Split from the original D129809.
Details
Diff Detail
- Repository
- rG LLVM Github Monorepo
Event Timeline
mlir/test/Dialect/Math/canonicalize_ipowi.mlir | ||
---|---|---|
440 | why are these tests so complicated? why do you need memref in here? |
mlir/test/Dialect/Math/canonicalize_ipowi.mlir | ||
---|---|---|
440 | Thank you for looking into it, @Mogball! Basically, I wanted to test multiple folding cases in one function, so I had to make use of the ipowi in some way (so that it does not get removed completely). I made the uses by storing the results into their own cells of the array passed as a pointer to the function. I guess I can just make the function return 11 results and get rid of the store operations. Or I can also split the function into 11 functions. Please let me know which of the options you prefer. |
mlir/test/Dialect/Math/canonicalize_ipowi.mlir | ||
---|---|---|
440 | The second one. The folder does look complicated. Do these tests cover each set of cases once? |
mlir/test/Dialect/Math/canonicalize_ipowi.mlir | ||
---|---|---|
440 | Yes, I tried to test each line of code at least once. For some of the cases I added more than one test, e.g. for positive powers I test negative and positive base values, whereas the folding code does not care about the base sign. |
mlir/test/Dialect/Math/canonicalize_ipowi.mlir | ||
---|---|---|
440 | SG. Please split these up and document which case you are testing with each. |
why are these tests so complicated? why do you need memref in here?