Add opt test

Update to move half float to inline instead of library call as compiler-rt doesn't currently support divhc3. This may be added at some point as this currently causes errors for clang when libgcc is not available, see this issue: https://github.com/llvm/llvm-project/issues/61914

Clang format

Update style to mach style guidelines

Fix case issue. As for tests for debug information being generated, these are already covered by various tests in flang/tests/Transforms, namely debug-line-table.fir, debug-line-table-existing.fir etc.
While the tests are invoked through fir-opt instead of a flang flag, it is the same pass that is verified to be run in the tests added so is covered.

Add common function addDebugKindInfo, update warning message

In D146814#4223967, @awarzynski wrote:

What's the overall design goal here? 100% consistency with Clang? Could this be documented?

Add test for invalid debug-info-kind and address other review comments

Remove references to non existing import

Update tests to check for warning at higher debug levels, as well as amend warning message

Add test to check that debug pass is run and not run with appropriate flags

Attempt 2 to add previous work as stacked patch

Clang format

Add dependence on D142347 and update references to clang::codegenoptions to llvm::codegenoptions

Rebase

Setting this revision as base for adding debug flag to avoid adding additional dependencies on clang.

LGTM, but someone else should approve. Only comment is that the fir test is currently called loop-versioing.fir. Should probably be loop-versioning.fir?

Add test to check for generating library call or generating inline code

Formatting

Remove change to scf

Remove redundant getNumRegionInvocations from scf.if and fir.if

Clang format has changed in between patches, appease it

Amend test name, add RecursiveMemoryEffects trait and limit fir.if to 1 else region

Add scf.if test back

Cleanup changes

LGTM

Appease clang format

Change findBoxDef and findMaskDef to use a common method findDef

Update fir tests to match new generated code and change naming scheme to avoid confusing logical ranks

Fix an issue where input arrays containing only huge(i) would return 0 instead of 1.

Style changes and some small improvements

Minor bug fix to avoid using uninitialised variables in for loops

In D142977#4119152, @vdonaldson wrote:

The following code has a regression. Output from this code is now T; the output should be F.

integer z(0,5)        ! integer z(0) is ok
print*, any(z .ne. 1) ! should be false
end

F18 clause 16.9.13 ANY (MASK) or ANY (MASK, DIM), paragraph 5, Result Value Case(i) addresses the case of a zero-size argument:

The result of ANY (MASK) has the value true if any element of MASK is true and has the value false if no elements are true or if MASK has size zero.

The test is ok if the declaration of array z is changed to z(0). The code may need a more general check for the array size.

Fix Typo

Change to add proper support for different kinds of logical

Fix failing tests by changing two lines

Change genBody for All to check for zero rather than one

Minor change to fir test names to make them unique

Refactor to reduce code copy-paste, other minor changes

Fix test/lower/array-derived.f90 fir test and minor changes to style

Fix minor issue with tests.

Add test and comment

Remove uneeded mlir::twine{}

Fix zero64/one64 values being swapped around

Change naming scheme to not include fast math strings in logical functions

Change to test for 0 in simplified count rather than add logicals

Minor changes to formating and documentation