LGTM, thanks.

Thanks for patch! It will be great to have this setup available.

I moved the logic for library functions to inferLibFuncAttributes, added a note to TargetLibraryInfo.def and updated the tests

@uenoku Sure. I'm working on moving the logic for library functions to inferLibFuncAttributes, but I can add that in later.

Rebased

Going to rebase this.

Hi. I'm seeing a cmake failure now, it looks like the last parenthesis was dropped. LIBDIR})"), LIBDIR})"

Thanks for the review!

I would definitely like to be able to specify a specific subset of the test-suite, however I'd try to avoid an additional control macro if we can extend the current ones.

Added additional tests and addressed the inline comments

@xbolva00
I ended up not adding a helper function after looking at the existing ones because it won't simplify the code. To follow the current pattern for the helper functions:

Fixed indentation, added a test and expanded the summary to explain the difference between early-cse and early-cse-memssa.

Thanks for working on this! LGTM

Rebased and updated to use new name for WriteOnly attribute.

Rebased. Updated wording in documentation

LGTM. Thanks for fixing this

I've added some additional test including semantic tests.

Unified the blocks that add WriteOnly and Const attributes

Updated documentation

Fixed typos and added assertion

Avoiding two dyn_cast.

Sorry about the slow turn around. I've updated the documentation to be clearer about this attribute's use.

We can't rely on the stdlib rand() when verifying. You can include an implementation to use (eg MultiSource/Benchmarks/DOE-ProxyApps-C/miniAMR/glibc_compat_rand) so the output can be verified.

I had removed the periods after getting a FP Comparison failed, not a numeric difference between '.' and '.' " error. They were not the issue and I have added them back in.

Added a definition that will prevent the benchmark library from building cxx03_test which will fail when using c++11.