Updated code through the use of newly introduced FindClauseMult() function, which returns an iterator from the multimap, in case where multiple occurrence of a clause happens within a directive.

OK, I will merge it on Monday unless @tskeith has any other comments.

In D97964#2604649, @kiranchandramohan wrote:

Thanks @arnamoy10 for this patch.

Did you try the other approach that I mentioned using FindClause? I am guessing FindClause will only return a single clause and hence cannot be used directly. Would adding another function (if a similar one does not exist) which returns multiple clauses (using equal_range) work? This way we do not need to use the clauseList.

Fixing build failures, also moved the pipeline check to an existing file.

Rebased on main

Taken out -module related changes, updated the commit message as well.

LGTM

Minor changes as per comments, also updated commit message to include description of changes related to Setters.

1. Updated the test case to include a check when the option is not given and make sure works with f18 as well (with Mstandard)
2. Changed variable names as suggested.

Addressing comments, by separating the search directories from fintrinsic-modules-path in a separate variables. Also added dummy modules for the test case, as using %llvmshlibdir did not work out.

Updating the test case to include -fc1 as well

Added a missing option test for -flarge-sizes

Set the option as a frontend-only option, also modified the test case accordingly.

• Rewritten the test cases
• Moved flarge-sizes to a standalone test file

Updated test case to check module content

Added some setters to work with.

• Added option for setting DOUBLE PRECISION when fdefault-real-8 is given
• Also added link to relevant gfortran doc

Addresses proper handling of sizes

Addressed latest reviewer's comments, no longer using a pointer for defaultKinds

Addressing reviewers' comments.

Sounds good to me, thanks for the help. In the meantime I will work on the fdefault-* family of flags, which will be dependent on this patch I think.

A few more comments addressed and a new test case added for write-module check.

Addressing reviewers' comments, adding -module-dir as the default option instead of -J

Added test for the driver-help, also contains all the changes that was done in the previous diff (Aliasing -J and -module-dir, changing data structures etc).

Addressing reviewers' comments with the following changes:

Rebased to the latest master of LLVM

@lebedev.ri Is there any instructions on how to land this? Sorry I am doing it for the first time

Nullptr check change

Creating a patch with more context

Adding more context to the diff.

Using WeakVH instead of Instruction to take care of dangling pointers. Had to replace the "insert" calls of SetVector with a SmallVector push_back

Thanks for the suggestion, as per your suggestion, I did the modification in the deleteDeadInstructions() function as follows:

Here is a diff for the content of DeadInsts just before and after a speculatePHINodeLoads()