[flang-commits] [flang] [flang][OpenMP] Allocate `reduction` init temps on the stack for GPUs (PR #146667)

[Tom Eccles via flang-commits]

https://github.com/tblah approved this pull request.

This looks okay to me, but you might need to do further work in OpenMPToLLVMIRConversion to be sure these allocas don't end up inlined inside of loops. This is complicated because we don't want to mix allocas and code in general so we can't just inline the init region at the allocaIP.

https://github.com/llvm/llvm-project/pull/146667