thing that the exynos case currently hits: we have an instruction where some mcpu can accept the instruction, and others can't, and all have `-skip-unsupported-instructions=parse-error` set, and it checks that the total cycle count is as expected? I would derive such a test case from the existing exynos case.


https://github.com/llvm/llvm-project/pull/90474


More information about the llvm-commits mailing list