[Libomptarget] Use new tripcount argument in the runtime.

[jhuber6]

Summary

The previous patch added an argument to the __tgt_target_kernel
runtime function which includes the tripcount used for the loop clause.
This was originally passed in via the __kmpc_push_target_tripcount
function. Now we move this logic to the kernel launch itself and remove
the need for the push function.

Depends on D128816

[jhuber6]

Rebasing and making special tripcuont -1 for the legacy API to use. If we have this tripcount we don't set it in target and instead expect it to be set by __kmpc_push_target_tripcount

[jhuber6]

Adding a static check on the struct size.

[jdoerfert]

openmp/libomptarget/src/interface.cpp

385

Doesn't target already do this? And we should get rid of this in a follow up.

[jhuber6]

openmp/libomptarget/src/interface.cpp

385

Oh yeah, I did it here and forgot to delete it after I decided to move it. I'll get rid of it. And why do we get rid of it in a follow-up?

[jdoerfert]

openmp/libomptarget/src/interface.cpp

385

we don't need this mutex and stuff anymore. Now the information is passed right into the target launch and not separately. See below.

openmp/libomptarget/src/omptarget.cpp

1556

And then delete everything that has to do with Device.LoopTripCnt

[jdoerfert]

OK, keep it till we branched 15 but don't use it when we come from the new interface.

[jhuber6]

Updating

[jdoerfert]

LG, one typo, see below.

openmp/libomptarget/src/omptarget.cpp

1514